Why must the baseline questions stay locked?

Attribution. If questions change between scans, a visibility shift cannot be traced to the market versus the methodology. The lock keeps every comparison like-for-like.

Can I add new questions after the baseline is set?

Yes — as a separate, parallel set with its own start date. The original lock stays intact, so history remains comparable while the new set begins accruing its own.

How often should a baseline be re-scanned?

It is a cadence decision, not part of the definition — the pressure comes from launches, active pushes, and how fast your market's evidence moves. Plan cadences are published on /pricing.

What is an Intelligence Baseline? A practical definition

Q: What is an intelligence baseline?

The fixed reference a first scan establishes: a locked set of real buyer questions plus the sourced record of who appears in the evidence behind each, dated. Later scans compare against it, so change reflects the market rather than the measurement.

Glossary · Continuous Intelligence · 4 min read · last verified 2026-08-11

Reviewed before publication Editorial board Independent commercial review

In shortAn intelligence baseline is the locked set of benchmark buyer questions a first scan establishes — the fixed reference every later scan is measured against. What goes in, what movement means, and how baselines go wrong.

An intelligence baseline is the fixed reference point a first market scan establishes: a locked set of real buyer questions, plus the record of which companies appear — with sources — in the public pages and AI answers behind each one. Every later scan is compared against it. The lock is the entire point, and it is worth stating precisely once: because the questions do not change between scans, any change in the results reflects movement in the market's public evidence, not a change in what was measured. Everything else about baselines is elaboration on that single principle.

What goes into a baseline

Three components, each with a quality bar:

The question set. Real buyer questions — the ones asked before choosing in a market, phrased the way buyers phrase them. Not keywords, which are proxies, and not questions built around your own brand name, which flatter rather than measure; what is a benchmark question set covers construction.
The presence record. For each question: which companies appear in the pages and answers behind it, recorded with a link for every appearance so any entry can be re-opened and checked. Where nothing is found, the record says so explicitly — an honest null is data.
The date. A baseline is a dated artifact. Undated, it degenerates into a general impression of the market; dated, it is the fixed point that gives every later delta its meaning.

Private data and internal assumptions stay out, because a baseline must be checkable by someone who does not trust its author.

What movement against a baseline means

Once the baseline exists, re-scanning the same locked set turns observations into attributable deltas: a company now appears for a question where it did not, holds its ground, or drops off — and each delta traces to specific pages that changed. This is what makes the baseline the start of a loop rather than a report: gaps become actions, actions become shipped work, and the next scan against the same questions shows whether the work moved anything. What is a re-scan covers the re-measurement step, and the locked benchmark methodology treats the measurement principle in full.

How often the loop turns is a cadence decision, not a definitional one — scan frequency varies by plan, and current cadences and prices live on /pricing.

Reading a baseline as a decision tool

A good baseline reads like a to-do list disguised as a report. For each question it answers three operational things at a glance: where you already appear and in what company, where you are absent while rivals are present, and where nobody has answered the buyers well — the open ground. Those three states map to three different plays: defend and refresh what you hold, contest the questions where you are missing, and claim the unanswered ones first.

That mapping is also what protects a team from the two classic misuses of market research: treating the first scan as a verdict on the marketing team, and treating it as a one-off audit to file. A baseline is neither — it is the "before" picture whose entire value arrives later, when shipped work can be checked against the "after." Teams that skip straight to acting without recording the before lose the ability to ever show their work mattered, which is a quiet but expensive loss: the next budget conversation runs on anecdotes instead of deltas.

How baselines go wrong

The failure modes are few and recognizable:

Keywords instead of questions. A baseline of search terms measures a different thing than buyer discovery, and usually the wrong thing.
Drift. Questions quietly added, dropped, or rephrased between scans. Every edit breaks attribution for the affected comparisons; if the market's language genuinely shifts, new questions belong in a separate, parallel set with its own start date, leaving the original lock intact.
Unverifiable entries. Presence claims without links, or built on data no one else can access. A baseline that cannot be audited is just an assertion about the past.
Reading one scan as truth. AI answers vary between runs; a baseline is the reference for a trend, and single-run wobbles are noise until the same locked set confirms them over time.

What an intelligence baseline is not

It is not a keyword target list, and not a snapshot that gets rebuilt fresh each quarter — rebuilding is starting over, not measuring. It is not a simulation or projection of buyer behaviour; it records only what is observably there, with sources. And it is not permanent as a description of the market: what it fixes is the measuring instrument, so that the market's changes — and yours — can be seen honestly against it. Ongoing measurement built on such baselines is the subject of continuous market intelligence.