Section 1: the question, confirmed in writing
Bad research starts with a topic ("analyze our competitors"). Good research starts with a decision ("do we cut price, hold, or reposition?"). The first page of a real report restates the question exactly as the client confirmed it, plus the scope boundaries: which competitors, which dimensions, and what was explicitly excluded.
This page is not bureaucracy. It is the contract that prevents the two classic failures: the report that wanders into everything and answers nothing, and the report that answers a question the client never asked. If a research vendor will not confirm the question in writing before starting, that tells you what the deliverable will be.
Example: the question page
Delivered format, condensedCLIENT ASKED: "We sell single-origin coffee online at $22/12oz. Three competitors keep undercutting us at $16-18. Do we cut price, hold, or reposition?" SCOPE: 5 named competitors. Pricing architecture, shipping economics, retention mechanics. OUT OF SCOPE (per client): wholesale channel. DELIVERY: 7 business days. Every claim source-linked [S1]-[S14], competitor pages archived as snapshots so the client can verify what we saw.
Section 2: the competitor table (facts only, no adjectives)
The table is pure observation: prices, terms, shipping, review volume, hiring signals. Two disciplines separate a real table from filler. First, same-day collection: prices checked on different days are not comparable, so everything is captured in one pass and archived. Second, verification over marketing copy: package weights from product photos and listings, not from the brand’s claims about itself.
The table is also where the surprises live, because verified facts beat assumptions. In the example below, the most important finding in the entire report is a bag size.
Example: competitor table
Delivered format, condensedROASTER A: $16.50/12oz, free shipping over $35, subscription -15%. Volume play: 14 origins, none scoring above 86. Ships in 3 days from one warehouse. [S1-S3] ROASTER B: $18/12oz, but bags are 10oz. True price $21.60/12oz. The undercut is packaging math. [S4, S5] ROASTER C: $17, VC-backed, last raise 2023, hiring freeze visible on LinkedIn. Pricing likely unsustainable. [S6-S8] ROASTER D: $24, thriving. Ships within 48h of roast, prints roast date on the box. 4.9 stars across 2,100 reviews; "freshness" appears 312 times. [S9-S11] ROASTER E: $19, no subscription. 60% of traffic from one viral video in 2024, declining since. [S12]
Section 3: what the sources actually say (including where they disagree)
This is the section template-filler research skips entirely: synthesis. Not a restatement of the table, but the pattern the table reveals, written in plain sentences.
In the example: the undercutters are not winning, they are churning. Roaster A’s subscriber reviews complain about staleness, because a single-warehouse model means coffee arrives two to four weeks off roast. Roaster B’s lower price is a smaller bag. Roaster C is burning runway. Meanwhile the competitor charging more than the client is growing, because it sells freshness as a verifiable fact (roast date printed on the box, 48-hour ship) rather than as an adjective.
Crucially, a real report names its contradictions instead of smoothing them over. Industry commentary said specialty coffee turned price-sensitive after 2024; Roaster D’s review volume and repeat-purchase signals contradicted that for the freshness-led niche. The report says so, says which evidence it weights more (the behavioral data), and says why. If a report you commissioned contains no disagreement anywhere, the analyst either did not read enough sources or did not tell you the truth about them.
Section 4: what we would do (the judgment section, with confidence levels)
Research that ends at "insights" outsources the hard part back to you. The judgment section commits to a position, makes each action checkable, and attaches a confidence level to every recommendation so you can see exactly where the evidence is thin.
Checkable means: "print roast dates on the box" is a task someone can verify happened on a Tuesday. "Lean into your freshness story" is strategy fog. And confidence levels are not hedging; they are the opposite. "High confidence on 1 and 2, medium on 3 because it assumes your margin at $18.50 stays above 38 percent, confirm before launch" tells you precisely which assumption to test before betting money.
Example: the judgment section
Delivered format, condensed1. DO NOT cut to $16-18. You would be matching a packaging trick and a dying balance sheet. [Confidence: high] 2. Steal Roaster D’s mechanics, not their price: print roast dates, commit to 72h roast-to-ship, say both on the product page. Your roasting schedule already supports this; it is a messaging fix, not an ops fix. [Confidence: high] 3. Introduce an $18.50 "roaster’s choice" blend SKU as the price-shopper catcher, protecting the $22 flagship. [Confidence: medium; assumes margin at $18.50 >= 38%. Confirm before launch.] 4. Watch Roaster C for 2 quarters. If they fold, their subscribers churn somewhere. Be the somewhere: a "switcher" landing page costs almost nothing now. [Confidence: medium]
Section 5: the source list and method note
The last section is the audit trail: every source as a working link, archived snapshots of competitor pages (so a price change next week cannot rewrite history), and a method note that states plainly what the report cannot tell you. In the example: competitors’ actual unit economics were inferred, and every inference is flagged as one.
The method note is the fastest way to evaluate any research vendor. Ask one question before commissioning: "what will the report say it could not determine?" A vendor with a ready answer does real work. A vendor who promises completeness is selling slides.
What this structure costs in the market: custom research firms start around $5,000 and routinely exceed $65,000, excellent and slow and priced for enterprises. Freelance analysts run $40 to $300 an hour with quality that is a lottery you manage. Our competitive scan at $290 and full report at $790 exist because AI made the gathering phase roughly ten times faster while judgment stayed human-speed: a person still reads every line, checks every source, and writes section 4 personally. The complete sample scan below is the real document, condensed only for length.
Common questions
How many competitors should a competitive analysis cover?
Three to six, chosen deliberately: your direct price rivals, plus the player winning at a higher price point (where the lessons usually are), plus any structurally fragile competitor whose customers may soon be available. Past six, depth per competitor collapses and the table becomes a directory.
How often should competitive analysis be refreshed?
For a specific decision, do it once, decide, and act. As an ongoing practice, a light quarterly refresh of the table (prices, offers, hiring signals) is enough for most small businesses; the full synthesis and judgment sections only need redoing when something structural moves, like a competitor folding or a new entrant with funding.
Can I just do this with an AI chatbot for free?
A chatbot accelerates the gathering phase and is genuinely useful for it. What it will not do reliably: verify prices on the same day from live pages, archive evidence you can audit, notice that a competitor’s bag is 10oz instead of 12, weigh contradictory sources, or stake a defensible position with confidence levels. If you use AI directly, you become the analyst doing those five things, which is a fine choice when the decision is small. When it is not, the verification layer is what you are paying anyone for.