How to run a market-position review that changes decisions

Guide · Continuous Intelligence · 4 min read · last verified 2026-07-21

A market position review is a scheduled, fixed-method comparison between the market's current observable state and a previously recorded picture of the same market. The output is not a description of where things stand — a single snapshot supplies that. The output is a set of differences from the last recorded state, each with evidence attached and a decision written against it.

Fix the method before running the first one

Nothing in a first review is comparable to anything, which makes the first run purely an act of construction. Three artifacts have to exist before the review is repeatable.

A frozen question set. Between eight and twenty questions covering position, alternatives, category language, and the objections that recur in sales conversations. These become the benchmark question set, and their wording is now fixed. Improving a question is not free: it severs comparability with every prior reading of that question.

A defined scope. Which competitors, which surfaces, which segments, which geographies. Scope creep between reviews is the most frequent cause of uninterpretable results, because coverage change and market change look identical in the output.

A recorded starting state. The first run establishes the intelligence baseline — not a judgment about whether the position is good, just a dated record of what was observable, with the raw material kept.

Step 1: Re-run the fixed questions

Ask the same questions, of the same surfaces, in the same order, inside a window of the same length. Resist mid-run edits. A question that reads awkwardly should be logged for the next method revision rather than fixed in flight, because the alternative is a review whose results cannot be compared to the one before it.

Step 2: Capture evidence, not conclusions

Record what was observed, not what it seemed to mean. That means the actual text, the actual page, the actual response, with the date and source of capture. The evidence trail is what makes a review defensible three months later when someone asks whether a claimed change was real.

A conclusion recorded without its evidence cannot be re-examined. It can only be believed or disbelieved, and by the next review nobody will remember which it was.

Step 3: Compute differences against the prior reading

Compare against the previous review, not against an impression. For each question, three outcomes are possible: no change, a change in degree, or a change in presence. Presence changes — something appeared, something vanished — are usually the most informative and the most likely to be lost in a summary that reports averages.

Before treating any difference as real, check it against the known variation of the measurement itself. A small movement that the system produces on its own is not a finding.

Step 4: Separate observation from interpretation

Run this as two distinct passes, ideally with a visible break between them.

The first pass states only what changed, using the evidence, with no explanation permitted. The second pass proposes causes. Keeping them separate prevents the standard failure in which an appealing explanation gets attached early and the group then interprets the remaining evidence to fit it.

For each proposed cause, write down what would confirm it and what would eliminate it. Causes that fail this test are guesses, and labeling them as such is more useful than dressing them up.

Step 5: Decide, and record the decision as data

Every difference resolves into one of four dispositions:

• Act — a specific change, with an owner and a date
• Investigate — the observation matters but the cause is unclear; commission the narrower work
• Watch — real but too small or too new; name what threshold would escalate it
• Dismiss — within noise, or irrelevant to current strategy; record the reasoning

Recording dismissals matters more than it appears. When the same signal reappears three reviews later, the prior reasoning is what turns a repeated observation into a pattern rather than a rediscovery.

Where an observation touches how the organization describes itself, it belongs in the positioning conversation rather than the operational one, which is the boundary how intelligence informs positioning works through.

Step 6: Set the next window before closing this one

Name the date of the next review before the current one ends. An interval decided in advance is a method; an interval decided when someone gets around to it is a mood, and it drifts toward whenever the market last felt alarming — precisely the sampling bias that makes a series uninterpretable.

What the review costs

Most of the effort sits in capture, which is mechanical and delegable. Interpretation is short when the capture is good — an hour or two for a scoped review, most of it spent looking at evidence rather than debating from memory. Reviews that run long are almost always reviews where the evidence was not captured and the group is reconstructing rather than reading.

Failure modes to watch for

• Changing questions between runs, then reading the resulting movement as market change
• Expanding scope quietly, so more coverage reads as more activity
• Recording conclusions without the underlying material
• Skipping a review during a busy period, which puts an unexplained gap in the series
• Ending without dispositions, so the review produces awareness and no decisions

What to do next

Write the question set and freeze it. Run once to establish a starting state and accept that this run produces no findings. Schedule the second review at an interval matched to how fast the market moves, and hold the method steady until there is a specific, documented reason to change it — then run both methods in parallel for one interval so the series survives the change.