Why Retention Is Important in Ad Monetization (feat. Ad LTV) | DARO

Putting an ad placement here... wouldn't that be a terrible user experience...?

When I look at this year's revenue targets, it hits me that our current ad placements just aren't going to cut it. Fiddling around with the app, I can't help but think, How great would it be to just sneak one little ad placement down here? We've built up a decent user base by now—would they really hate it that much if we added just this one? After firing up Excel and running a rough revenue simulation, this non-existent ad placement already starts feeling like my own child. With that energy, I go into a meeting with the product team, passionately pitch the projected revenue, only to hear back: "That's a really bad user experience. Can we rethink this?"

I'm not trying to take sides, but if you work at a company of a certain size, you've probably found yourself in this exact situation at least once. In previous posts, we've shared our thoughts on balancing UX and revenue ( The Ad Monetization Dilemma: Balancing UX and Ad Revenue) as well as a strategic decision we made to temporarily sacrifice revenue to protect Alarmy's product experience ( Why We Chose Usability Over a 2x Revenue Increase).

Typically, when running an experiment on adding ads to the user journey, we look at user retention alongside revenue as a metric. We measure whether the added ad disrupts the user's flow and whether users are feeling so inconvenienced that they actually stop returning to the service, and then make our decision. Today, I want to expand on this concept and discuss why retention and LTV are crucial for ad monetization, and how we can go about measuring them.

‍

Should We Bring These Users In or Not?

Ad LTV = Lifetime (days) * Daily Ad Revenue (ARPDAU)

LTV stands for Lifetime Value, which is a metric representing the revenue generated by a customer from the moment they enter the service until they churn. When calculating ad LTV, you can measure it by converting the user's lifetime into days and multiplying it by the daily ad revenue (ARPDAU).

Once you start measuring ad LTV and monitoring ad revenue this way, you gain the benefits of 1) measuring the relationship between ad revenue and service usability in a single metric, and 2) evaluating the value of user acquisition much more precisely.

When analyzing experiment results with revenue as the core metric and retention as a guardrail metric, we often fall into a dilemma over the tradeoff between revenue and retention. It would be fantastic if ad revenue went up while retention stayed exactly the same. In reality, if ad revenue jumps by 10% but retention drops by 3%p, should we keep the treatment group? How much of a revenue increase is equivalent to a 1%p difference in retention? These are the kinds of questions that pop up.

Lifetime = SUM(D+N Day Retention)

LTV helps navigate this dilemma much more easily. The "Lifetime" in LTV refers to the average lifecycle, and if you are already measuring daily retention, you can substitute the average lifecycle with the sum of the average daily retention. In other words, if you can calculate the difference by deriving Lifetime using overall retention—rather than just retention over a specific period—you can compare the delta in Lifetime with the delta in ARPDAU to gauge whether the user's actual LTV is increasing.

ROAS = LTV / CAC

Furthermore, once you start measuring the ad revenue generated over a user's lifecycle, you can also build an efficient User Acquisition strategy. By dividing LTV by CAC (Customer Acquisition Cost), you can quantitatively determine whether you are taking a loss or making a profit when bringing in a user based on current standards. Also, if you segment and monitor this by channel or country, you can formulate specific hypotheses about which countries you should focus on, or which ones require action to boost daily ad revenue.

For instance, if ROAS is high, you could take action to drive more traffic from that country. On the flip side, if ROAS is low but CAC is also low—meaning user acquisition is relatively easy—you might try more aggressive ad placements to boost LTV.

‍

Wait, How Do You Even Calculate Ad LTV...?

Ad LTV = Lifetime (days) * Daily Ad Revenue (ARPDAU)

The tricky part of measuring ad LTV is the Lifetime aspect. Even if you try to shorten the measurement window to one or two months, you still have to wait out that entire period just to check the retention metric, which slows everything down. This makes the metric less actionable, leaving you with that nagging feeling that you're just staring at a lagging indicator.

To maximize the timeframe covered by approaching Lifetime as SUM(D+N Day retention) while minimizing the waiting time for measurement, I recommend calculating and using predicted retention values. You can easily do this using the curve_fit function provided by scipy, or by building your own model using Linear Regression, Tree-based, or Deep Learning algorithms. The key here is to identify the points where your service's retention curve converges and find a function or model that can predict those specific points as accurately as possible.

For example, if you have a service where retention drops sharply up to D7, flattens out starting at D14, and remains consistent after D28, it's crucial to ensure that your chosen model accurately predicts the data points at D7 and D14, and also projects retention to stay flat after D28.

While the timeframe will vary depending on the model, with a bit of refinement, you can look just at D7 or D14 and still predict the sum up to D365 quite accurately. This allows you to calculate a much longer lifecycle using a relatively small number of observations.

An additional method I suggest is measuring retention based on the occurrence date (green) rather than by cohort (red). You can think of this as a way to look at more actionable metrics, as mentioned earlier. If you wait seven days based on a cohort and measure LTV using retention up to D7, you're essentially looking at a week-old metric, which inevitably makes you think, Isn't it already too late to take action?

By tweaking this just a bit and calculating LTV every single day using the D1 to D7 retention corresponding to that specific date, you can derive LTV using the freshest retention metrics available. Whenever there's a shift in LTV, you can analyze the variance across each retention point in the same way to pinpoint the cause. At first glance, you might think, Isn't this just six of one, half a dozen of the other? But from the perspective of both delivering and receiving metrics, knowing that you are referencing the most up-to-date data significantly increases the chances of generating hypotheses that actually lead to action.

‍

Ad Revenue and Service Usability

As I mentioned earlier, ad revenue and service usability are like the chicken and the egg—inseparable no matter how hard you try. In situations where you constantly have to weigh the tradeoff between the two, LTV is a metric that can serve as an excellent compass. At first glance, it might seem tough to calculate consistently and tricky to monitor regularly. However, if you pool your efforts and start measuring it, I can confidently say it's a fantastic metric that not only elevates your thinking around pure revenue and usability but also refines your approach to user acquisition, empowering you to make focused and strategic choices.

‍

⏰ Curious about DelightRoom's secret to ad monetization?

‍

Originally published at https://daro.so.