The era of algorithmic marketing: 6 mobile marketers comment on Google’s UAC & more

This guest post was originally posted on MobileDevMemo.

The move to algorithmic systems like Google's UAC and Apple's Search Ads Basic represents a fundamental and pivotal transition in the way that marketing dollars are invested in mobile marketing campaigns. No longer do mobile marketers have the ability to choose who, where, or how their ads are shown to users -- instead, algorithms decide these logistics, guided by few inputs, such as CPI bid and budget.

What do marketers think about this shift?

Here are six voices on the topic covering the good and bad, how it affects marketers, and what it means for the future of marketing.

Allison Guidetti - Head of Global Marketing, The Weather Company, an IBM Business

AI is listed as one of the 12 global threats against humanity by the Future of Humanity Institute. However, unlike the other 11 threats -- items such as extreme climate change, nuclear war, and an asteroid impact -- AI can also work to combat the other threats to save humanity.

So, my fellow marketers, how does it feel to have a hand in leveraging not only one of the greatest threats against humanity, but also one of its potential saviors? With great power comes great responsibility.

Our entire world is controlled by algorithms that few people truly understand. Maybe that’s a bit hyperbolic, but Google Maps controls where I go, and Amazon controls what I buy. Facebook and Buzzfeed go even further, suggesting who and what I think about while eerily reinforcing my existing beliefs. And, as a new mom, Amazon has an increasingly heavy influence on my life at around 3 a.m. (Seriously, packages of get-your-baby-to-sleep products show up at my door that I do NOT remember ordering).

With an algo-controlled world, it makes sense for all of us to leverage algo-assisted marketing. Let’s be clear, though. You can have marketers without algos and algos without marketers, but it's the combination of both that creates magic and synergy.

In a recent Ad Age article, Simon White posited that "ROI increases are an addictive drug, one that's hollowing out the brand and leaving an empty carcass." To add to this, a risk with algo-controlled marketing -- while it allows for smarter user insights -- is that it simply gets us to those empty carcasses faster. To mitigate this risk, it’s up to marketers to control the algos, not vice-versa.

Know what “motivates” the marketing algos -- money. If you aren’t careful, algos could bring in a bunch of low-cost, but also low-value, consumers to your products. To help control the algorithms, remember the following:

• It’s up to marketers to ask the right questions.Algorithms offer amazing options, but remember they are only as good as their inputs. Flawed input data will result in flawed outputs. Learn and understand how your inputs impact the output. What’s your goal? For example, do you need an app install or deeper action by that user once in the app? Know what data you need first, and then choose the right inputs, monitor results to confirm the strategy, and course correct when needed.
• It’s up to marketers to make the final decision.Don’t rely on the algorithm for more than data or insights. We’re talking about human decisions, and only humans can understand the whys behind consumer behavior. The winning formula includes BOTH algos and humans to equal synergy. It’s a risk for marketers to think otherwise, so do not take yourself out of the equation.
• It’s up to marketers to build brands and cultivate trust.Brand trust becomes even more important in a world of increasing AI and algorithms. It takes time to build and maintain brand trust, but that trust can be eroded in an instant. Consumers look for a company that cares about them and doesn’t betray the relationship being built. Think beyond the short term, such as bringing someone into your funnel at a very low cost. Marketers must balance short-term wins with long-term profitability and brand building.

The way marketers work is evolving. We must move fast or get stuck in the past. If you haven’t already, start to invest in algo-assisted marketing as it can save you time and allow you to focus on more strategic tasks… but not if it’s at the expense of the people who make those tools effective.

And if you already are investing, ensure you are being really strategic about your inputs and KPIs, understand how the algos work, and analyze the results using your own knowledge and brand expertise. Algos need human interpretation to be most effective. Only humans can understand the “why” behind the “what” to craft the game-changing insight that drives decisions and builds brands.

 

Thomas Petit - Growth Marketing, 8fit

In the past 12 months, Facebook released its “App event optimization” bidding strategy, Google forced advertisers to only use its “Universal App Campaigns” campaign structure, and even Apple introduced “Basic”, a simplified version of Search Ads. These moves make a lot of sense for these companies to try and attract advertisers who lack advanced marketing expertise and in-house resources, with the promise of saving the hassle of setting up and controlling campaigns permanently. I see them as the first attempts in an irreversible trend in which the machine will eventually substitute the hard work of monitoring and tweaking campaigns on the go. On the mid-long term, machine-controlled app acquisition is inevitable. Resistance is futile: AI will win over human marketers on both input costs and output deliverables.

But we’re not there yet -- actually, we are quite far away. Such schemes come with frustration for app marketers, not only because of their complete loss in control, data and learnings, but because these products are still in their infancy and rarely deliver on their promises. With Facebook and Apple, you can choose to use them or not, while Google unilaterally forced all advertisers into its UAC before reaching product maturity.

As a result, many app marketers struggle to reach past KPIs and are left with barely any leverage to optimize their ROAS. Those levers do exist to some extent (creative diversity, negative keywords, event tweaking…), but they’re very limited in comparison with manual bidding allowing to target specific audiences, queries, placements and use bid modifiers to adjust to every app specific reaction, like locations, devices, day of the week or time of the day to name only a few. This product has been developed to cater for the largest customers, usually gaming, dating and e-commerce companies with extensive budget and broad audiences, and at this point it’s mostly unsuitable for many other cases like complex funnels, free trial subscription apps, limited population geographies, niche audiences apps, etc.

In the short term, I’d recommend considering if your setup and goals are compatible with the way optimization is done in such systems. In particular, app devs mainly operating on iOS, with low advertising budget (machine needs volume and a learning period), targeting events with low completion rate (typically a 1-5% IAP conversion), lacking compelling video creatives, unaware of some dubious UI default options (location targeting, negative keywords etc), and careful about their brand term install volume should think twice about allocating budget there. Combined, such cases are particularly common if not defining among indies and early stage startups.

The likelihood that the “learning period” will never reach the promised optimization stage is very high, and devs will be left with money gone and zero insights. Given the very high value of the learnings made early on about the specific placements, queries, demographics or creatives that resonate with one’s own audiences, I’d recommend starting with laser-targeted manual campaigns, and only move on to machine-driven if/when budget allows and previous criteria are met. In the meantime, one can only hope that as time goes by, machine driven app acquisition products will evolve with optimizations on reporting, settings and input to soften the learning period risks.

 

Gabe Kwakyi - CEO, Incipia

With markers like the release of Apple’s Search Ads Basic and Google’s famous move to UAC, the industry has shifted to a new era of marketing -- one in which ad platforms, gorged on regressing trillions and trillions of data points, have started to decide what’s best for their advertising customers’ dollars. Algorithmic marketing promises to do away with tedious targeting work in favor of saving more time and getting better results than humans are capable of. Sounds great, right?

Yes and no. For unspecialized marketers such as indie dev teams, this as a positive inflection point. Yet for more savvy marketers who are better equipped to leverage certain optimization levers, the transition of control from humans to algorithms is not all good.

Regardless of how many data points consumed or Go matches won, machines still underperform humans in many cases of context. Think: classifying cat pictures.

While optimizing for trends of good / bad performance in a standardized cost / conversions sense is within a machine’s capabilities, the fact that machines still can’t accomplish a task as simple as identifying cat pictures (as one example) underscores that fact that machines still do not know better than humans. So why do away with the ability for humans to decide on, say, targeting?

There certainly are pros and cons of leaning more on machines for marketing, and this comment is by no means a call for a rollback of the algorithmic marketing wave. Some of these changes are indeed useful. Standardizing creative inputs is a boon. Automating targeting for exploration purposes is nice. Automating bidding decisions based on controlled logic is wonderful. Paying a max CPI is fantastic.

But as the users of these new algorithmic marketing tools, it’s important for us in the industry to make our feedback known, good and bad, in order to force these companies to stay focused on building products we advertisers can get behind, rather than chasing quarterly profits without regard to the consequences.

So Google: bring back negative keywords. Bring back the ability to AB test creatives. Allow us to adjust bids or budget for groups of users we have more context (or business knowledge) on than the algorithm. Allow savvy marketers the option to still pull the levers we need to access in order to do a good job. Give us an “all features vs basic features” selection, as you did with search campaigns, and like Apple did with ASA advanced vs basic.

The good news is that companies like Google have listened to advertiser feedback; the bad is that this has stemmed from very negative circumstances (the hate speech ad placements snafu).

The era of algorithmic marketing doesn’t have to be an either humans or machines dichotomy, nor should it be. Listen to our feedback and we’ll build a better product in the end.

 

Eric Seufert - Owner, Mobile Dev Memo / Platform at N3TWORK

The shift by the duopoly (Facebook and Google) to event-oriented bidding methods for mobile advertising in 2017 astounded and alarmed some mobile marketers. These value-based bidding models extend the domain of the advertising medium down into the analysis process: sophisticated mobile advertisers are used to handling all ROI analysis with their own tools and their own data, and the duopoly re-engineering their platforms to make that irrelevant surfaces some understandable skepticism.

There exist real reasons for concern with this trend but also illegitimate ones. A well-founded concern is that Google, with its newly-agglomerated Universal App Campaigns (UAC) format, removes the ability of the advertiser to select the channels that perform best for them: YouTube, AdWords, Google Play search ads, etc. are all now rolled into one advertising channel that can be bid against on the basis of event completion. For advertisers that possess the infrastructure and domain expertise to evaluate each advertising campaign at a very deep level of granularity, UAC appears a "black box": an opaque system into which money is funneled and events are extracted. Aside from creative variation, there's little an advertiser can do to optimize performance on UAC now — many advertisers don't even adjust bids, since the machine learning algorithms that power UAC are meant to optimize against completion of the in-app events they are being fed by the developer. Any time a vendor changes their service in a fundamental way that reduces transparency, apprehension from customers is justified.

But an illegitimate concern is that these changes will simply extract more money out of customers without providing additional value. From a market perspective, this isn't a credible fear. Facebook and Google, collectively, own an enormous share of the digital advertising market —- in 2016, they captured 89% of all revenue growth -— but it's important to remember that these companies are fierce competitors. If both Facebook and Google moved into value-based bidding, it's because they see that format as a means of poaching clients from one another. This isn't OPEC: each of these companies are doing their utmost to steal market share from the other.

It's also important to remember that for most advertisers, value-based bidding is nothing short of a godsend: very few companies have the wherewithal to build out end-to-end analytics and marketing automation infrastructure. For these companies, the burden of needing to create an entire team with a function that is wholly orthogonal to the purpose of the business is counter-productive and distracting, at least at the early stages. From this perspective, value-based bidding isn't rent seeking but rather an expansion of the market via reduced barrier to entry.

And the duopoly isn't the only one pursuing value-based bidding: every major mobile advertising network has, at least to some degree, integrated in-app events into their delivery algorithms (the level of integration varies from network to network, but that's a separate topic). And so far as the mobile advertising ecosystem's development has emulated that of the desktop advertising ecosystem more than a decade ago, value-based bidding isn't really a surprise: it's the closest approximation of the advertising pixel that mobile can accommodate (for a review of the systemic differences between desktop and mobile, see this article).

For the sophisticated advertisers who feel that Facebook's app event optimization (AEO) and value optimization (VO) bidding types and Google's fully-subsumed UAC channel have rendered their tools and infrastructure less valuable, the developments of 2017 serve as a poignant reminder that the mobile advertising ecosystem is evolving rapidly and that flexibility is a competitive advantage. The shift to value in mobile advertising bidding emphasizes the imperative of powerful and engaging creative, strong resonance between ad creative and an app's user experience, and a finely-tuned value proposition for the app that creates a superb and unparalleled experience for the consumer. This considered, it's hard to argue that value-based bidding works against the most important participant in the advertising dynamic: the user.

 

Elad Amit - SEM Specialist, Playtika

Anyone who has been in the online marketing industry for some years now couldn’t help but notice how all ad networks are moving from a somewhat manual form of optimization to an algorithmic-based one. While everyone on Playtika's marketing team knew that this transition would change the technical aspects of running mobile ad campaigns, we were surprised by the extent to which it changed our entire User Acquisition strategy.

We’re now constantly researching which kind of events are best for optimization; how much any specific action a user may take is worth for us. This kind of down-the-funnel optimization forces our marketing department to work closer than ever with the other teams – Monetization, Product, and Analytics just to name a few.

An additional significant effect UAC had on our daily work is that we now allow ourselves to get more creative with our designs. Since UAC picks the best creatives for each campaign, we want to give it plenty of different possible variants to choose from. This learning curve also means that we now allow more time for campaigns to reach outstanding performance, as we understand the system needs to thoroughly study the ads and audience.

While a few years ago the main requirement from an online marketer was to dive into the specific data sets of each campaign, quickly and efficiently, the 2018 online marketer needs to possess a new set of skills. Algorithm-based campaigns necessitate a much more creative form of thinking: as marketers think of down-the-funnel events, they should stop thinking about users, and start thinking about players.

Soon we expect to see a shift from a single-event-based bidding to a multi-event-based bidding. This means marketers would be able to bid separate rates for users who reached event A (reached level 5, for example), and are likely to perform event B (deposit within the next 14 days, for example). We look forward to being able to bid not only for actions players have taken, but for actions they will take in the future.

 

Eduard Krecmar - Performance Marketing, AppAgent

Right now we are in a very specific evolutionary time in the field of mobile user acquisition. The already-steep technological curve for consumer tech companies is moving us rapidly toward functional Artificial Intelligence (AI). This is extremely relevant when we talk about media buying on the biggest channels, including Facebook and Google.

Google's UAC was the first true blackbox ad product. Far from being welcomed, it was received with concern. A blackboxed ad product demands an incredible level of trust between the media and advertisers, which is – in the case of UAC – nascent. Your success in user acquisition is already tied to a couple of platforms, and now you are being asked to blindly trust algorithms that promise to deliver better – and more relevant -- results than those you are able to achieve on your own.

This naturally generates more questions regarding inventory, quality of placement and targeting. Ad fraud, brand safety and data protection are claimed by Google reps to be 100 % secure, but advertisers have zero ability to verify these promises. As a business, we are being forced to fully trust the ad platform – and we’re talking about multi-billion dollar operations, not flower shops around the corner.

The loss of control and insight we experience when comparing automated campaigns to manual campaign management is painfully real. The leap of faith is not going to happen within one year, and frankly automated ad products have to provide more information and some meaningful level of control to UA managers to gain the widespread trust of the market.

A less critical point of view is that today a wider range of advertisers are able to setup well-performing campaigns without a significant knowledge of performance marketing. This simplification in marketing naturally puts more importance on the product itself, from which end-users benefit. Also, daily user-acquisition management can be pretty repetitive, so automating it with a pinch of AI is a blessing. Successful mobile companies, such as the larger mobile game developers, already have a high level of automation in-house, so I for one welcome our binary overlords with a positive attitude.