Product intelligence is the practice of keeping what your users said, what you decided and why, and what you shipped in one record. The point is that every product decision can show its evidence and be checked afterwards. It sits between two older categories: product analytics, which counts what users did, and feedback tools, which collect what they said. It adds the decision and the result.
The definition
A product decision is a claim: build this, for these people, because of this, and expect that to change. Product intelligence is whatever lets you make that claim with evidence and check it later. In practice it is a record with four parts.
- What users said, in their words, with the source and the date. A quote from an email in March is evidence. Your summary of it is not.
- What you believe and how strong the evidence is. A belief with praise behind it is weaker than one with a payment behind it, and the record should say which.
- What you decided and why, including what you chose not to build and what you expected to change.
- What you shipped and what happened, so the prediction can be compared with the result.
The word intelligence is doing a specific job here. It does not mean clever software. It means the kind of record an analyst keeps: sourced, dated, with the confidence stated. Absence is written down as absence. "No evidence yet" is a valid entry.
How it differs from product analytics
Analytics counts behaviour: sign-ups, activation, retention, the funnel. It is precise about what happened and silent about why. Nobody's words are in it. When retention drops, analytics shows the drop; it cannot show the three support emails from the week before that explained it.
Product intelligence keeps the counts and adds the words. It also adds the decision. Analytics has no field for "we expected this to reach 40% by June and here is why". Without that field, a result cannot be a lesson, because nobody wrote down what they expected.
The two work together. Analytics is often the source for the result at the end of a decision. Product intelligence is where the decision, its evidence and its prediction live, so there is something to compare the result against.
How it differs from feedback tools
Feedback tools collect what users said: a portal, a widget, a board with votes. They are good at volume and at sorting. They stop when a request is tagged, counted and put on a roadmap.
Product intelligence treats a tagged request as the start, not the end. Three questions follow. Who asked, and are they the people you build for, or a loud few from outside the segment? What did you decide, and what did you expect to change? Did it work, and were the people who asked told?
Feedback tools also tend to count everything equally. A compliment and a payment are both "positive". A public thread and a paying customer are both "mentions". Product intelligence ranks your own users first, treats praise as the weakest evidence, and shows public sources without counting them.
What an evidence-based product decision needs
Strip the category words away and an evidence-based decision needs six things, whatever tool holds them.
The words, with their source
Exact quotes, with where and when they were said. Summaries drift toward what you wanted to hear.
Counts of your own users
How many of the people you build for have this problem, from which channels, over what period. Public chatter is context, not a vote.
A grade on the evidence
Praise, a past pain, a request for access, time given, an intro made, a payment. The grade is what people did, never what they said.
A written decision with its reason
Including what is out of scope and what would change your mind. Decisions without reasons get re-argued every quarter.
A prediction before the build
The metric, the number you expect and the date. Written first, or the result cannot teach you anything.
A check afterwards
The prediction against what happened, from a source you chose in advance. Then one line about what you learned, kept for the next decision.
Most teams have two or three of these in different places: quotes in a notes app, counts in a spreadsheet, decisions in someone's memory. Product intelligence is the habit of keeping all six together, and the honesty to leave a field blank when there is nothing true to put in it.
Where MindBacklog sits
MindBacklog is the brain for your product. It keeps everything your users tell you, shows you what to build and why, and hands it to your coding agent with the evidence.
The record above is what MindBacklog keeps. Feedback is sorted into problems ranked by who has them, with the quotes behind each. Beliefs sit on an evidence ladder where praise never counts as proof. Decisions are logged with their reasons. Each initiative carries a prediction written before the build, and Results checks it against what happened. The build itself goes to your coding agent as a Build Pack that carries the why and the evidence.
It runs as one loop. Learn: what users said, in their words. Decide: the belief, the evidence grade, the decision with its reason. Build: the Build Pack your coding agent pulls, with the metric and the must-haves. Launch: tell the people who asked, in their words; you send it. Check: the prediction against the result; you decide what it means.
Mind, the AI in MindBacklog, writes the drafts and shows its sources. It says "not ranked" under thin data, "not found in what we read" about a competitor, and "nothing in your data on this yet" when that is true. The verdicts that matter are counted from people, never written by the model. That is the whole point of the practice: the evidence stays visible, and so does its absence.
Read how the loop works, or start with one of the guides: the evidence ladder, spec-driven development for founders, or the product-market fit survey. The glossary defines every term used in the new MindBacklog.
