A Tour of Targeting

The general principles of targeting followed by a tour of the available options on the major platforms

This article is part of my Entreprenerd: Marketing for Programmers book, which is currently available to read for free online.

Perhaps the most basic goal of any advertisement strategy is informing potential customers that you exist. But not every Tom, Dick, and Harry out there is a potential customer for your particular business, so it’s important to direct your advertising towards only the most relevant and promising of leads. This is the art of targeting. To wit: The solar panel seller who takes out advertising in a renewable energy monthly magazine is sure to win far more sales than the competitor who spends the same advertising budget driving around housing estates and announcing the solar panels with a megaphone.

Good targeting saves you money because you don’t waste your advertising budget on people who have no interest in what you are selling. Good targeting respects the public’s time because they will be subjected to fewer messages about irrelevant products and services. Good targeting treats the public more humanely because you are less likely to bombard the infertile and the childless with advertisements for nappies and children’s swimming lessons. Good targeting even helps the platforms you advertise with (such as Google AdWords) because their users are more likely to click on ads and thereby generate revenue for the platform. And because advertising platforms appreciate when you make them money, they reward you by further reducing advertising costs. (See, for example, Google AdWord’s Quality Score).

I believe your choice of targeting mechanism is the most important contributor to your online advertising success, more so than the advert creative. Finding the right targeting combination is a bit like panning for gold—the best advice is to try out all sorts of weird and wacky ideas until you stumble upon a keeper.

The modern interface of an online advertising platform is a fighter jet cockpit filled with a bewildering array of controls for honing in on your potential customers. These controls are mostly filters that limit your advertising to only show in certain geographical locations at certain times of the day, to people who fall into certain demographics or exhibit certain behaviours detectable by the advertising platform.

Knowing how to zone in on your ideal customers from the cockpit of one platform is no guarantee that you’ll be able to locate them when piloting another. Each platform is equipped with differing controls, so you may need to test drive a few different models until you find a platform that adequately meets your needs.

Indeed, there is already some implicit targeting going on as soon as you choose to advertise with platform X rather than platform Y. Already, the universe of people who might see your adverts has been narrowed down to “people who use the internet AND happen to congregate on Platform X”. If the demographics of Platform X are less than well matched with those of your product’s market, this could be quite the unwelcome restriction.

The most fundamental thing that anyone studying targeting ought to learn is the inventory of targeting options available on each platform. For instance, suppose you have a business selling textbooks to medical students. If you know that Facebook lets you target by university degree, then your work is mostly done and you will do admirably well. If, on the other hand, because of inexperience with Facebook’s platform, you are unaware of this targeting option, then you will struggle with less-effective targeting options and flounder at the heels of better-informed competitors.

(Proviso: The second half of this chapter is devoted to giving you a high-level guided tour of what’s out there right now. But this guide is neither comprehensive nor guaranteed to stay up-to-date. You would do well to supplement it with an intuitive rule of thumb that applies whenever an advertising platform is built on the back of a regular website/social network (e.g., Facebook/Twitter): Expect that any data gathered from the website’s users will eventually reappear as a targeting option for advertisers. For instance, Facebook asks its members about their relationship statuses. This data is then fed to advertisers in the form of a targeting option.)

The gap between your intended audience (the people out there who would love your product and to whom you want to advertise) and your targeted audience (the people who actually see your advert as a consequence of your chosen targeting options) is a constant source of frustration and disappointment for advertisers. Ideally, you want both groups to be one and the same so that you advertise to all the right people—and not another soul. But that’s not going to happen. There will always be waste—sometimes more and sometimes less. Instead, we must live with calculated levels of imperfection. In some cases, it’ll be worth advertising to a group where only 1 person in 100 is in your intended audience. This will be so when you reckon that the costs (financial or otherwise) of advertising to this group are lower than the expected profits. The logic is the same as that of the blackjack player who counts her cards and places a bet whenever the odds are in her favour.

The brain of the marketer thinking about targeting is filled with Venn diagrams. Every additional targeting option that they apply adds another circle into the mix. The marketer then adjusts and moves these circles until the shaded area paints out a space as dense with potential customers and as sparse with waste as can be achieved.

General Principles

Off-target targeting

The controls of your cockpit are hardly those of a Rolls-Royce; the levers are rusty, the gears sometimes jam, and the numbers on the speedometer are off by a worrying amount.

There is no guarantee that your customers are as they present themselves on Facebook or Twitter or any other website. And because most advertising platforms base their targeting on data their users input, we advertisers ought to stay vigilant about the difference between how our customers are in actuality and how they appear within the house of mirrors that is the advertising platform.

The sources of inaccuracy are many: privacy concerns, public posturing, laziness, or staleness (e.g., someone’s circumstances change but they don’t update Facebook). Facebook has this author on file as a 90-year-old man who is a fan of trashy reality TV shows, but only one of these two “facts” is true.

I don’t want to exaggerate the problem of dodgy data because a sizeable proportion of internet users are upfront about themselves. Nevertheless, it’s still worth pausing to think about how your targeting controls might be unreliable.

Dressing for the occasion

Once we’ve honed our adverts in on the people most likely to become future customers, our targeting tools help us in a different way: They divide our targeted group into hyper-targeted subgroups which each receive marketing messages tailored just for them. The car manufacturer might determine that people in their 30s value safety and space for children above all else, whereas people in their 20s and are more interested in ecological footprint and electronic gadgetry. By emphasising the right benefits to the right groups, the manufacturer tells everyone exactly what they need to hear before they buy.

This sort of advertising segmentation is akin to owning multiple sets of clothes—for example, a conservative suit for job interviews and outrageous hipster attire for music festivals. Wear the wrong clothes to the wrong occasion and you won’t have a good time.

The AND gotcha—avoiding unnecessary targeting

Because it’s rare for a single targeting option to sufficiently hone in on the intended audience, we often layer options together so as to combine their effects. On most platforms, each layered targeting option is related to the others with an implicit AND logical operator. If you aren’t aware of what this implies, you risk putting a low ceiling over your potential reach.

Suppose a business wished to reach UK-based law students to sell textbooks written in English. They conjure up an image of their typical customer, and, based on this image, they set the following targeting options:


University Country UK
University Degree Law
Age 18-28
Language English
——————- ———

Because these targeting options relate to one another with an AND operator, advertising will only be shown to people who simultaneously match all four criteria. But this high hurdle will overly restrict the reach of these adverts. Plenty of people in this business’s intended audience (UK-based law students) may fall outside of the targeted audience, as defined by these targeting filters.

  • Just because someone has their Facebook account set to a language other than English doesn’t mean that they don’t speak English. Indeed, the fact that a person studies law in the UK (as required by the other layers of the targeting) indicates that they almost definitely speak the language proficiently. The inclusion of this targeting option only served to exclude foreign students.

  • Just because someone is over 28 doesn’t mean that they don’t study law. As it stands, this business is completely ignoring mature students—and these are the ones with the most money to spend.

This business would have been better off thinking about what criteria are implied by others, then throwing off whatever turned out to be superfluous.

Let’s look at another example. Our law textbook seller has been trying to advertise to law students on Facebook for months but is increasingly frustrated because many people leave their degree subjects blank. The seller then has the idea that anyone who likes the Facebook page of a university law society is probably also a law student, regardless of whether this person registered their degree with Facebook. The seller expresses this realisation through the following options:


University Degree Law
Likes An exhaustive list of law societies
—————— ————————————

For the same reasons as those given in the previous example, these targeting options leave the business short-changed in its reach. The implicit AND between the options limits advert reach to people who both registered law as their degree and also liked a law society. Because the intersection of these two groups contains fewer members than the group of people who just list law as their degree, this business’s advertising reach would be even worse than it was before they decided to target law societies.

What the business really wanted was an OR between their targeting criteria such that anyone who EITHER studied law or liked a law society would be delivered the advert. Some platforms let you sandwich your targeting options with OR relations. When this possibility isn’t present, you can achieve the same effect by splitting your adverts into two side-by-side ad groups that differ in their targeting options (one will target law society “likers” and the other, law students).

Oblique targeting

The most obvious targeting configurations are often oversubscribed. When pursuing these advertising avenues, expect stiff competition from other businesses with bigger budgets and deeper pockets than yours. In hotter markets, heavy competition may completely price new contenders out of the mainstream targeting options.

Dissatisfaction with typical targeting options can also occur in the total absence of competition. It’s possible that you fully exhaust your leads from straightforward options, then find yourself in dire need of a fresh stream of customers.

This is the kind of environment that rewards the creative thinker. Oblique targeting is all about unearthing new pockets of leads in cobwebby corners or down grassy cul-de-sacs. If priced out of targeting law students directly (say by targeting university degree), how else can the marketer reach that same audience? This requires that the marketer imagine what it’s like to be a member of that audience: What magazines, books, and blogs do law students read? What movies do they watch? What institutions are they members of? What tangential interests do they tend to have? Eventually the marketer will shout “Eureka” at having found some clever way to target by configuring some oddball combination of controls in their advertising cockpit. For example, the marketer realises that law students often engage in a special form of debating just for lawyers-in-training (“mooting”), and this realisation can be leveraged by targeting anything and everything related to it.

This kind of oblique targeting may not be as cost-effective as targeting directly, but it might nevertheless be profitable enough to be worth doing.

Negative targeting

Some sets are easier to define in terms of what they don’t contain instead of what they do contain. When someone loses the “Q” tile in Scrabble, they don’t complain that they only have “9 A tiles, 2 B tiles, 2 C tiles…” remaining. They simply say, “The Q is gone”.

Negative targeting is a bit like this. Instead of defining who your intended audience are, you define who they are not. This is awfully useful.

For one, negative targeting is used to describe sets of customers that are impossible to describe with the available positive targeting options. If there is no option to target people who are “single”, you could reach much of the same group by negatively targeting people who are “married”. If there is no option to target people who don’t watch TV, you could (theoretically) negatively target those that show no signs of interest in TV or TV shows.

In other situations, negative targeting can be used to ease the administrative entry of targeting configurations that would be awkward to describe in the positive. Without negative targeting, anyone who wants to target people living outside the USA would have to spend an hour in the advertising editor exhaustively listing every country other than the USA.

Another use-case for negative targeting is to fine-tune overeager positive targeting so as to better focus advertising budget on the underlying intended audience. In Google AdWords, negative keywords act like fussy editors who scrutinise and double-check targetings suggested by positive keywords. For example, a sweet shop owner who positively targets the (broad match) keyword “candy shop” will waste money on people looking up the 50 Cent rap song of the same name. By adding negative keywords related to the artist names or to the music domain (such as “50 Cent”, “Olivia”, “song”, “listen”, “youtube”, ”music”, etc.), the sweet shop owner streamlines her targeting and saves money.

How are you to figure out what negative targeting options will fine-tune your audience? After all, not everyone is up to speed on their pop culture references…. A good general strategy here would be to start out with a purely positive targeting advertising campaign, leave it running for a while, then study the performance reports provided by your platform. Assuming the data is detailed enough, you’ll notice that some subgroups have atrociously bad click-through rates; these you trim out of future targeting through negative targeting.1

Negative targeting isn’t available on every platform yet; let’s hope it gets rolled out more generally soon.

The edge

Keeping with the general societal trend of technological acceleration, advertising platforms are constantly outfitting themselves with brand new targeting possibilities. Not only this, but brand new platforms are forever arriving on the scene offering fresh opportunities for the bold. All of this novelty is a playground of potential for the intrepid advertiser quick on her feet.

Whenever a new targeting possibility is released to the world, it begins its life alone, massively underrated and undersubscribed by other marketers. This often means deliriously good deals for the nimble early adopters who show up at the great big auction house of online advertising. If you are quick to familiarise yourself with the latest batch of possibilities, you’ll gain an edge—at least until your competitors crowd in after you.

In fact, it’s important to stay up-to-date on targeting developments no matter what. Even if you arrive too late to the auction to avail of the bargain prices, it may still be worth your while to pay full price for use of a newly released targeting option that identifies your intended audience more finely than any previous technology could.

In addition to all this, there are special benefits accruing to advertisers willing to test the waters with newly launched platforms. The adverts shown on these platforms are just as new to the platform’s regular users as they are to the advertisers paying for them. As a result, the platform’s users haven’t yet developed platform-specific ad-avoidance behaviour, which makes them (temporarily) more responsive to adverts. What’s more, deep-pocketed corporate competition will be dissuaded from advertising during a platform’s earlier days because it probably won’t launch with the APIs and other tools that heavy buyers need and expect.

The inflationary effect of competition on advertising costs has been remarkable. During my first three years advertising with Google AdWords, I saw my cost per click rise by close to 300%. After this happened, I diverted much of my budget to Facebook advertising, which was then available at prices significantly less than what I was paying Google AdWords. Eventually though, Facebook advertising costs ballooned and stopped being the bargain they once were. Today, these advertising platforms are, at best, modestly profitable for my business, whereas they were once magnificently so. The moral of the story: The effort required to stumble through early, unpolished, and under-documented new advertising platforms is justified by the heavily reduced costs.

Synonym bombing

As mentioned earlier, much of the data that underlies targeting options originates from self-reported factoids given by users on the consumer-facing side of advertising platforms. When this data comes from options users select from pruned drop-down menus, this poses no problems. But complications arise if it comes from user fields that accept unrestricted input, such as text boxes.

Suppose that you are a seller of cat food and you wish to advertise on a platform where users list their likes in whatever ways they please. The relevant data, when re-appearing as a targeting option, won’t be packaged into neatly standardised little semantic parcels. Some people will have liked “cat food”; others the differently worded “cat feeding”; or the more specific “food for Siamese kittens”; or a brand of cat food such as “WHISKAS”; or a blog intended for cat owners, like “The Reluctant Cat Owner’s Journal”; or the idea of bargains, as in “discount cat food”.

The infinite granularity of expression afforded by unrestricted input has its pros and cons for targeting. On the plus side, advertisers won’t be pigeonholed into the advertising platform’s preconceptions about what should be available to target. This gives advertisers the power to go after impossibly small niches (e.g., fans of Hodor from Game of Thrones), or to respond quickly to novelty (e.g., a band that just made its first public appearance). On the down side though, granularity creates work. Advertisers won’t maximise their reach unless they target the various ways their audience refers to whatever core concept is being targeted. I call this “synonym bombing”. Although the technique isn’t named this way elsewhere, it’s still an officially sanctioned practice. For example, LinkedIn advises advertisers targeting by job title that, “If you’re looking for an Account Manager, also try targeting a Senior Account Manager, Junior Account Manager, Director of Accounts, Account Strategist, and so on.”

Undersubscribed targeting groups

Just because a targeting option perfectly fitted to your business exists doesn’t mean that it’ll be worth a sod for your business. Presence of a targeting option doesn’t guarantee presence of any targets. The group may well be undersubscribed. For example, Twitter’s behaviour targeting gives advertisers the power to display adverts to American homeowners whose properties are valued in certain ranges (e.g., “under $99,000”, “above $500,000”, etc.). The total number of Twitter users in every property value band combined comes to 13 million people. But Twitter has 65 million users in America.2 Given that about 67% of Americans own homes3, the math implies that Twitter is missing home value information for 4 in 5 of its users. For many advertisers, under-subscription of this degree will fatally dilute the targeting option’s power. Cruder alternatives, like demographic and education information, may turn out to be more discerning.

The sum is sometimes more than the parts

As we saw in the last section, advertising platforms don’t have complete information about their users. This problem occurs again and again—not all companies publish their workforce sizes on LinkedIn, nor do all Facebookers volunteer their relationship statuses. This leaves large gaps of unspecified data across many different targeting dimensions. Hopeful advertisers who target a specific company size on LinkedIn (10–100 employees) or relationship status on Facebook (“single”) will therefore fail to reach people who are in reality a good match but cannot be identified through targeting. The fallout can be substantial—sometimes even the majority of platform members.

Many platforms have a special targeting setting for reaching the “unspecified”—those who fall between the gaps. The “unspecified” are worth advertising to if you have reason to believe that a satisfactory proportion of them are indeed part of your intended audience. For example, if you wish to reach unmarried people under 24, it may be worth targeting not just those whose relationship status is “single”, but also those whose status is “unspecified”. This is because it’s reasonable to assume that young people are mostly single anyway, and that this fact will hold within the unspecified group—perhaps even especially so since those who are married are likely to broadcast this fact.

On the flip side, you are unlikely to have much luck efficiently reaching niche sub-groups (e.g., “open relationship”) through targeting “unspecified”. Those in open relationships only make up a tiny proportion of the underlying population, so you shouldn’t expect them to appear with any worthwhile frequency in the “unspecified” group, at least not without the assistance of additional targeting layers.

Research advertising ease before committing

Before committing to any new business venture, there’s great value in knowing whether you’ll be able to meaningfully market through paid advertising. When available in sufficient volume and at tolerable prices, paid advertising makes launching a new business a thousand times easier.

Many advertising platforms give you estimates of how many people a given combination of targeting options ought to reach, and at what expected cost. By poking around in here, you get a feel for whether a proposed business will be easy, difficult, or impossible to advertise.

A Tour of Targeting Options

Let’s take a guided tour of the main targeting options on the market right now. Once you’re done with this section, you’ll know your way around the modern territory as well as an old pro.

Geographical location

Most advertising platforms let you target countries, counties, cities, and postcodes; some even let you draw an N-kilometre radius around a chosen epicentre.

These geographic targeting features are most obviously of promise for businesses with a physical presence. There is a limit to how far people will travel to buy something, and advertising outside these limits is wasteful.

Not many people in small-town Spain will buy a guide to negotiating with a New York landlord, nor will a New Yorker buy a kitchen oven designed for the Spanish power system. Geographic targeting helps advertisers of region-bound products to restrict their adverts to the appropriate geographical bounds.

Residents of rich countries have larger disposable incomes that those of poorer countries. When you advertise to the whole world, you might discover that 90% of your advertising budget gets spent in India, yet you don’t get a single sale there. This isn’t a fictional example; it actually happened to me.

The concept of geographic location seems self-explanatory, but there’s a hidden gotcha: It doesn’t always make sense to classify someone as present in one location. Imagine a student who spends their weekdays at Oxford, their weekends with their partner in London, and their summer holidays in their hometown in China. No matter which of these three the student lists as their location, they will fail to capture the full reality.

Some platforms calculate the location from the user’s IP address or network activity. This gives us a concept of “current location”. But current location isn’t necessarily the key consideration. While some businesses (e.g., local restaurants) do want to target people currently nearby, others are more interested in people who live in some place or other (e.g., house painters). The fact that someone is googling for painters while holidaying in France doesn’t change the fact that they want their house in the UK painted.4

As we can see, which location is relevant depends upon the advertiser’s specific motives. For this reason, it’s sometimes best to completely remove location targeting whenever you’ve applied some other targeting dimension that implicitly restricts location. For example, if you wish to reach “law graduates in Oxford”, it’s enough to just target people who graduated any of the law universities in Oxford. An additional layer of location targeting would be superfluous.

As with all other targeting options, there’s the potential for geographical targeting to help you reach your intended audience through oblique and convoluted means. For instance, Google AdWords doesn’t let advertisers directly target employees of a particular company, but there’s nothing stopping an advertiser from limiting their advertising to the smallest allowable radius around their target company’s offices.

Time

Every online advertising platform targets by time in some form. The lowest common denominator of this is the power to manually switch a campaign on or off.

We most typically think of seasonality as something measured on the scale of winters and summers, and it’s true that most of your efforts in targeting time will be spent adapting your adverts to these macro trends. If you sell fancy dress costumes, you’d be mad not to tailor your campaigns for Halloween, Christmas, and any of the other major holidays. If you sell ice cream, you’d be mad not to freeze your campaigns in winter.

Modern online advertising tools have brought us to the age of micro-seasonality, an age where the peaks and troughs of demand can be measured by day of week or even time of day. It would hardly be surprising if there were a correlation between night owls and potential customers of sleep-aid products. Neither would it be a creative stretch to suggest that work-related products would be more effectively advertised during office hours or that products intended for students would remain largely unseen on a Friday night.

How do you figure out these correlations? Through a combination of Analytics and detailed reports from previous advertising campaigns.

Gender

Some products are bought disproportionately by members of certain genders. To name a few: boxing shorts, tampons, bras, and protein powders. Targeting by gender can help you better zone in on your likely audiences. If you know that 97% of people buying your product are men, then advertising it to women doesn’t make much business sense.

There’s no need to go crazy with gender targeting though; in the presence of more discerning targeting options, it may be completely unnecessary. Even if an advertiser knows that it’s mainly women who buy female birth control pills, it’s patently worth advertising the pill whenever someone who isn’t a woman expressly shows interest by googling prices for the pill. Maybe they are buying on behalf of their partner.

Only a handful of products are strictly for one gender or another; most merely show slight statistical differences in buying patterns which only justify proportional modifications to bid sizes for each gender (e.g., the advertiser might bump up bids 20% for women).

Sometimes you’ll need to vary your advert creatives so as not to alienate one gender or the other. The seller of disposable razors should know that men are more likely to shave their faces and women more likely to shave their legs. Not tailoring their images and copy may hurt the retailer’s conversion rates.

Age

For many of us, life unfolds in a predictable progression of stages. As kids, we are schooled. As young adults, we study or begin working. During this time, we move out from our family home and develop new interests. But with these new freedoms and abilities come new problems—we must now rent and maintain property, feed and insure ourselves, and earn enough money to support our lifestyles.

Age targeting is a crude targeting option that relies on these statistically typical paths through life. In a sense, its effectiveness is a barometer of cultural norms.

Age targeting is best used when combined with other options. Some time ago, I wanted to target undergraduate students. Back then, Facebook let you target members of a particular university but didn’t give you any way to target level of study. (It added this feature a few years later). Despite this, I was able to reach undergraduates (skipping postgrads and staff) by restricting the advertisements to only students who were under 23. (Yes, I would have missed mature students, but this was the trade-off I chose to make.)

The typical users congregating on some platforms are a good generation older than those on others. For this reason, advertisers need to factor in the implicit age targeting that occurs on each platform. When I was a teenager, I used a social network named Bebo that was hugely popular in my home country, Ireland. I left for Facebook when I went to university, but Bebo continued for years afterwards to be thronged with early teens. By contrast, my godmother—nearly 30 years my senior—has only ever signed up to one social network: LinkedIn. As a woman working for a British corporation, her choice of social network is a professional necessity. Aged-related platform usage patterns go beyond social networks. My dad searches the internet with Bing because it was the default search engine installed on his Windows machine, but no one reading this book would dream of using anything other than Google.

Device/computer/browsing technology

The advertisers most grateful for this targeting option will no doubt be the software vendors whose products are exclusively designed for some particular piece of technology. Why should they waste money advertising Mac-only applications to Windows users or Google Chrome plugins to people who browse with Safari?

Even if a software vendor sells a cross-platform product, these targeting options remain important for directing users to the landing page that matches each user’s tech stack.

Depending on what device someone is browsing with, they are either more or less in the mood to purchase. Consequently, emphasising or de-emphasising adverts on certain devices (phones vs iPads vs laptops) can affect revenue in surprisingly large ways. At time of writing, customers are unlikely to complete long buying processes involving credit cards from their mobile phones, so many ecommerce businesses decrease their mobile advert bids. But other companies, such as call-out taxi services, would leave their mobile bids as high as their laptop bids, if not higher.

Interests ( / Likes)

Interest targeting is available in many different guises across the various platforms, so it’s best to begin by sketching out the important strands which distinguish one manifestation from the next.

Recency

People’s interests change over time, so each platform’s way of categorising someone as “interested” matters a lot. Just because you were into mortgages six months ago doesn’t mean you are today; once you’ve bought the house, you won’t give financing another thought. Similarly, just because teenage-you liked a band on Facebook doesn’t mean that your tastes as an adult haven’t since matured.

But your history of Facebook likes only gets longer. And the interests Google guesses you have (may) stay frozen in time.5 Theoretically, the perfect advertising consumer would systematically prune their likes and meticulously register and deregister their interest elsewhere. But no one, at least no one remotely normal, does that.

As a result of this danger, advertisers targeting by interest should watch out for targeting algorithms that display adverts to people whose interests have grown stale. Your suspicions should be raised when the algorithm relies on data that could be months or even years old. By contrast, some algorithms are inherently current by design. For example, advertisers on Reddit who target by whatever subreddit someone is currently browsing are likely assured of engaged readers.

Private declarations

Some platforms harvest interest information from public user activity (e.g., timelines in Facebook and Twitter), whereas others pick up the info from private sources (e.g., browsing histories on Reddit or Google). This distinction matters because people skew their interests when they are in public.

For one, there’s often an element of embarrassment or shame in expressing your true concerns. Not many people go to Facebook and like “cures for back pain”, “ending heroin addiction”, “how to meet the love of my life”, or “pros and cons of telling religious parents you are gay”. This means that the advertiser targeting by (public) interest will have a tough time reaching their intended audience. Instead, they will probably have to look for indirect routes, ones that rely on socially acceptable public activity trails. Perhaps the person with a bad back likes a brand of chairs known for their excellent spinal support. Or perhaps the recovering drug addict follows the author of a personal empowerment book that comes highly recommended within the community of recovering addicts but is rarely mentioned elsewhere.

Social pressures and duties are another source of inaccuracy in publicly stated interests. People like things on Facebook (or tweet them) because of social pressure to conform, or social signalling, or the want to publicly support a friend’s new venture which otherwise holds no special interest, or amusement at something said by a company (as opposed to appreciation of the goods they sell).

Granularity

In general, the more you can narrow down your targeted audience to match your intended audience, the better. The cat food retailer should prefer the advertising platform where they can target people interested in “cats” over the one that only lets them target people interested in “pets”. This saves the retailer the wasteful expense of advertising to dog owners and fish owners, who won’t have the slightest interest in cat food.

“Similar to”

Some platforms automatically calculate lists of people that they believe have similar interests to the people you are explicitly targeting. Frankly, as of early 2016, I have not once advertised with any platform’s “similar audiences” feature without regretting it. But that’s just my experience; you might have better luck.

Many platforms give advertisers more than one mechanism for targeting interests. By comparing these using the factors we just discussed, you’ll be better equipped to choose the one best suited for your needs. For example, both Facebook and Twitter have concepts of precise interests and broader predefined interest categories. These differ only in granularity. Precise interests are ultra-granular (e.g., specific football teams, bands, etc.), whereas the predefined categories are more general (e.g., the concept of “football” or “rock music”).

Let’s close by looking at a sneaky little tactic: Advertisers who have well-known, established competitors can piggyback off their rivals’ previous followings. The luxury handbag startup will target Prada lovers, just as the soft drink upstart will target Coca-Cola lovers. Be warned that this tactic may be illegal (check with a lawyer). And even if it is legal, it can sometimes backfire for another reason: Your competitors’ fans may well be duds. I know of household brand names that hire shady companies to inflate their like and follow counts by using bots and outsourced like-farms.

Keywords

Famously, advertisers on Google Adwords can target based on search queries. The great thing about this targeting method is its to-the-second recency. Whenever someone googles “buy sandwich toasters”, this, as you can imagine, is a damn good time to advertise that you’re selling.

Less well known is Google’s Display Network. Over two million websites have partnered up with Google (via their AdSense program) and agreed to rent out space on their pages in exchange for a cut of whatever Google earns by selling it to advertisers. Google is essentially a broker who automatically inserts ads whenever the on-page text of the partner’s site matches a targeted keyword. The reach is phenomenal, with Google claiming that it encompasses 90% of all internet users.6

Keyword targeting isn’t exclusively the domain of search giants. Twitter released their own keyword targeting based on the phrases recently tweeted or searched for through their platform. Should you happen to own a pizza delivery business, you could advertise to people who recently tweeted to their followers asking for a “pizza delivery” recommendation.

Keyword targeting should really be called keyword targetings. There is a whole family of differing “match types” which fine-tune when keywords count as being triggered. These keyword match types are (roughly) similar between the platforms, so I will focus on describing Google’s.

Broad match

The most eager of all match types. If you don’t set a match type when setting up your keyword targeting, Google sets it to broad match by default.

Broad match keywords match not just search queries that exactly match the original text entered, but also misspellings, different pluralisations, acronyms, abbreviations, synonyms, related searches, or searches that contain additional words. For example, targeting “sandwich grill” would cause the advert to display for

  • sandwich grills - plural

  • sandwich grilling - changed stem

  • sandwich gril - misspelled

  • grill sandwich - order changed

  • buy sandwich grill - additional word at the start

  • sandwich high-speed grill - additional words in the middle

  • sandwich grill for sale - additional words at the end

  • sandwich maker - synonym for one of the contained words

  • panini grill - synonyms for both of the contained words

Modified broad match

This is a less extreme version of the broad match, the only difference being that synonyms and related searches are no longer targeted. But other than this change, modified broad match behaves just like broad match does. Out of the list of matching search queries we gave in the section for board match, everything except the following two would continue triggering adverts:

  • sandwich maker

  • panini grill

The motivation behind modified broad match is to give advertisers more control whenever synonym and related-search matching trigger too many irrelevant variants and waste profitability—which is basically all the time.

My recommendation to advertisers is to begin with broad match keywords for a few weeks and gather data about how well each synonym/related search performed. After this initial research period, create modified broad match keywords for all your top performers, then remove the original broad match keywords so as to prevent their wastefulness.

Phrase match

This match type is tighter than the modified broad match in that it requires that the search query contain the keywords as a single unit (i.e., in the same order and without words interspersed in the middle). But phrase match still matches search queries that have minor variations of the keywords (e.g., pluralisations, misspellings).

As such, the only difference between this and the modified broad match would be that now advertisements won’t show for:

  • grill sandwich (because the order changed)

  • sandwich high-speed grill (Because a word was inserted in the middle of the phrase. By contrast, words added to the start or end wouldn’t cause any problems because the unity of the phrase is still intact.)

Exact match

This is the tightest match type, but it isn’t as tight as its name suggests. As with all the other match types, exact match still triggers advertisements whenever the search query contains the keyword with minor modifications (e.g., pluralisation, misspelling, altered stem). The difference between exact match and phrase match is that exact match has a zero-tolerance policy for additional words in the search query, no matter whether they appear before, after, or in the middle. This means that exact match won’t display adverts for the following searches (which were fine with phrase match):

  • buy sandwich grill (which showed ads with phrase match)

  • sandwich grill for sale (which showed ads with phrase match)

  • grill sandwich (which did not show ads with phrase match)

  • sandwich high-speed grill (which did not show ads with phrase match
    either)

Remember that there is also negative keyword matching that the marketer can use to prevent adverts from displaying. For example, if the retailer of a sandwich maker knew that a model could not prepare paninis, then she might negatively target the keyword “panini” while positively targeting broad match “sandwich grill”. This means her advertisements wouldn’t show for “panini grill”, but would for other synonyms like “toastie grill”.

Just as with interest targeting above, keyword targeting also opens up the possibility for you to piggyback off your competitors’ brand names/trademarks in the hopes of convincing their leads and customers to switch. Google specifically states that they will not investigate or restrict the use of trademark terms as keyword targets, even if they receive a trademark complaint.7 Despite Google’s hands-off approach, this practice is still in a legal grey area; you risk legal consequences if you choose to pursue it.

Your fans

Plenty of businesses have a social media presence which they use to build a following. Their efforts are motivated by the promise of free marketing—the possibility to reach all their hard-won fans without paying a further dime. In the early days of social media, this actually worked; marketers experienced a golden age where almost anything they posted on social media would reach their fans.

This was not to last. Nowadays, marketers would be lucky if even 10% of their fans saw their social media posts; such is the jump in competition for a user’s attention. It’s commonplace for Facebook users to have friend counts in the thousands, each of whom posts more content than ever thanks to the ubiquity of share buttons, powerful new mobile devices, and the continuing myth that photographs of cappuccinos are of any interest to anyone. What’s more, every Facebook member has been accumulating an ever-growing mass of group and page memberships for as long as they’ve been part of the network. Each of these pages continue churning out content in the struggle for attention, so now, instead of competing against 10 other pages, you’re competing against 100.

All profitable businesses are born out of scarcity, and this lesson hasn’t been lost on companies like Facebook, who are now capitalising on this increasingly hot commodity. In the manner of a trendy New York nightclub, advertisers can now tip the doorman and skip ahead of the penniless multitude vying for a chance to be seen. One such bribe is Facebook’s option to “boost” a page’s post, which basically gives a brand’s social media the same airtime as it would have gotten before the competition got steep. Or alternatively, you could target by “connections”, whereby businesses can target their regular adverts to existing fans of their company’s own pages, apps, and events.8

Connection targeting isn’t just limited to reaching people who liked your page/app/event. It optionally extends to their friends. This feature is worth enabling when you believe it likely that the interest you’re targeting will be shared by a significant proportion of a person’s (probably very extensive) social network circle. (Remember, this circle probably includes their family, childhood friends, university friends, one-night stands, work colleagues, etc.). Just because someone who is otherwise civilised has a shameful obsession with the Jersey Shore TV show doesn’t mean that more than a pair of their Facebook friends have the remotest interest in such things. On the other hand, I can imagine situations where the overlap of interests would be strong. For example, the average computer science student is likely to have a generous helping of friends they’ve met on their course who share similar interests.

Idealistic advertisers may scoff at the thought of paying to reach fans they had already won over organically. It feels desperate and dirty, like paying your brother to come to your own birthday party. But consider the pragmatic realities: Your fans—assuming they were legitimately earned—know and love your brand. They are the people in the world most likely to buy from you. Sure, you already invested heavily in winning their initial affection. But now that they love you, the marginal cost of converting them into paying customers is going to be far lower than the cost for fresh leads. Therefore, spending a little extra money on reaching these existing fans is not only rational, but optimal.

There’s another huge benefit to connection targeting: It leverages social proof. When you notice that your buddies are going gaga over an advert you’d otherwise ignore, that’s going to grab your attention and encourage you to give it a look. As such, Facebook is wise to slap a Like button on their adverts and highlight which friends appreciated it.

Other people’s fans

Twitter has a wonderful targeting option that lets advertisers target the followers of anyone on Twitter—be they competitors, enemies, celebrities, bots, politicians, or your company’s CEO. Other people’s Twitter accounts now become user lists for you to advertise to, opening up endless possibilities for creative targeting. This is the marketing equivalent of picking up a Mario Kart power-up.

The most obvious use-case would be targeting the followers of a competitor, thereby poaching some of their hard-earned followers. But even if you have no direct competitor to piggyback off of, you could achieve much the same advantage by targeting businesses that serve different niches of the same market (e.g., a fitness supplement manufacturer might target followers of a chain of gyms).

Another use-case, this one more for the branding-conscious advertiser, would be to target the followers of a thought leader relevant to their customers. For example, a trendy shoe designer might advertise their pairs to the followers of a street fashion guru.

Uploaded lists of customers/leads

Advertisement isn’t just about finding completely new customers—it’s also about reaching out to existing ones and enticing them with new offers.

Facebook, Google, and Twitter let advertisers upload lists of email addresses then target these people through their platforms. The obvious limitation is that the advertising platform has to have someone in its system who matches that email address. The key metric here is the “match rate”, and this varies quite a bit by platform. In 2015, Larry Kim did an experiment with 350,000 email addresses and found that Google and Facebook both matched about 50% of them, whereas Twitter only matched about 10%.9 Regardless of which platform you’re using, you’ll be able to nudge the match rate upwards if you keep all your customers’ old email addresses around instead of deleting them.

Businesses without email addresses on file for their customers won’t necessarily miss out on this party. Facebook lets advertisers upload other types of identifying information, such as phone numbers, Facebook user IDs, and mobile advertiser IDs.10

Even when emails are collected by a business, these alternative identification methods are still worth collecting because they help bump up the match rate even further.

Leaving aside potential regional-specific issues with data privacy laws and website policies, there’s nothing technically stopping an advertiser from procuring a list of customers from a third-party source. Two businesses wishing to reach similar audiences might thus agree to share their uploadable customer lists to mutual advantage.

Many business transactions still occur offline (e.g., in high street stores). This creates a lacuna for marketers wishing to advertise to their previous customers online. Luckily, uploaded customer lists can assist in smoothing over this problem. The trick is to unite a customer’s online and offline identities through something that both existences share. This something is usually an email address. Having gathered the email address at the store, the customer can then be targeted with follow-up online advertising.

Uploaded customer lists also play a role in negative targeting. For whatever reasons, some customers will be unlikely to buy again. This could happen, for example, if your website is dedicated to selling just a single book; once someone has bought it, what’s the point in continuing to advertise to them?

Finally, some platforms have features to find “similar audiences”. This is closely related to a targeting mechanism we already looked at, that of similar interests.

Snipering

So far, we’ve spoken as if uploading lists of previous customers is the only way to leverage uploaded list targeting technologies, but this isn’t quite true. Advertisers who so desire can alternatively choose to compile lists of likely leads using publicly available information. For example, an advertiser could research a list of Twitter handles they wish to advertise to (e.g., @jackkinsella, @WhiteHouse), then upload this list through Twitter’s Tailored Audiences.

This is the online advertising equivalent of cold-emailing, and will no doubt be of great value to salespersons wishing to get their messages in front of decision makers they’ve already identified.

LinkedIn also has a way to achieve a similar effect, albeit on a smaller and more labour-intensive scale. It would be a stretch to call this advertising pure because it doesn’t involve regular advert creatives with pictures, headlines, and all that jazz. Instead, the marketer sends out a special (paid) LinkedIn email (called an “InMail”) to any LinkedIn member they want, even if the marketer doesn’t know that person or have any connections with them. This tactic is very useful to the recruiter struggling to fill a position or the accounting software salesperson hoping to reach a company’s CFO. If the CFO has no publicly available contact details elsewhere on the internet, InMail may be the only way to get through to her directly. This tactic risks bordering on spammy, so stay tasteful in your message and keep in mind any potential legal consequences.

Behaviour

Behaviour targeting is a messy hodgepodge of next-gen targeting options based not on information someone keys in about themselves on some social network or other, but rather their actions, as reported by offline sources like customer loyalty programs or online sources like purchasing activity or subscriptions. Here’s an example (from Twitter’s documentation) about how behaviour targeting has been applied:11

Nestle’s @Butterfinger leveraged partner audiences for the launch of its new product, Butterfinger Cup Minis. By targeting Promoted Tweets to Twitter users who frequently purchase peanut butter candy in grocery stores, convenience stores and other in-store destinations, @Butterfinger achieved a 52% lift in engagement rate compared to Nestle’s overall performance in 2014.

Behaviour targeting is available on Twitter and Facebook at present, although they don’t gather the data themselves—this comes through partner companies. At the moment, behaviour targeting is undersubscribed (i.e., data is only available for a fraction of users, so the reach won’t be what it should be). But as commercial transactions are increasingly digitalised in coming years, this number is sure to rise.

There are far too many different behavioural targeting options to list here, so I’ll just mention some of the most interesting examples on each platform. Be warned too that the available options depend on what country you are advertising into. This is because data availability through partner companies differs from country to country. Places with strong data-privacy rules, such as Germany, won’t have half the options available as a place with laxer rules, such as America.

Twitter

  • Creditworthiness

  • Yearly income

  • Home value

  • Liquid assets

  • Investment ownership

  • Net worth

  • Shoppers who spend in certain retail categories (e.g., “fitness” or “electronics”)

  • Shoppers who buy a particular brand (e.g., Reebok or Vans)

  • Donation habits (e.g., “animal charities” or “political liberal causes”)

  • Cat or dog ownership

  • Political party affiliation (US only…this one must be great for raising donations!)

  • Automobile ownership (e.g., how old someone’s car is and what brand they own)

  • Sporting interests (e.g., people likely to attend live soccer games, go camping, or play tennis)

Facebook

  • Expats living in a specific country (e.g., in Germany or China)

  • Travelling behaviour (e.g., frequent travellers, business travellers, commuters, users of travel apps, people currently travelling)

  • Facebook specific digital activities (e.g., people with histories of making Facebook payments, creating events, admiring Facebook pages, or playing Facebook games)

  • Other (non-Facebook) online activities (e.g., players of console or browser games, uploaders of photos)

YouTube

Through Google’s Display Network, businesses can direct their advertisements to appear on specific YouTube videos or channels. This exciting targeting option is useful for a range of marketing purposes such as promoting your YouTube channel to viewers of a rival channel, advertising your music instrument ecommerce store in videos that talk about synthesisers you stock, or promoting your jazz school on videos dedicated to teaching piano improvisation through video.

YouTube also has a video remarketing feature that enables video uploaders to target people who’ve “interacted” with their videos in the past. Possible interactions include watching videos, taking an “action” on a video (like, dislike, comment, share, etc.), or subscribing to the YouTube channel.

TV

Whenever a TV program airs its latest episode, fans buzz about it all over the web. From the flurry of activity on its own platform, Twitter can easily tell which of its users are engaged with which TV programs. Twitter then harvests all this info and packages it up into neat little TV targeting options.

TV targeting comes in two variants. The first, new airing targeting, restricts advertisements to displaying around a show’s airing window. New airing targeting must be contrasted with continuous targeting, which instead displays the promotional messages at any time during the advertising campaign—during TV show airing or otherwise.

TV targeting will be most useful to businesses already buying advertising space on TV and wishing to supplement their messages there with Twitter campaigns. I imagine too that businesses selling merchandise related to specific TV programs will also want to tinker with TV targeting.

Major world event

Whenever some newsworthy event is happening—a major conference, an election, or a movie release—you can be sure that people will talk about it on Twitter.

Unsurprisingly, advertisers want to insert themselves into the middle of these conversations. Even for companies that don’t sell products directly related to the event, timely and relevant wit can be remarkably effective at winning over the affections of the general public. For instance, during Super Bowl XLVII, a blackout struck and Oreo quickly quipped, “You can still dunk in the dark”. This tweet, despite being promotional, was famously well received.

Twitter has a special targeting option just for these moments. Marketers are given a pre-populated calendar of major upcoming events around which they can schedule their advertisements. Behind the scenes, this is powered primarily by tweet consumption and engagement data.

This feature appears to still be in an early state of development. As I write this, the number of events on offer is only a few hundred. This rather paltry selection is organised into the following categories

  • Conferences (e.g., Microsoft BUILD, ad:tech)

  • Sporting events (e.g., the Boston Marathon, the FA Cup Final, the Miami Open)

  • Entertainment events (e.g., the Billboard Music Awards, the Cannes Film Festival)

  • Movie releases (e.g., the latest X-Men movie)

  • Political events (e.g., Super Tuesday)

  • National holidays (e.g., St. Patrick’s Day, Mother’s Day)

Relationships and family

Targeting by relationship status will most obviously be important to advertisers in the dating space. Facebook lets advertisers target not only the classic denominations (e.g., “single” or “married”), but also a selection of narrower options including “divorced”, “open relationship”, and “long distance relationship”. These could come in handy for advertisers of niche services, such as software that facilitates couples watching synchronised Netflix videos together from afar.

If you want to create appropriately enticing creatives for dating services, you’ll want to know your audience’s sexual preferences. To this end, Facebook lets advertisers target by whichever sex a person registers they are “interested in”.

Freshly engaged couples often publicise their plans to marry on Facebook, and, as is to be expected, Facebook capitalises on this juicy titbit by letting advertisers target by engagement status. Wedding planners, photographers, and catering companies love this.

One thing leads to another in relationships, and sometimes that other thing is actually a person. Through Facebook’s knowledge of its members’ family relationships, it is able to provide advertisers with the option of targeting to parents who have children in certain age ranges (e.g., “parents of preschoolers”, “parents of pre-teens”, or “parents of teenagers”.

Users of your own app

App developers often make their money by convincing users of their free app to upgrade to a paid version or complete an in-app purchase. For this reason, it’s sometimes worth their while to advertise to their existing app users, even when these users have logged out of the app and are hanging out somewhere else on the internet. This functionality is no pipe dream—it’s available on both Google and Facebook.12

Let’s talk specifically about Facebook’s feature set, it having more developed functionality that Google. Before using it, you’ll need to do a mandatory bit of integration work to ready your app and establish a connection between it and Facebook. Once done, you are already able to advertise to your app users. But this might not be good enough, because all the users will be clumped together into an undifferentiated whole.

You can further refine your targeting by pinging Facebook with in-app events. These come in two varieties: custom and predefined. The predefined ones are designed to cover the basics, such as:

  • App install

  • App launch

  • Search

  • Add to cart

  • View item

  • Purchase

  • Complete registration

  • Complete tutorial

  • Level achieved

  • Achievement unlocked

  • Rated

Most event notifications also accept additional parameter(s). For example, the “search” event could also send along the query string; the “add to cart” event could lead to the cart value, line item, and currency; the “achievement unlocked” event could result in a description of the exact achievement. You can then filter by this information to target the most profitable leads.

If needs be, you can also create your own custom in-app events that capture whatever is most important to your business for segmenting and targeting users.

Through Facebook’s lookalike audiences feature, you can further your advertising reach beyond your original lists of app users. Facebook’s lookalike audiences are supposedly quite good when fed app user lists (one company claimed that it increased their app’s revenue six-fold). But of course, your mileage may vary.

There are some apps that only work when groups of friends have all downloaded it (e.g., messaging apps, work collaboration apps, or multiplayer games). These app developers will love Facebook’s option to target the friends of whomever has installed an app.

Users of other people’s apps

If you are launching a new app, there is no better group to advertise to than users of competing or similar apps.

Although this might seem hard to believe, Twitter collects and maintains a list of all the apps that its iOS and Android users have installed on their phones. They dress up this cheeky data grab as the “app graph” and use its contents to power their “installed app” targeting option. This lets advertisers target users who install apps that fall into certain predefined categories (e.g., business apps, finance apps, health & fitness apps, or reference apps).

Google has a different approach. Instead of only displaying ads to users of its own websites/apps (as Twitter does), Google syndicates its adverts to users of other apps, so long as those other apps are in cahoots with Google through its expansive Display Network. Impressively, Google lets advertisers target named apps (or “placements”, in their lingo). At last, you can rest assured that the fine users of Extreme Fart Recorder PRO will hear your promotional messages! Because specific app targeting happens through Google’s Display Network, these ads can be configured to appear on apps running on any platform (including iOS). Unfortunately, there’s a big limitation to Google’s otherwise compelling app targeting: The ads only show whenever a user is currently within that app, which may not be often at all.

Not everyone is willing to spend money paying for apps. In recognition of this, Google has added a targeting option that limits advertisements to only display to users who have previously purchased a paid app or made an in-app purchase from the Google Play Store.13

In addition to targeting specific named apps, advertisers on Google can also target broad (predefined) app categories, these being the same as those used on its app store.

Visitors of your own website (remarketing)

There are good reasons to advertise to people who’ve already visited your website.

First is the possibility of segmenting your audience so as to target only the visitors who have the highest likelihood of buying. This is done by filtering your visitors based on some action they performed on your website, such as adding a product to their cart or spending more than 20 minutes browsing. If your remarketing list consists of only the hottest of leads, you can confidently bid more per click when you need to beat out a competitor in the auction for advertising space.14

Second is the idea that advertisements only peak in effectiveness after being seen multiple times. By continuing to deliver your messages after someone has left your website, you bolster your brand and help transform mild interest into a compulsion to buy.

Third is the trick of making your business seem larger than it really is. By showing follow-up adverts on Facebook, Twitter, and (especially) the zillions of websites in Google’s Display Network, you create the impression of omnipresence. This inspires trust in conservative old customers who are scared of dealing with brands they haven’t heard of. (As internet users become more and more accustomed to remarketing, this effect will probably wear off.)

Configuration

The first step in getting started with remarketing is to integrate the advertising platform’s JS snippet/remarketing pixel into your website. Next is to update your privacy policy to alert visitors that they will now be subject to remarketing. This is necessary because remarketing has strong data privacy implications. Once that’s done, create skeleton remarketing lists as soon as possible, ideally at website launch. Delay in setting up these lists may leave you with significantly less leads to remarket to than you would have had if you’d gotten your act together earlier.15 The last bit of remarketing configuration consists of choosing the trigger conditions that will cause a visitor to be added to your lists. The usual default trigger is someone visiting a particular URL (e.g., the URL “/checkout”). Other possible trigger mechanisms are the visitor having spent so-and-so long on your website, or the presence of special custom variables which your website sends the advertising platform (e.g., you might have a “frequent-visitor” variable that gets sent whenever someone has been to your website five times in the last month).

It’s important to mention that you can choose how long someone stays on each remarketing list. Lists can have short or long stay times. Short times signal recent interest/engagement and are useful to advertisers who want to strike while the iron is hot. Longer stay times, by contrast, are more like capturing email addresses. These lists are useful for customer-lifetime marketing strategies, such as announcing sales or launching new services of interest to old customers. Often it’s worth having parallel short- and long-stay lists, with each list serving a different marketing end (e.g., you’d have one list for people that visited the checkout page in the last seven days and another for everyone that ever visited it).

Must-have starter lists:

  • All visitors - A handy catch-all list useful for general advertisements (e.g., “sale now on”, “introducing our newest service”).

  • Cart abandoners - Contains people who were on the verge of buying something but didn’t follow through. It’s worth advertising to these folk because you have a good chance of convincing them to go all the way and complete that purchase—especially if your remarketing campaign offers them a little discount.

  • Previous customers - In many businesses, past customers are likely to return and buy again. Ideally, the adverts you show them should emphasise products that are often bought together with whatever that past customer already ordered.

  • Lurkers - People who spent a certain minimum amount of time on your website—say “2 minutes” or “6 minutes”. This is an indicator that they are seriously considering your offer.

  • Churners – People who used to subscribe to your service but have since cancelled. When you release new features or reduce your pricing, you might be able to convince these people to start their subscriptions again.

  • Window shoppers - People who viewed a particular product (or category page). This is really a bunch of lists, one for each major product or category you stock. Having these lists around will enable you to segment your leads and only remarket products that each lead previously showed interest in.

It would take ungodly amounts of time for companies with expansive product ranges to set up hundreds of thousands of remarketing lists so as to segment their leads properly. To this end, Google created something called Dynamic Remarketing. If you’ve ever visited Amazon and looked at a Bose speaker only to see the same model advertised a minute later on another website, then you’ve been exposed to this technology. Dynamic Remarketing works by pulling product data from a feed on the advertiser’s website and automatically generating remarketing ads.

Remarketing overload

There is a risk that advertisers employing remarketing end up bombarding their audiences with adverts everywhere they go. This can come across as annoying, or even stalkerish. In gauging how pushy you’re being with your remarketing efforts, remember to consider the total effect of your remarketing across all platforms. If a customer sees your remarketing ads through Facebook, Google, and also Twitter, their exposure to your advertisements is tripled….

One way to combat this effect is with frequency capping. This works to limit the number of ad impressions shown to one person in one day. Personally, I configure each platform to show only one impression per day.

The second way to soften the potentially pushy vibes from remarketing is to limit how many days your campaign runs for. This can be controlled most easily by advertising to suitably short-stay remarketing lists (e.g., seven days). After this time is up, any given individual on that list would no longer see the advert (unless they visited your website again…). The advertiser without the benefit of short-stay lists can exert cruder control by simply turning the whole remarketing campaign off after a number of days.

Negative remarketing

There are also negative remarketing lists. These can be combined with other targeting options (or even positive remarketing lists) to further narrow the targeting of a campaign. The classic use-case would be companies selling a single product. Once someone has bought this item, there is no chance of them buying again; therefore, it makes no sense to continue advertising to them.

Visitors of other websites

Targeting visitors of other websites is perhaps the most basic of all the targeting mechanisms ever built, one that is as old as the internet itself. Even to this day, it’s possible to directly buy advertising inventory from any website offering it. That said, inventory is nowadays mostly sold through bigger brokers, with Google being the biggest of them all. Through their Display Network, you can rent space on any of the gazillion websites signed up to Google’s AdSense program. Indeed, thanks to their “automated placements” feature, you needn’t choose the exact websites yourself; instead, you can rely on other targeting options (e.g., keyword matching) to pick out appropriate websites.16 After leaving this running awhile, you can swoop in and promote the most profitable placements, and shoo away the bums. If this sounds a little familiar, that’s because it is: We already encountered “placements” in the section about targeting users of other people’s apps.

It’s possible to go even deeper than choosing individual websites in your manual optimisations. Websites today are rather diverse affairs, and perhaps only a fraction of their pages are relevant to your business. One way around this waste is to ask Google to limit the targeting to named folders (or specified URLs) of the placement website. If you are a seller of digital cameras, there’s little point in indiscriminately advertising on every page of a review portal—but it wouldn’t be a bad idea to advertise on the sections pertaining to photography.

Of course, pruning placement folders may quickly become tedious, so you may prefer the automated option of combining placement targeting with keyword targeting. Now, advertisements will only be shown on pages that also match for keyword content.

Because webmail clients count as websites, it is possible to shoehorn your adverts into email inboxes. To advertise in Gmail, set up a placement for mail.google.com then create some special Gmail Ads from their Ad Gallery.17 Once approved, your ads take an expandable, email-like form and show up under Gmail’s “Promotions” tab.

*Source: Google.com*

The websites displaying the adverts in the Display Network come from all walks of life: extreme right-wing newspapers, obituary listings, parked domains, lobby groups for legalising euthanasia or banning abortion, etc. Needless to say, juxtaposing your brand with these websites may be undesirable. Exactly how undesirable depends on how sensitive or family-friendly your brand is intended to be. Edgier brands, such as food delivery company Eat24, actively seek out cheap advertising options on adult/porn advertising networks, buying up advertising real estate that other brands wouldn’t touch with a 10-foot pole. Another important consideration is your target demographic: Young internet users are usually aware that there’s an element of randomness to where advertisements show up, and know that an advert’s presence on an undesirable website doesn’t imply that the advertiser was aware of or approving of the connection. The same cannot be said for older, less technologically savvy web users.

Google offers various options for taming the range of placement websites. The most surgical and precise of these is blocking specific unwanted sites through negative placements. Also rather precise (and tedious) is the option of blocking websites that contain certain “negative keywords”. For those who don’t want to fiddle endlessly with these options, there is an easier general solution: Google has high-level filters available in their “site category options” which let you disable anything from military content to advertisements placed on error pages. The current list is printed below:

*Source: Google.com/adwords*

I’d like to finish with some cautionary words about the technical format of placements: There is a meaningful distinction between domain names with and without the preceding “www”. Google’s documentation tells us that excluding example.com is general enough to also exclude www.example.com, but excluding www.example.com means your adverts will still appear on example.com.18 Because many websites respond to both “www” and the bare domain name, your placements (or exclusionary placements) could end up far less comprehensive than intended.

Language

Given that most platforms offering language targeting also offer country targeting, this feature only has limited applicability. One such application is tailoring the ad copy in countries where multiple languages are spoken. For instance, an advertiser in Belgium would show ads either in Dutch or French, depending on the primary language of whoever’s on the receiving end.

Language targeting also shines when targeting subpopulations within a country that are identifiable through their speaking a language other than the national norm. Someone residing in Spain but speaking only English is probably an expat, and is therefore a good target for adverts for Spanish courses or English-speaking tax advisers.

Education

Facebook and LinkedIn let you target those who have attended a given educational institution, studied a specific field, or attained a certain degree. This sometimes extends to negative targeting. LinkedIn gives the example of a degree-awarding education provider negatively targeting people who already have postgraduate degrees.

Facebook has a few especially nifty education targeting additions. For one, you can specify a range of undergraduate years, so that adverts only show to one particular cohort. This enables you to run the same campaign every year, but to different cohorts, excluding from targeting those who saw your adverts in the previous academic year. Facebook also lets advertisers target by general education level, such as “at high school”, “at university”, “graduate”, “doctorate degree”, etc.

As you’d expect, education targeting is of most use to those in the business of education: MBA providers, online course purveyors, textbook publishers, or tutoring services. Another major consumer of these features would likely be recruiters who hope to snap up fresh graduates for their clients.

Workplace and job targeting

The primary purpose of these targeting options is to reach companies. Because companies are nothing but collections of individuals, the B2B advertiser has only to reach employees at a particular company in order to reach that company itself.

There are, of course, cheekier use-cases for workplace targeting. I know of a PR consultant who saturates employees of relevant trade journals and magazines with messages about his clients, thereby subconsciously kindling these publications’ interest.

LinkedIn, as you would expect, does workplace targeting best, not least because its members have profile info that is more complete and up-to-date than anything on offer elsewhere.

LinkedIn (along with Facebook) lets you target employees of a particular company (e.g., Microsoft or Nike). This is a godsend once you know which companies you want to pitch to or partner with. Both advertising platforms also offer targeting by job title and industry. Targeting by job title is fully granular, so you have to synonym bomb a list of variations in the hope of matching the myriad self-reported job titles on file. Targeting by industry isn’t granular at all; instead, you target broad predefined categories like “Healthcare”, “Education”, or “Tobacco”.

LinkedIn (alone) lets you target by job seniority (e.g., Entry, Manager, Director, CXO, Partner) and job function (e.g., Accounting, Legal, Marketing, Operations). These options let you hone in on the decision-makers working at the relevant parts of the company you’re pitching too. There wouldn’t be much point in your financial services company advertising to your potential client’s kitchen staff.

Sometimes you want to advertise to only the biggest players, and sometimes you only want to reach the little guy. LinkedIn lets you do both by means of its targeting by company size. Through this option you can choose to advertise to companies with only one employee all the way up to behemoths with over 10,000.

LinkedIn members can form “groups” for whatever purposes they like (e.g., local networking, professional associations, topic-based support). Unsurprisingly, LinkedIn lets you target these. To get some ideas about which groups to target, check out LinkedIn’s groups directory and search.

Recruiters, professional training course providers, and so on will be fans of LinkedIn’s targeting by skills. If you’re organising a conference for a certain technology (e.g., Rails or CSS), then this is the targeting option for you.

  1. Given that Google AdWords is the most popular of all advertising platforms, I’ll describe a specific workflow for gathering negative keywords here. AdWords compiles a report called the Search Terms Report that tells you what actual search queries triggered your adverts. This report will inform you that your bid on the keyword “candy shop” ended up displaying adverts 20 times for the query “50 Cent candy shop”, 10 times for “download candy shop”, and so on. This report also tells you the click-through rates and cost per click for each of these actual search queries. After leafing through it, you’ll know what negative keywords you ought to set up. 

  2. http://www.statista.com/statistics/274564/monthly-active-twitter-users-in-the-united-states/ 

  3. https://www.wikiwand.com/en/Home-ownership_in_the_United_States 

  4. As if there weren’t already enough ways to think about location, Facebook introduced the concept of “recently moved”, which they define as people who recently changed the city listed in their profile. 

  5. If you’re curious about what interests Google thinks you have, check out their Ad Settings page: https://www.google.com/settings/u/0/ads/authenticated. 

  6. https://support.google.com/adwords/answer/2404191?hl=en 

  7. https://support.google.com/adwordspolicy/answer/6118?hl=en 

  8. It’s not possible to employ “Connections” to target fans of pages others own. 

  9. http://www.wordstream.com/blog/ws/2015/10/05/adwords-customer-match 

  10. Mobile advertiser IDs are basically unique device identifiers for mobile phones. Apps can be programmed to detect these IDs and shoot them off to the marketing department. 

  11. https://blog.twitter.com/2015/introducing-partner-audiences 

  12. Google offers this feature as part of its remarketing toolkit; Facebook as part of its custom audiences. 

  13. https://plus.google.com/u/0/+GoogleAds/posts/Wj6AjvvsW8d 

  14. But bidding more per click when there is no threatening competition is unnecessary and will increase your cost of advertising over the long term. Therefore, it’s important to know whether you are losing a share of available impressions to your competitors. In Google AdWords, the extent to which this is or isn’t happening is summed up in a statistic called “impression share”. If you already have an impression share of 90–100%, there’s no point in raising your current bids. 

  15. May leave you with significantly less leads” because some types of remarketing lists on some platforms can be populated retrospectively. That said, most cannot, so it’s safer to create skeleton lists as soon as possible. 

  16. https://support.google.com/adwords/answer/2580292?hl=en 

  17. There are many gotchas; see this video walkthrough from Google: https://support.google.com/adwords/answer/6105478?hl=en. 

  18. https://support.google.com/adwords/answer/2454012?hl=en 


More Articles:

SEO in Image Search

SEO-ed *text* enters you into the main event, but you might also want to compete in Google’s alternative races, such as Google Images, Google Videos, or Google Maps.


Check Out My Guide to Statistical Significance on Kissmetrics

Fun, example-filled guide, aimed at building intuition


SEO at the Whole Website Level

Internal linking strategies, sitemaps, and HTTPS