The sign of times to come
A few years back, I attended an event Google India had organised for mobile marketing practitioners. They were pitching a then-new product called Universal App Campaign, which had already been acronymized as UAC. The idea was simple: you can use just one campaign to promote your app across all of Google’s varied inventory such as Google Search, Play Store, YouTube, websites and apps on its ad network, and what have you. No need to make separate search, video, and display campaigns for the same objective. Hence, ‘universal’, I suppose.
The pitch for UAC to marketers was: less work for you. But judging by the response of the early users among the attendees, we did not want to do less work. People seemed to be genuinely unhappy with the product. Post-presentation Q&A went something like this:
Marketing Managers: ‘How do I add interests to target my users with?’ Google Account Manager: ‘You can’t.’ Marketing Managers shake their heads.
‘How do I add gender for targeting?’ ‘You can’t.’ An even more vigorous shake of the head, accompanied by a scoff.
‘How do I upload a list of users to target?’ ‘You can’t.’ You get the picture.
Turns out this new product UAC was taking away a lot of targeting levers marketers seemed to really want, under the guise of ‘less work for you’. So, why was Google doing this? Surely the good people at Google Ads would seek the feedback of their customers i.e. mobile marketing practitioners. After the session, attendees were already predicting the death of the product. “Well, it’s easy to set up and is good for newbie users. But it’s just not for us.”
Fast forward a few years: In March 2020, AppsFlyer Performance Index report put Google Ads ahead of Facebook Ads in its power ranking. And with AppsFlyer being the gatekeeper to a good chunk of mobile ad spend by marketers, it perhaps reflected the ground reality of ad market share. The report attributed the turnaround to UAC.
Google, in particular, got off to a relatively slow start (in 2015), but has since ramped up its presence on mobile.
Since launching its Universal App Campaigns product a couple of years ago, and in particular in the past year, Google has leaped forward.
So, what happened? Were the marketers wrong? Or did we get something we did not yet know we needed? Cars instead of faster horses?
A Brief Anatomy of Digital Advertising Campaigns
To answer this question, we have to understand how digital ad campaigns are set up. Broadly, there are 3 levels to be defined by a marketer to run a digital ad campaign:
- Campaign objective, budget, etc. That is: what do you want from the campaign? Do you want app installs, or likes, or videos views, etc? How much are you ready to spend?
- Audience targeting levers. That is: what are the demographic (location, age, gender), or behavioral (interests, occupation, etc.) characteristics of your prospective customer?
- Ad creative. That is: what is the written and visual communication that you want to show your prospective customer?
What UAC essentially did — and this is a way more central concept to understanding UAC than the ease of campaign creation or unification of inventory — is it took the 2nd lever i.e. audience targeting levers (apart from location) away from the marketer.
This idea might seem like a non-starter to a marketing purist. Surely, defining the audience is the most important aspect of the campaign creation. But as the success of UAC would now tell us: it perhaps isn’t. What are we missing?
This question had come up during the aforementioned event too. Someone asked: my product is only for parents. Surely, I have to set age targets. Someone mentioned their product was only for women. And, so on. Everyone seemed to have a valid use case to have some audience targeting.
At one point, a Google Account Manager asked a Marketing Manager: how do you know your product is only for so-and-so groups? Because it’s our product and we know only this persona can use it, came the incredulous and understandably irritated reply. And our campaigns will learn that too, the account manager batted the concern away.
Our campaigns will learn that too. That’s a great one-line explanation of how digital ad campaigns work.
Two things that maketh a great ad platform
An ad platform is essentially a marketplace. On the supply side, it has businesses who want to monetize the traffic on their websites and apps by showing them ads. On the demand side, it has businesses who want to show ads to prospective customers. The ad platform facilitates this transaction. (Google and Facebook are known to command a duopoly in this market. However, their supply and demand side names are different. For example, in case of Google, the name facing the demand side is Google Ads and the name facing the supply side is AdMob.)
But there is more to being a successful ad platform than just having some demand and supply. If I monetize my website by letting Google run ads on it, I will keep using them as long as they are paying me fairly. However, retention of advertisers is more difficult. As an advertiser, I will come back to a platform to run a campaign again, only if the previous campaigns have delivered results for me. Aka relevant users. Aka ROI. Aka ka-ching!
This is why I refuse to get excited when ad sales reps tell me their platform has 100 million unique active users in India and delivers 1 quintillion impressions every hour or something like that. That’s not enough.
2 things that make a successful ad platform are:
- Differentiated supply: Only Google Ads can show ads to YouTube users. Only Facebook Ads to Instagram. Only Bytedance to Tiktok.
- A system that has the ability to predictably deliver desired results to the advertiser at scale over a period of time.
Point 2 is under-appreciated. As important as Facebook’s news feed is to the history of Facebook Inc, I’d opine that the system powering Facebook Ads platform has been even more important to their success.
Imagine the complexity: A shopping app has given mandate to Facebook that they can spend 10k dollars over the next 30 days as long as they can bring 1k new shoppers to the app. So, which Facebook-Instagram users should Facebook’s system show the ad of the shopping app to? Surely not show the ad to any random 1 million users and hope 1% click on the ad to install the app and 10% among those make the purchase? Definitely Facebook’s system needs to predict which 1 million, from amongst its users, are the most likely to complete the series of actions. So, how does it make the prediction?
In absence of knowledge of inner workings of Facebook’s prediction algorithm, let’s make an educated guess: How does anyone make a prediction? By training a model with past results, of course. So, the campaign needs to get enough data points to build a model that takes in Facebook user’s attributes (that it has collected on them over time) as input variables and the probability to convert as per advertiser’s objective as the output. Once, the model has been trained, i.e. the campaign has learnt enough, it needs to now deliver the desired result at scale.
The above is an extremely simplified view of how things might be working down at Facebook Inc in any case, but I have to point out one key thing that’s missing: constraints. Constraints such as auction bid caps, ad schedule, and, most importantly, audience targeting that might have been put by the marketer (reducing the super set available for the model to train or to target).
The prediction job that this system is doing is hard enough and then the constraints are making it even more difficult. Only if they could get out of way. Only if marketers stopped putting them.
Only if there was a way.
The machines took our jobs (or did they?)
A Facebook employee once told me the deceptively simple strategy of a successful advertiser: put accurate campaign objectives, put as few constraints as possible, give enough time and let the platform do its thing. Very Buffett-esque. No hidden tricks.
That’s what UAC is enforcing on everyone: you won’t be able to put the constraints if we altogether remove the option to do so. While this is an extreme position taken by Google Ads (albeit being an accurate reflection of how machine learning is powering and driving digital advertising under the hood), it has not been an overnight phenomenon.
In the 2010s decade, when Facebook was suddenly gaining ground in the ad market despite the lead the incumbent Google had taken in the 2000s with their Search ads (Facebook’s ad revenue in 2009 was 0.7 billion dollars compared to Google’s 23 billion dollars; By 2019 Facebook’s ad revenue had grown 100x to ~70 billion dollars while Google had grown 6x to 135 billion dollars), it was down to a lot of innovations Facebook Ads platform had made with behavioral targeting. The crown jewel, perhaps, was Facebook’s Lookalike Audiences product.
The idea for Lookalike Audiences is simple: instead of telling the platform what you know about your desired customer persona, you just give the contacts of your best customers to Facebook and let them find their lookalikes. Just like UAC, it’s not very confidence-inspiring but it works.
And, even with creatives (3rd layer in the abstract model of digital ad campaign described above), Facebook and Google have been encouraging marketers to not come up with fixed ideas on what their ads were going to look like and instead giving their system enough options to mix and match and to learn from before scaling. Unthinkable for, say, traditional TV ads.
Essentially, it’s been happening gradually for a long time. The keys have been in process of being handed over to Google and Facebook by advertisers for some time and the car is now driving itself for the most part. We do choose to get in the car at a time of our choosing, and we get to punch in the destination, but that’s about it.
So, coming to the titular question: where is digital advertising going in 2020s? With ad platforms being more and more driven by machine learning algorithms, and fewer levers being given to the advertiser, is digital marketing, as a field, meaty enough?
Core vs non-core parts of the job
Business strategists love to ask: is so-and-so thing core to your business or not? If not core, outsource it to someone who does it at scale, and better. It reduces cost and saves mental energy required to run those operations.
Marketing is about distributing your product. That couldn’t possibly be non-core to a business. And yet, marketing and advertising are not the same thing. While, marketing as a process is about understanding the market and its needs, defining the value that product offers to meet the market’s needs, and building awareness about the product in the market; advertising is only the last part.
So, yes, marketing is not dead and is unlikely to ever be. We don’t build our customer persona merely for using it as audience targeting on ad platforms; we build them primarily for our product planning, our go-to-market model, and for measuring our relevant awareness.
And, while the levers of control in digital advertising (as we should accurately call digital marketing) will keep going down, it merely means we can focus on the core parts of our business more (which objectives are important to my business, for example) as well as the more creative parts, and less on understanding the nuances of a proprietary ad platform. So, yes, less non-core work for us.
The verdict: dead or not?
In summary, digital advertising is undergoing the same transformation that other sectors are going through. There are non-bespoke parts of jobs that are going away. But the field itself is not going anywhere. And the work will get more creative and more fulfilling, with more of the job getting done under the hood.
