3 Sep 2016

App Stores Should Automatically Optimise Revenue For Developers

If you're selling an app, it's a good idea to A/B test different prices to see which one brings in the most revenue. App stores should offer to take care of this for you. Instead of setting a fixed price, like $2, you would leave that to the store. The store would A/B test various prices like $1, 2, 5, 10, 20 and 40 [1]. It would then pick the one that has resulted in the most revenue. It would email you a report telling you what it tried, what the results have been, and what the best price turned out to be.

It would then repeat this experiment every quarter, adjusting the price as the market changes. This is something a human may forget, or not be able to find time for among the umpteen things one needs to do to make an app.

The optimal price may also be different in different countries. Maybe lower in India than the US, given lower purchasing power in India. You can't manually run 150 A/B tests, one for each country, but an automated system can [2] [3] [4].

In addition to testing different prices, the store can also A/B test different payment models like a one-time payment vs a subscription, to see which makes more money. And subscriptions of various durations like monthly, bimonthly, quarterly, half-yearly and yearly. The store can A/B test all of them, different combinations of price and duration [5].

You as a developer would be able to make some decisions while leaving others to the store. Say you've decided that you want annual subscriptions. You could then leave it to the store to test different subscription prices to see which one works the best, and how to vary prices for each country. Or if you know you want subscriptions, you can also leave the duration to the store to figure out.

Ads are another important payment method. The app store can test to see if ad revenue is more than explicit payments. In fact, it can test three options: ads only, paid only, or leaving it to the user's choice to either pay or see ads. All three possibilities can be A/B tested by the app store. This requires an ad library made by the platform owner, like iAd [6]. Before you add the ad view to your UI, you'd check if it's enabled for this user:

if (iAd.isEnabledForThisUser()) parentView.add(new iAdView());

That's all it should take to support both ads and payments in your app, compared to supporting only ads.

In summary, app stores should work on developers' behalf to maximise income by automatically A/B testing various prices and pricing models like free, one-time payment and subscriptions.

[1] Powers of 2 are a good idea, since they cover the greatest range with the least data points.

[2] Alternatively, app stores should offer APIs so that third-party services can do the A/B testing for you. But since maximising revenue is something most developers care about, it should be a part of the store itself. It's not a niche service.

[3] If you're testing 6 prices over 150 countries, that's 900 prices, way more than it makes sense for a human to test.

[4] Country is only one factor that may affect the optimum price. Maybe it makes more sense to charge a lower price for older models, on the theory that those users wouldn't be willing to pay a higher price? That is, charge iPhone 6s users $2, but iPhone 5s users only $1. Or offer a lower price to someone who has a 16GB iPhone than a 64 one, on the theory that the latter is less price-sensitive?

[5] If you have six prices and five subscription durations, that's 30 combinations to test.

[6] Which unfortunately has shut down.

No comments:

Post a Comment