Automated insight: optimized SKAN conversion models alert you to smarter SKAN set-ups

If today’s AI is so smart, why doesn’t it do more of your work? Especially with SKAN conversion models, which are notoriously challenging to get right.

Welcome to optimized conversion models for SKAdNetwork from Singular.

Why do you need optimized conversion models for SKAN?

• 35% of apps are not using the most optimized model
• 20% of apps are using a significantly inefficient model
• Using a poorly optimized model costs you missing data … for 7-day revenue, sure but even for 1-day revenue (!!!)
• Singular’s optimized SKAN conversion models technology has found up to 60% better conversion models for some clients
• And … it’s all automated

Artificial intelligence (or, as I like to call it, the machine doing your work for you) should actually be useful. After all, the Industrial Revolution was 200 years ago. The Intelligence Revolution, Forbes says, is here. Shouldn’t your technology do more to help, especially with something as challenging, finicky, and hard-to-figure-out as iOS attribution via Apple’s SKAN framework?

SKAN conversion models that work: it all starts with accurate data

You already know: data is hard to get under SKAdNetwork. 

First off, full, accurate, and complete data is literally impossible to get, thanks to privacy thresholds and the single postback that SKAN 3 offers, or the 3 that SKAN 4 will offer. So you need SKAN Advanced Analytics to model missing data, enrich insights with data from privacy-safe sources, make revenue predictions, and — bottom line — wring every last ounce of value out of the data you actually do receive.

But you also need to ensure that the baseline data SKAN does provide is the best, richest, and most accurate it can possibly be.

And that’s where optimized SKAN conversion models come in.

“The key component to getting accurate data, whether it’s revenue or whether it’s modeled revenue, is having the correct model,” says Singular product manager Evyatar Ram in a recent Growth Masterminds podcast. “Everyone understands that for modeled revenue, if you’re trying to get 7 day predictive, you need a super optimized model. But I think the interesting thing here is that even if you’re just trying to get accurate 1 day revenue that’s fully encoded in SKAdNetwork, even for that use case selecting the right conversion model impacts the accuracy of the number that you will get at the end of the day.”

We’ve already seen that only 9% of mobile marketers think they have SKAN figured out.

• What if you have the wrong model? 
• What if there’s a more optimized model that captures more data?
• What if you could get greater accuracy?

And, if you have the right model, the most optimized model with the highest level of accuracy possible … how might that improve campaign optimization, spend allocation, partner selection, creative optimization, and all the other things that allow mobile marketers to deliver outsized growth?

But perhaps the more important question is: how would you know? Or even: could you find out if you’re actually using the most optimal model without a long, difficult, and expensive (opportunity cost!) round of testing?

Cue the machine.

Automatic always-on background SKAN conversion model optimization

Singular has developed technology to rank SKAN conversion models for effectiveness: their ability to deliver accurate data, Ram says. With always-on SKAN conversion model optimization checking in the background, you’ll now get a notification if there’s potentially a more accurate model to try.

“For every app out there, we actually check the data once a week,” Ram says. “So every week we’re going to see if you have a better model or not. We will look at our most optimal model, and we’ll look at what model the customer already has … if we think that our model is significantly better than the one that you might be using, we’ll recommend it.”

In other words, if it’s 2% better, the technology won’t waste your time.

But if there’s a 20% or 40% or 60% better option, you’ll get a notification.

The benefit here is that …

• as the world changes (economy, politics, pandemic) 
• as your app changes  (UX, UI, monetization, capability, onboarding)
• as your marketing changes (partners, creative, cohorts, offers, spend, bids)

… you don’t have to worry that your conversion model for getting attribution and conversion data back from SKAdNetwork is out of date, wrong, or unoptimized.

You’ll be getting automated checks on it weekly. The machine has your back.

SKAN 4: automated optimization will be even more important

And what about SKAN 4?

Automated optimization technology will be even more important there, because you’ll potentially have up to 4 separate conversion models for SKAN operational at any given time:

1. Full granularity model if you surpass crowd anonymity thresholds for postback 1
2. Coarse conversion model for postback 1 if you don’t surpass crowd anonymity thresholds
3. Coarse conversion model for postback 2
4. Coarse conversion model for postback 3

“SKAN 4 does a lot of really great things, but in some respects, managing your SKAN set-up in SKAN 4 might become more complex,” Ram says.

Ram is being diplomatic. SKAN 4 simply will be more complex to implement: there’s more to instrument, and there’s more to analyze. Technology like this, however, will help marketers let MMPs like Singular do all the heavy lifting.

Revenue models first, mixed models next

There is a caveat here. Optimized conversion models for SKAN is initially only available for revenue models.

There are super-useful mixed conversion models that combine multiple factors like revenue, events, engagement, or funnel progression. The optimized conversion model technology isn’t available for that yet, but it is coming. 

“That’s definitely something that’s on our roadmap and hopefully we’ll have some things to share soon,” Ram says.

Optimized SKAN models matter

It can’t come soon enough. 

As I mentioned off the top, 35% of apps Singular has checked are NOT using the most optimized SKAN conversion models, and 20% are using highly inefficient models. That means 1 in 3 apps could get better, and 1 in 5 apps are really missing out on data accuracy, severely impairing their ability to advertise, market, and grow effectively.

Plus, for some apps, Singular’s new optimized conversion models technology for SKAN has found a 60% better model. That’s not small. That’s not insignificant.

That’s simply game-changing.

“At the end of the day, it’ll help you make better decisions,” Ram says.

Subscribe to Growth Masterminds

We talk to industry leaders about key changes happening in the mobile adtech industry.

Tune in by subscribing to our YouTube channel here, or by subscribing to our audio podcast on pretty much every platform podcasts are available on. You can find more details on Growth Masterminds here.