Keep, fix or rebuild: how we decide what to do with your app

Every app we review ends up in one of three boxes: keep, fix or rebuild. The free 48-hour verdict tells you which box yours is in, in writing, before you spend anything. This is how we decide.

Keep: it works and it is safe, just hard to change

The data is protected, payments reconcile, and users can do what they came to do. What is missing is the scaffolding that makes change safe: tests, error tracking and a safe way to deploy. In one agency's audit of 21 AI-built apps (AxonBuild, June–July 2026), at least 18 had no working automated test, 17 recorded errors nowhere, and at least 17 shipped every change straight to live users. Those three are the cheapest things to add, and they change day-to-day life with the app the most.

Fix: the core works, specific things are broken

Sign-in, payments, data-access rules or webhooks are broken or unsafe, but the data model underneath is broadly right. We fix those items and leave everything else alone. One item is a Single Fix ($300 to $1,500). The whole fix-now list is a Fix Sprint ($3,500 to $6,000). Both are fixed price.

Rebuild: only when you have to

The data model is wrong, or every fix breaks two more things. Then we rebuild the unsafe core and keep what you learned: the screens, the flows, the product decisions. Rebuilding costs more than fixing on every published price we have seen, so we recommend it only when the evidence leaves no other choice, and we say which evidence.

What the verdict looks like

• One line at the top: keep, fix or rebuild.
• Fix-now items, each with a fixed price.
• Can-wait items, so you can plan them for later.
• The items you can safely fix yourself, and the ones you should not: data access, payments and secret keys.