The Final Hurdle of Subscription: The Paywall

A hurdle you can overcome more easily than you think with an inductive approach

Once you've roughly figured out how to structure your subscription products (duration, price, etc.) and when to prompt users to subscribe, it's time to design the paywall that appears at that exact moment. The paywall is literally where the user makes a purchase. We need to make sure that when they visit the paywall, they can buy without any hesitation.

Like any other product spec, should we dig up problems from our users' VoC, user interviews, and product data to plan this bottom-up? Or should we just slap on elements that are supposedly good for subscription conversion based on endless rumors? Neither.

Unlike other screens in the product, the paywall is a screen where optimization from the supplier's perspective is crucial. Department stores lacking windows and clocks wasn't a design requested by consumers. The fact that many price tags in the market say 9,900 won instead of 10,000 won wasn't what consumers asked for, either. The paywall is the same. You have to find the optimal design that leverages the consumer psychology of potential subscription users from the supplier's point of view, rather than the consumer's.

Finding a seed from those rumors might be one way to go about it, but there's a high chance it won't fit our product. The right approach is to run at least a simple experiment to see how much it actually helps with subscription conversion.

Let's take a look at how to find the perfect paywall for our product's subscription.

Taking an inductive approach

The paywall is a great surface for inductively combining optimal variables.

1. Since usually all users who first install the app are forced to go through it, it's a screen where experiment sample sizes build up quickly.

2. Even if you run experiments frequently, it's a screen that has less impact on the app's overall core usability.

3. Experiments on the paywall are tasks that require very little design and engineering effort.

Frankly speaking, if you validate only the title in one sprint, only the main image (illustration/image/photo) in the next, and then the appeal method (user reviews, number emphasis, etc.) in the next... you can roughly find the optimal paywall through about 6 experiments within a single quarter. If you discover 3 winners and improve the subscription conversion rate by just 30% to 50% each time, you can create more than a 2x increase in conversion rate. (The less growth optimization you've done, the higher the experiment hit rate and the larger the increment per win will be.) Our product also found its optimal paywall by going through a cumulative 30 or so test groups over the past year.

Instead of racking your brain and spending a month carefully planning the perfect paywall, it's a surface where running a simple experiment even a day earlier is more efficient. Of course, the important thing here is that whether each experiment wins or loses, you must design it by well-controlling the variables so that learnings can accumulate properly. That way, existing growth increments won't be lost and can steadily stack up.

Running unorthodox experiments in parallel

If you keep accumulating learnings like this, you might accidentally fall into the trap of local optimization. This means there might be another path to create bigger growth, but you end up just staying in the alley you've already entered and settling for optimization within it. If there comes a point where it becomes hard to generate even a 10% increment from paywall experiments, it's safe to say it's about time to boldly try a completely different paywall.

Looking back, our product's paywall has had steady growth experiments while at the same time seeing unorthodox changes periodically. We probably attempted changes more frequently, but a winner emerging from these unorthodox test groups seems to happen on about a 1-year cycle. The test group located on the far right was an unorthodox one we really had no confidence in, but it became a thankful winner that delivered a 1.4x conversion rate compared to the existing control group. Now, it proudly stands as our strong control group, and new growth specs are underway based on it.

It's not easy to derive multiple test groups for this kind of unorthodox strategy, so at times like this, we usually reference the paywalls of other companies that are doing growth well. Sometimes we get fresh ideas we hadn't even thought of, but more often, it's a "dark under the lamp" situation where we discover elements that make us think, "We should've tested this ages ago!"

On the other hand, during our research, we often find paywalls that seem to have referenced our product's paywall. (Sometimes we spot the September 2021 version of Alarmy's paywall, and other times the early 2023 version.) It makes me feel secretly proud, but also a bit worried whether they thoroughly tested it before internalizing it... Anyway, I hope it helped with their subscription conversion.

Even after testing so many different groups, our paywall-related backlog is still stacked. Unfortunately, because the cumulative growth increment is already substantial, the expected impact of subsequent backlogs isn't that big. Recently, the impact of product specs outside the paywall area has been much larger, so it's naturally getting pushed down in priority.

Still, the fact that the paywall is a screen where you have nothing to lose by testing, and that it requires little effort, makes it quite a shame to just leave it as is. It's the kind of screen I'd love to modularize and automate just to run continuous tests... (If anyone is already efficiently optimizing their paywall in their own way, please share your learnings..!)

If the paywall in your current product is being neglected without any testing, I highly recommend running even a simple experiment like changing the title, or a quick test to change the main color. A seed to increase your subscription conversion rate more easily than you think might be right around the corner..!