While building Mengya, one set of numbers stayed with me.
Among 72 registered users, 49 actually began a conversation with the AI, and 18 continued for more than ten turns. In interviews, one person said that their scattered experience and abilities had been organized for the first time. Another used a gentler phrase: I felt seen.
If we looked only at those moments, the product seemed to have found its value. But the next day, only 15% of users returned, far below the 30% target we had set in advance.
Why was one good conversation not enough?
Being moved and forming a habit are different
A first experience answers one question: did this create a moment of value for me?
Continued use asks something else. Why should I open it tomorrow? Does it correspond to a problem that will occur again? When I return, will I know immediately what to do?
Mengya’s clearest value was helping users extract skills, organize experience and see possible directions for growth through conversation. That moment could be moving, but career reflection is not a high-frequency daily need. Someone might talk deeply on a confused evening, then wake up with no action that requires the product again.
Conversation also has a cost. A user has to organize their thoughts, explain their experience and allow the AI to ask questions layer by layer. For someone with a clear goal, depth is valuable. For someone who only wants to try the product, it can become a burden.
An Aha Moment therefore proves that value exists. It does not prove retention.
Retention is not reminding someone to return. It is leaving a relationship unfinished.
When retention is weak, it is tempting to add reminders, check-ins, points or more features. But if a user has no second reason to open the product, those changes only make a one-time experience noisier.
I began to care about another question: when a user leaves, are they carrying something that still needs to be completed?
It might be tomorrow’s task, a plan that has not yet been finished, feedback that updates as they act or a system that understands them better only through continued use. The real reason to return is not that the product calls the user back. It is that something in their life remains open.
Mengya’s interviews revealed another problem. Users did not only lack an analysis of themselves. They lacked a way to carry a plan into everyday action. That insight later inspired a separate product, Snowball, which shifted the center from “seeing yourself” to goals, plans, daily tasks and feedback.
Snowball was not the next version of Mengya. Its definition, architecture and technical implementation began again. The only thing inherited was an insight: users do not only want to be understood. They also need support in moving forward.
Some products do not need high frequency, but they must define themselves honestly
Low retention does not always mean failure.
If a tool exists to revise a résumé once, explain a medical report or create a travel plan, a user leaving after finishing may mean the problem was solved. Forcing daily activity on such a product can distort its purpose.
Mengya, however, was not intended to produce a one-time ability report. It promised an ongoing relationship with growth. Under that promise, 15% next-day retention was not a number we could explain away. It meant acquisition and first value had been validated, while continued use had not.
So we closed the product.
I am now grateful for the result precisely because it was not beautiful. It taught me that validation is not the art of explaining every number as success. It is the willingness to stop when the evidence does not support the assumption.
A good conversation can make someone remember you.
But a product becomes part of life only when it enters a problem that truly recurs and leaves behind a clear next step.