Tech Guru

Trusted Source Technology

Surviving the mobile marketing winter

Surviving the mobile marketing winter

Three forces, in combination, present immense challenges to performance marketers at the moment:

  • Apple’s App Tracking Transparency (ATT) privacy policy has eroded the efficacy of ads targeting across various direct response advertising channels, but especially on Facebook and Instagram. ATT reached distribution on a majority of iOS devices in July of 2021, but its impact became acutely apparent in Q1 of this year, for reasons that I outline in this interview;
  • Macroeconomic weakness that is coinciding with various inter-related drags on consumer spending such as supply-chain shortages and inflation (consumer prices were 8.3% higher in April 2022 than in April 2021);
  • Changing consumer behaviors as social norms revert to pre-COVID patterns and time spent with digital products subsides.

It’s impractical, if not impossible, to try to tease out the individual burden of any of these dynamics on mobile advertising performance, generally. And it’s also largely beside the point: it is the “perfect storm” combination of these three conditions that compounds to such painful detriment to advertising performance.

As a result, commercial performance is indisputably deteriorating:

  • Snap updated its Q2 guidance lower in May, just one month after the company announced Q1 earnings, citing macroeconomic headwinds as the reason that it expects lower topline revenue and EBITDA in this quarter;
  • Google saw decelerating year-over-year growth in direct response advertising revenue on YouTube in Q1 2022, which the company attributed to a difficult comparison quarter related to COVID and to ATT;
  • In its Q1 2022 earnings call, Shopify’s president proclaimed that, “We do believe [ATT has] created some friction for merchants and advertising and [has] lowered their return on ad spend.” As of this writing, Shopify’s stock is down 75% for the year;
  • According to SensorTower, numerous app developers have seen app installs decrease globally across both iOS and Android between December 2021 and May 2022: Zynga by 21%; Voodoo by 16%; Activision Blizzard (which owns King, the developer of Candy Crush) by 11%;
  • Netflix lost 200,000 subscribers in Q1 of 2022 and expects a net loss of 2MM subscribers in Q2 2022, due in part to formidable competition from newer-entrant streaming services but also as a result of changing consumer habits as COVID restrictions ease.

Look to any public DTC company’s stock — Smile Direct Club, StitchFix, Blue Apron — on a year-to-date basis and the combined devastating consequences of ATT, the COVID overhang, and lurking economic fragility are apparent. Again, the individual contributions of these three forces are irrelevant. Taken together as a composite, they have engendered unforgiving barriers to growth. A mobile marketing winter rages.

In this article, I’ll identify three broad, strategic initiatives that marketers can undertake to navigate the mobile marketing winter. Some of these are covered in the Advertising strategy in a recession presentation, and some are touched upon in the recent MDM podcast of the same name. In this piece, I’ll attempt to provide color around why these particular propositions are prudent given the circumstances.

Re-establish measurement baselines

Measurement is probably the operational gap on which most marketers’ attention is currently fixated. The combination of COVID overhang and ATT laid waste to much of the analytical infrastructure that advertisers use to direct budget:

  • Among other things, ATT renders creative-level measurement infeasible (although that may change with the introduction of SKAdNetwork 4.0), meaning that the granularity at which many teams previously optimized ad spend is no longer possible. Since user-level revenue data cannot be attributed to ad creatives, or even really to campaigns as a result of the deficiency of SKAdNetwork, advertisers are left to unify unattributed revenue data from their products and top-of-funnel ad engagement metrics from ad networks and ad platforms;
  • Because the COVID pandemic was so protracted and globally pervasive, advertisers’ ROAS models and general engagement behavioral models are seeded with years worth of data that is no longer relevant as COVID norms dissipate. The models that dictated advertiser spend for the past two years cannot be used to describe current or future consumer behaviors.

These issues are thorny: they necessitate wholly new (and more complex) analytical methodologies both for advertising measurement / attribution but also in-product behavioral modeling. For marketing measurement, many teams are shifting their infrastructure toward macro-level predictive tools such as media mix models or incrementality models which attempt to use variations in performance and ad spend over time to measure channel performance. These approaches can be effective, but they can also be difficult to implement.

Confounding matters further is the complication that changing, post-COVID behaviors and recessionary threats introduce. With ATT, the behavior of new users cannot be taken for granted because ATT primarily disrupts targeting and causes ad platforms to deliver less relevant traffic to advertisers. But changing post-COVID norms and economic softness will impact the behaviors of existing users, meaning that older cohorts cannot necessarily be expected to behave as predicted at the time of their acquisition. Advertisers could see existing cohort engagement diminish at the same time that they see advertising performance deteriorate.

Marketing teams must establish new measurement baselines, or reliable models of advertising and product economics, as the broader commercial environment changes. ROAS models that are mostly informed by COVID-era data cannot be trusted; the same is true of cohort forecasts. I outline the process of establishing a ROAS model from a cold start in It’s time to retire the LTV metric:

I believe the better approach to thinking about LTV in the abstract, terminal sense is to buy traffic at a 100% ROAS target against an early-stage benchmark until enough data is accumulated to establish new, later-stage benchmarks. An example is outlined below: the marketing team sets a ROAS goal of 100% at the Day 15 benchmark and accumulates enough data to understand how the ARPU curve evolves to that point. Using that curve, the team estimates the bid level for 100% recoup at the Day 30 benchmark and buys against that (reducing the Day 15% ROAS to 80%). The team then accumulates data until they can estimate the bid for 100% ROAS at Day 60, and so on.

This process is slow but it’s necessary in order to demarcate new bid and budget thresholds, especially if the attribution methodology is changing at the same time. I have seen marketing teams use media mix models for channel-level performance measurement and on-platform ad engagement analytics for creative-level optimization. This is a sensible strategy — or, at least, it’s worth experimenting with — but it creates two entirely new optimization siloes that the team’s legacy model can’t possible accommodate. Going through the process of seeding the performance model is a critical step in adapting to the new environmental reality.

Improve ARPU through in-product personalization and strategic cross-promotion

I explain why in-product data takes on expanded influence in the post-ATT environment in Why in-app personalization, not probabilistic attribution, is the future of post-ATT advertising:

Prior to the deprecation of device identifiers, all app personalization was outsourced to advertising platforms. These platforms reached the most relevant audiences based on aggregated behavioral profiles sourced from other properties. Now, absent those profiles, ad platforms are less able to locate relevant users for exposure with ads…Given this limitation, it is incumbent on app developers to find ways to parse apart the broader, more heterogeneous accumulations of users that are presented to them as cohorts by ad platforms into meaningfully-defined groups.

Essentially, prior to ATT, the personalization function was outsourced to ad platforms by advertisers: the advertiser supplied a monolithic product, and the ad platform supplied a relevant stream of vetted, curated users on the basis of behavioral profiles. Since that is no longer possible — or is, at least, less effective — then the advertiser must take on the effort of personalizating the in-product experience. Product personalization is an expansive topic, but here I refer to any effort to bring the user content based on their specific traits and habits in service of increased retention and/or monetization.

The goal of this exercise is to increase ARPU such that increases in CAC are compensated for. Another approach to achieving this is to unlock multi-product LTVs through strategic cross-promotion: moving users from one product to another before the point of churn. The ARPU benefit is clear: instead of acquiring users against their projected revenue in one product, a multi-product sum of projected revenues can be used, which should hopefully increase advertising bid levels and make up for increased CAC.

The mechanics of executing cross-promotion successfully are no different than in serving personalized offers or algorithmically-curated content: in the case of cross-promotion, the content being served to the user on the basis of relevance is an ad for another product owned by the same company. I chronicle my effort to scale such a portfolio-wide cross-promotion system in this 2016 GDC presentation: ultimately, the initative resulted in a much higher proportion of users installing in-network apps, creating a “portfolio LTV” effect that increased average user value. I believe that this recognition of the need to support a portfolio LTV is driving much of the current consolidation activity in the mobile gaming category.

Establish demographic-targetable audience segments through qualitative research

In the halcyon days of the pre-ATT era, automated, conversion-optimized targeting tools like Facebook’s AEO and VO bid strategies and Google’s generally opaque UAC platform mostly alleviated the need for advertisers to speficy any targeting outside of custom audiences for lookalike targets. These systems grouped users into audiences based on behavioral commonalities (mostly related to conversions, eg. purchases) that would not be possible for advertisers to observe or aggregate themselves. More importantly, these groups were totally inscrutable and not organized on the basis of any perceivable set of demographic traits.

The ability for platforms to do this type of behavioral targeting is significantly diminished by ATT (and the fast-changing privacy landscape more generally). Only very broad demographic features are now available to target against on ad platforms: things like geography, phone type, potentially age and gender. These features (and combinations thereof) tend to be poor, or, at least, non-optimal predictors of engagement and monetization.

But those demographic features are the only levers available to advertisers for targeting now, and products must be adapted to that new constraint. Aligning the product with the broad targeting parameters available in advertising is primarily a product marketing exercise for existing products (how can the product be altered to best resonate with Men, aged 18-35?) but an audience development exercise for new products (what kinds of products best resonate with Men, aged 18-35?). These initiatives require a sensibility that many direct response advertisers don’t possess, as again, ad platforms previously actively managed audience targeting on the basis of conversions, with no assumptions about demography required from the advertiser.

“How can a product be adapted to a specific audience once it is live?” is a common response I get to this propostion. The anser is that the product’s functionality and core use case probably can’t (and shouldn’t), be, but its aesthetic (its “packaging”) can be.

I speak to this in The Power Triad of Resonance for Mobile Games, although the concept is pertinent to any category of consumer tech products: the core use case, the tone, and the theme of a product must all align in order to be resonant. Theme and tone changes might penetrate into the product in terms of design, but they might also merely be applied at the ad creative layer to reach the intended audience (although a discordant aesthetic between ad creative and product could create other performance problems).

Concretely, the team should decide which set of intersecting demographic targets best captures the most relevant audience for the product and find ways to guide the product in that direction, predominantly aesthetically but potentially functionally.

Photo by Alex Gruber on Unsplash