Are you accidentally killing your ROAS? Part 1: Treating all networks the same

On paper, it looks like scale. In reality, it was a slow leak worth $4M. We've seen a studio spend $11M, chasing 100% ROAS at D365. Fast forward 6-months and ROAS flatlined at 65%.

Are you accidentally killing your ROAS?

On paper, it looks like scale. In reality, it was a slow leak worth $4M. We've seen a studio spend $11M, chasing 100% ROAS at D365. Fast forward 6-months and ROAS flatlined at 65%. The result wasn't just missed targets; it triggered an 80% drop in monthly spend from their peak.

So why did this happen? And how can you avoid it? That's why we sat down with Deconstructor of Fun and discussed practical solutions so you don't kill your ROAS.

In this series we'll discuss 3 problems:

1. Assuming networks will all behave similarly → today's article
2. Managing organic installs in analysis
3. Managing the ad network algorithm

The problem

UA teams frequently fall into the trap of evaluating new acquisition initiatives (such as new networks or geographies) using ratios or extrapolations derived from existing UA performance, leading to misleading conclusions.

In many cases, this behavior is well-intentioned. In the absence of reliable data on how a new channel will perform, UA teams make the best possible decisions using the tools and signals available to them.

Why it matters

• New channels can diverge materially in long-term performance (e.g. 90+ days post-install) with little or no early warning
• As a result, UA spend can be effectively misallocated for months, with failure only becoming visible once the initial cohorts fully mature.

What is the solution

• Overall, you cannot avoid all risk, but the aim should be to at least understand, quantify and manage the amount of risk that your UA strategy is taking on
• Beware of the most common change types; new networks, new countries, changing campaign types and creative styles
• Speak to your account managers before you start a new network, and ask for example ROAS curves for games in your genre.
• When launching, start by focusing on short term targets, and increase spend only as your window of knowledge increases
• Set aside specific test budgets where all stakeholders internally understand and agree that less than 100% ROAS is expected while ad networks "learn" your target audience
• Try to keep all other factors the same — when launching on a new network, use your best performing creative assets
• If you are already using predictive analytics, make sure you understand the confidence intervals for a new network

If you find yourself struggling to solve the above, we'd be happy to help. Kohort works with studios to build predictive models that contain all the nuances of every campaign, country and network. Our models have been honed on over $6bn in UA spend.