💎 Growth Gems #97: Paid UA and Retargeting
Hey,
This week I’m sharing gems on:
These insights come from Marcus Burke, Yoann Pavy, Kate Lovejoy, Ava Savitsky, and Warren Woodward.
Enjoy!
🥇 TOP GEM OF THE WEEK
1. Meta ads for subscription apps
Gems from Marcus Burke (Performance Marketing & Growth Consultant), Yoann Pavvy (CMO at Nude) and Jake Mayell (Head of Growth at Hex Digital) in Meta Ads, Imposter Syndrome and Media Mix
Marcus was an early subscriber to Growth Gems, so he’s a good person 🤣
But more importantly, he’s been in marketing for over a decade and has significant experience in performance marketing: first for games, then for subscription apps (e.g., Blinkist, Tandem).
And the last couple of months, he’s been sharing many insights on LinkedIn (at a very high frequency - not sure how he does it!).
He was a guest on the Meta Ads, Imposter Syndrome and Media Mix episode of the relatively new podcast In Growth We Trust.
Gems below!
Targeting and analysis
💎 Everyone uses broad targeting, which makes sense if you can set purchases (your business goal) as a conversion event. When optimizing for trial starts, that signal has a much lower quality which means you want to add targeting restrictions.
(33:10 by Marcus Burke)
💎 Buying metrics and cost per trials can be more expensive for a lookalike audience, but the trial -> paid conversion is often better. So you need to monitor your product data to understand this in a SKAN environment (Mixpanel, Amplitude).
(33:10 by Marcus Burke)
💎 Look at conversion rates with different breakdowns in your product analytics tool, based on the questions you ask at onboarding or the data you collect. Spot any outliers (good or bad) in order to exclude or double down on some audiences. This can let you know spending more in some places is ok. Example: using zip code to spot which cities have the highest conversion rate and target those cities.
(35:36 by Yoann Pavy)
Event optimization
💎 With SKAN, you only have the first 24-hour window to send back a signal. Try to find a meaningful activation event that correlates more highly with conversion and create a combined event for “trial start + activation event“.
(39:40 by Marcus Burke)
Ok, you trigger the event and send it to your MMP. I followed up with Marcus to clarify the next steps.
He recommends adding this new event as an additional higher-priority event to your SKAN conversion schema (not to “replace” your current start trial event). That way, you can map both to Facebooks events to create a funnel in your ads manager
There’s most likely a difference when you map the event to Meta’s “purchase” vs. “start trial” event. While he did not have a definite answer for the post-ATT context, he shared that Meta’s “start trial” is nicely specific to sub-products but sits upper funnel (low quality). “Purchase” is a more generic but higher quality event that most companies (subscription apps, games, and beyond) send to Meta, which could help find audiences that are actually spending. This made him (and now me) lean more towards mapping the new event to “purchase”.
💎 Choose the same optimization event in your creative testing campaigns as in your scaled campaigns. Example: don’t do app installs in creative testing and start trials + activation event in the scaled campaign.
(50:32 by Marcus Burke)
Creative testing
💎 On Meta, until you reach $5k/month in spend, everything is a test: just one campaign and one ad set, where you make the best calls possible on creatives.
(45:05 by Marcus Burke)
💎 Keep in mind that on iOS, Facebook doesn’t have creative-level data (SKAN 3.0). So it looks at the engagement metrics and overall trials for the ad set, then attributes accordingly. If you test creative on iOS, only test ads with similar concepts/angle/audience in the same ad set. Use a separate ad set to test a new creative angle.
(45:50 by Marcus Burke)
💎 Use Meta’s automated testing for headlines and body text. Otherwise, that job just often doesn’t get done.
(51:40 by Marcus Burke)
2. Retargeting: measurement and creatives
Gems from Kate Lovejoy (US Managing Director at Adikteev), Ava Savitsky (Creative Director at Upptic), and Warren Woodward (Co-founder at Upptic) in Creatives in Retargeting and User Acquisition.
Retargeting took a hit on iOS with ATT.
However, many apps still run them to bring back users into their game/product.
On iOS, it seems this is done only for the cohorts where IDFA is available.
SKAN 5 will support measuring re-engagement, which brings some hope to the future of retargeting (even though the targeting aspect is still TBD).
So the timing seems fitting to share insights from Creatives in Retargeting and User Acquisition on how some companies run retargeting and how retargeting creatives differ from creatives in user acquisition.
Measurement
💎 To measure the impact of retargeting efforts, divide the user base into a control group and a target group and compare the average revenue per user and conversion rate. Adikteev provides verifiable logs and formulas to ensure the accuracy of the results.
(06:20 by Kate Lovejoy)
This seems logical, but I feel like, too often, people don’t try to measure the real incrementality of retargeting and re-engagement efforts.
How creatives differ
💎 In retargeting, you know the user has played the game before, so your creatives should focus on generating a “gentle” brand reminder and FOMO that can bring users back to the app.
(08:40 by Kate Lovejoy)
💎 Retargeting is also a great opportunity to promote new content (e.g., new levels, new characters) and welcome-back offers.
(09:45 by Ava Savitsky)
💎 Static ads are very good to present specific offers, something you want users to get immediately. Video is a better way to showcase new content.
(09:57 by Ava Savitsky)
Here are some (fake) example of how a static offer could look (like in emails, the surprise element often helps with click rate):
💎 Retargeting ads can show up in a lot of different places. Unlike UA, where you focus on having content native/specific to each channel, in retargeting, you focus more on the moment in the user journey rather than the context in which it’s shown. Example: if a user reaches level 60, you want to show elements from more advanced levels than if a user is a beginner.
(11:45 by Kate Lovejoy)
I couldn’t get an example as specific as the above, but here are other (fake) ones about unlocking a new character.
Focusing more on the moment in the user journey sounds very compelling. Unfortunately, if we can’t even target people that have already downloaded the app with SKAN 5.0, targeting them based on their progress in the app feels like wishful thinking (SKAN 10.0?).
💎 In retargeting, you must cover all formats (video, static) because you are reaching users in many different places. Example: weather apps often only have static ads enabled.
(13:15 by Ava Savitsky)
💎 The most successful advertisers using retargeting have a two-prong strategy:
Evergreen content (e.g., based on users’ level, etc.)
Campaigns around particular promotions
(15:00 by Kate Lovejoy)
💎 The groups you’re going after with retargeting are so niche that it’s unlikely you’ll need different creatives on iOS and Android. Users' past behavior and actions are more important for retargeting and predicting success than the type of device or platform. Example: when retargeting users that have spent $50-$60 in-app, it doesn’t matter which platform they’re on.
(22:05 by Kate Lovejoy and Warren Woodward)
💎 Experimenting with intentionally off-putting ads created with AI is proving effective in grabbing people's attention, and it will be interesting to see how the style evolves over the years.
(30:20 by Kate Lovejoy and Warren Woodward)
This insight goes beyond the retargeting topic, but it was a good callout. Manson Chen has been sharing several attention-grabbing visual hooks generated with AI (on LinkedIn and in his newsletter). I’m not sure they match the “off-putting” definition, but it’s pretty close (you’ll need to read this on the web).
Targeting
💎 For iOS retargeting, Adikteev only goes after users sharing device ids and users that have opted-in ATT. With this approach, they’re seeing very similar results to what they were seeing pre-ATT.
(23:45 by Kate Lovejoy)
If you don’t want to wait for a SKAN solution, check your opt-in rates to see if this is worth considering. Here is a pretty recent benchmark from Adjust.
💎 A lot of advertisers used to use rule-based logic (e.g., lapsed 7 days, lapsed 8 days, etc.). It’s a good proxy, but it can perform better to do some level of churn prediction at the game level (low, medium, high) using in-game events (openings, levels users are getting to, etc.).
(25:49 by Kate Lovejoy)
🎁 Bonus Gem
💎 Re-engagement is the area that was the most hit by ATT. Removing deferred deep linking didn’t make sense then because it hurts the experience. With SKAN 5.0, we can speculate that Apple is considering providing a privacy-centric solution for re-engagement and possibly retargeting.
Jihyo Kim (Director of Partnerships at Moloco) in Mastering ROI for iOS Apps Advanced UA Strategies and Hybrid Measurement
And before I leave, here is a quote on the Sub Club podcast on not forgetting that the platforms’ goal is to make money:
“The unit economics have to work for Facebook and TikTok, and you’re in competition with millions of advertisers.” - Thomas Petit (Growth Consultant)
See you next time. Stay curious!
⛏️ Sylvain
🔗 Sources: