Index8 gives its customers conclusions about their businesses: what changed, why it matters, what to do next. Conclusions deserve scrutiny. This page explains exactly how ours get made, so you can question any of them and check our work.
The short version: nothing arrives as a verdict. Every conclusion in the workspace carries three things you can inspect: the sources it was built from, a confidence level we are prepared to defend, and the assumptions it rests on. If any of the three looks wrong to you, the conclusion should lose your trust until it earns it back.
- Signals
facts with dates, not opinions.
- Analysis
a question, a hypothesis, and both sides.
- Insights
conclusions with a confidence level we defend.
- Recommendations
three moves, each with its receipts.
- Outcomes
the record grades the whole chain.
Signals: facts with dates, not opinions.
Everything starts with observations. A competitor's pricing page changed. Two reviews named the same complaint. Quote turnaround slipped from two days to four. Each signal is a fact Index8 observed, stamped with when it was observed, where it came from, and how important it looks.
Signals are deliberately boring. They contain no interpretation, because mixing what happened with what it means is how businesses talk themselves into comfortable stories. The register keeps them separate on purpose, and you can read every signal behind any later conclusion.
Analysis: a question, a hypothesis, and both sides.
Related signals become an investigation. An investigation has a research question ('is the rival's promo a price war or a plan funnel?'), a working hypothesis, and, importantly, the evidence on both sides. Supporting signals and contradicting signals sit in the same file.
Keeping the contradicting evidence visible is the discipline that separates analysis from advocacy. A hypothesis that has never faced a contrary signal is not a finding yet; it is a hunch with good posture.
Insights: conclusions with a confidence level we defend.
When an investigation firms up, it becomes an insight: a conclusion stated plainly, with why it matters to this business in particular, and a confidence level of low, medium, or high.
Confidence is a number it defends, not a mood. High confidence means multiple independent signals point the same way and the contrary evidence is thin. Medium means the reading is sound but rests on fewer legs. Low means worth knowing, weakly held. When the evidence is thin, the insight says so before you ask, and every insight lists the assumptions that would break it if they turned out false.
Recommendations: three moves, each with its receipts.
The weekly brief closes with the three actions most likely to matter, ranked by expected impact against effort. Each recommendation carries the insights it came from, an expected impact stated as a number where one exists, a cost and effort estimate, and how the outcome will be measured if you approve it.
Three is a design decision. A list of ten recommendations is a way of recommending nothing. Ranking forces the reasoning to commit, which is also what makes it checkable later.
Outcomes: the record grades the whole chain.
Every recommendation ends in your call: approved, deferred, or declined, with your reasoning attached. Approved actions get measured against the expectation that was written before the outcome existed. Expectation met or missed, the result goes on the record.
This is where the method keeps itself honest. When a prediction misses, that miss is recorded next to the reasoning that produced it, and the lesson feeds the next recommendation. A system that never admits its misses is not reasoning; it is marketing.
How to challenge any conclusion.
Open it. Every insight and recommendation in the workspace unfolds into its evidence. Three questions do the work: Are the sources real and current? Does the confidence match the number of independent legs under the conclusion? Do I know something about my business that breaks a stated assumption?
That third question is the important one. You know things no observer can: the handshake deal, the retiring technician, the customer who is really three customers. When your knowledge breaks an assumption, the conclusion should change, and telling Index8 so is how the picture gets better.
What this method refuses to do.
It refuses to publish statistics it cannot support: no invented benchmarks, no 'businesses like yours' numbers without a named source. It refuses to state conclusions without evidence attached. It refuses to make your decisions: Index8 briefs, you decide, and the calls that need a lawyer, an accountant, or another qualified professional get pointed their way.
These refusals are the product. An analyst you cannot check is just a confident voice, and the world has plenty of those.