There is a pattern anyone who has worked the rating layer long enough learns to expect. A team rallies around a slipping rating — a push on support, a wave of fixes, a concerted effort to win back the number. It works. The rating climbs, the chart goes out in the monthly update, attention moves to the next fire. Two quarters later the number is back where it started, and the effort is remembered as something that "didn't really stick."
The effort was real. What was missing was an account of why the rating fell in the first place — and why the forces that pulled it down did not stop operating just because the number moved.
A rating is a flow, not a level
The figure at the top of a listing looks like a level, but it is the surface of a flow: every day, new reviews arrive and shift the mass underneath it. Whether the number holds depends entirely on the balance of that daily inflow — the mix of satisfied and dissatisfied voices, and which of the two is more motivated to speak. Dissatisfaction is almost always the more motivated party. A user whose payment failed writes tonight; a user whose payment worked writes never.
This is why a recovered rating is not a fortified position. It is a swimmer holding ground against a current. The current — the underlying complaint patterns, the support gaps, the friction that generated the negative inflow — keeps flowing unless something changes it. Stop swimming, and the water decides where you drift.
Where the pressure comes back from
Regression rarely announces itself. It comes back through channels that look unrelated. An acquisition push brings in lower-intent users who review more harshly than the early adopters did. A release ships a regression that takes three weeks to surface in the review stream. A pricing change irritates a vocal segment. Seasonality shifts who is using the product and what they expect of it. None of these are rating problems in origin — all of them are rating problems in destination, because the review stream is where every other function's rough edges eventually land.
A one-time recovery has no mechanism for noticing any of this until it is visible in the headline number. And by the time a six-figure review base moves visibly, the inflow that moved it has been running for weeks.
The expensive round trip
This is the arithmetic that makes maintenance the economically serious half of the work. A recovery that is not defended is a round trip: the cost of the climb, then the quiet resumption of the conversion tax, then — if the business still cares — the cost of climbing again, this time against a review base that has grown larger and slower to move. Each cycle is more expensive than the last. On a large base, an unheld recovery is not a partial win. It is a more expensive way to arrive back where you began.
What holding the floor involves
Maintenance is unglamorous by design, which is why it is underbought. It means reading the review signal continuously — not at the point of alarm — so that a shift in complaint mix is caught while it is still a trickle. It means knowing which categories of complaint actually move the rating and which merely make noise, so attention lands where the number is decided. It means keeping review quality ahead of acquisition volume when campaigns scale, so growth does not erode what it is meant to capitalise on. And it means treating the threshold as a floor to be defended month over month, rather than a summit that was reached once.
The question that matters is not whether a rating can be lifted. It usually can. The question is what happens in month four, when attention has moved on and the current is still flowing. Recovery answers for the climb. Only maintenance answers for the floor.