Launching Alarmy's Subscription Model via Growth Hacking

This post is ported from an article originally written on Brunch on November 24, 2019.

We recently launched a new subscription model in Alarmy. I'd like to share this process based on our growth hacking steps.

Source: Sean Ellis, Morgan Brown. Hacking Growth. Currency, 2017

Analysis #1

We learned that Alarmy's core user base consisted of heavy sleepers who were highly satisfied with "sure-fire wake-up" features like various alarm missions and loud ringtones. Accordingly, we decided to structure our subscription model around premium features that could meet these needs.

However, since there were no other services selling alarm features as a subscription (leaving us with nothing to benchmark against), it was tough to gauge the potential of our chosen premium features—in short, whether people would actually pay for an alarm app. It was especially tricky to predict the ROI (Return On Investment), given that it would take about four months to develop and launch the subscription model.

Ideation #1

To figure this out, various ideas were proposed. We brainstormed methods like predicting the business's likelihood of success indirectly by benchmarking similar utility apps or conducting user surveys/interviews. Another idea was to launch a subscription model containing just one or two core premium features first to predict the conversion rate and revenue.

Prioritization #1

We reviewed these options using the Impact, Confidence, and Ease factors of the ICE Score System, which is commonly used for prioritization. The idea we ultimately decided on was to run a smoke test (a test to validate an idea without actually building the product) on our users.

It was an experiment where we didn't build the actual features, but merely set up a purchase page using minimal engineering and design resources to analyze buying behavior. When users attempted to buy, they were informed about the upcoming launch schedule and prompted to enter their email addresses for launch notifications and perks. We also used the collected emails for follow-up surveys to ask users which features they were most excited about and what price they considered reasonable.

Experiment #1 & Analysis #2

The user response (clicks on the purchase button) in this experiment was deemed significant, so we started turning Alarmy's subscription model into a real product. At the same time, we brainstormed ways to use the remaining time before the actual development and launch to build a more accurate and sharper knowledge base for this new business venture.

Our conclusion was that an experiment to determine the product pricing should be the top priority. Product price is the monetary translation of the value proposition it offers to users. A seller sets a price thinking their product is worth that much, and a user buys it when they judge the product to be worth that price. So, how should we set the initial price when launching our product for the first time?

Ideation #2 & Prioritization #2

Ideation and prioritization for this question were decided relatively easily. The best way to find the optimal price is to ask actual users and set it accordingly, and luckily, Alarmy had already built a sandbox (smoke test) to ask exactly that.

To measure the conversion rate more accurately than in Experiment #1, we polished just the purchase page to a production-level design. We then ran an A/B test with three different price points as our experimental groups.

Experiment #2 & Analysis #3

For the three price groups A, B, and C (in order of price), the conversion rate and revenue (conversion rate * price) results came out as follows.

- Conversion rate: A = B > C

- Revenue: A < B < C

Based on these results, we conducted additional experiments with new price groups X and Y (where A < B < X < C < Y in order of price). By selecting the final winner X, we were able to optimize our pricing before the official launch.

Currently, the Alarmy subscription model has been launched and is live on iOS (the Android version is scheduled for release by the end of 2019). Analyzing the actual metrics post-launch showed that they were similar to the smoke test data, which confirmed the validity of our testing method.

For reference, we couldn't obtain data for the "purchase button click → actual payment rate" step via the smoke test. For this part, we referred to benchmark numbers for different subscription types (Monthly, Yearly) found in various open case studies and analysis reports. The numbers didn't stray far from those benchmarks, allowing us to predict the conversion rate relatively accurately.

DelightRoom is always hiring for various positions, so we'd love your interest and applications.