Face recognition product launch experiment
We made a product experiment and want to describe how we done it and what we have learned.
Why face recognition?
In recent years I and my colleague noticed several trends
- face recognition technologies advanced to a degree when it’s safe to use them in payments operations, surveillance, and so on
- there were several apps dedicated to search-by-face like SearchFace, Verasity, TinEye which were viral (but then were blocked for privacy issues)
I was curious — is there a niche for a new product? Is there a demand and a way to fulfill it?
So we decided to make a product experiment to explore this a little bit.
Product experiment
We are using Osterwalder’s framework for the product experiments to define and track what we are doing. So, that’s how we defined an experiment:
We believe that users want to find someone by photo so desperately, that they will pay for the app before the search results and will pass through complex boarding
To verify that we will make an MVP with search-by-face technology and will include an in-app subscription at the boarding phase, right after the install
And measure organic traffic from the Appstore, cost per install, % of users who purchased an in-app purchase.
We are right if there will be users who install the app and pay for the subscription and we will get relevant feedback from users.
Minimal Viable Product
We have a limited set of requirements for the MVP
- being able to pass AppStore review
- provide analytics relevant to the experiment setup
- theoretically give a value (one person can find another) for user, so we can market it
- retain users for future development and allow them to post feedback to us
The app is now available on the Appstore: https://apps.apple.com/us/app/pipl-find-and-text-friends/id1212534307
What have we learned?
There’s a strong demand for such app in the AppStore in several countries
Appstore impressions by territory without much ASO efforts, just “search by face” gives relevant traffic from Russia & the USA
7% conversion is in the top 70 percentile for this metric.
8.38% of users who installed the app made a subscription
Our conclusions
- There is a strong demand to search by face app
- There is a strong privacy concern — the search database should be crafted from audience’s photos, not third-party sources
- We should try to process images on phone directly and work only with hashes to provide a private way of handling images.
Our next steps
As a next step we are trying to provide real value for users — in order to do this, we are trying to move recognition technology to the mobile app itself so we will be able to scan the user’s gallery for his friend’s without transferring images to our server. We hope it will allow us to craft a database for search based on audience data.