Competitor analysis

What this is for

To settle a decision you would otherwise make on instinct: what to claim, what to price against, what to say when a prospect names somebody else in the room. Analysis that changes no decision is a deck, and a deck is stale the week after it is presented.

The one rule: every fact carries a source and a date. That single discipline is the whole difference between intelligence and folklore, and it is also what makes a refresh possible. You cannot re-verify a claim whose origin nobody wrote down, so undated intel does not decay visibly. It just quietly becomes wrong while everyone keeps quoting it.

How to think about it

• Buyers define the set, not you. The right competitors are the ones that turn up in lost-deal notes, in sales calls, and in the answers AI engines give next to your name. The company you find most interesting is rarely one of them.
• Their customers outrank their homepage. A rival's marketing tells you their strategy. Their reviews tell you the truth about their execution, and the gap between the two is where every usable finding lives.
• Attack where their customers are unhappy, never where they are happy. Nobody believes you over their own experience, and trying costs you the credibility you needed for the point you were actually right about.
• The alternative that is not a company deserves a battlecard too. For most products the biggest rival is non-consumption: the spreadsheet, the intern, the way it is done now. It wins more deals than any named competitor and gets studied the least.
• Say when they are the right choice. Precisely, in writing. It is what makes the rest of the card believable, and it only ever costs you deals you were going to lose anyway.

Failure modes worth naming: stamp collecting, where a board hoards observations until nobody reads it; paraphrasing a rival's claim so it sounds a little worse than they said it; repeating a weakness they fixed two quarters ago; and producing a battlecard no salesperson has ever opened.

Numbers to hold it against

Competitive work has fewer clean published benchmarks than a channel does, so most of these are working rules of thumb rather than medians. They still beat doing it by feel.

• Three to five competitors in the deep set. Past five, the refresh cost outruns the value and the board turns into a museum. Everyone else gets a name, one line, and a link.
• Thirty to fifty recent reviews per competitor is where themes stabilize. The loudest two or three usually appear inside the first twenty; past fifty you are mostly re-reading them. A theme that has not repeated by review thirty is an anecdote wearing a pattern's clothes.
• Recent means the last six to twelve months. Older complaints expire. Repeating a fixed weakness in front of a prospect who has just been shown the fix is the fastest way to lose a room.
• Read the middle ratings first. Review distributions are strongly J-shaped: most reviews sit at five stars or one star, and almost all of the diagnostic detail is in the two, three and four star reviews nobody reads.
• Ten to fifteen recent win/loss conversations gets you the pattern, and it has to include wins. Interviewing only losses produces a board that explains why you never win, which is a different thing from the truth.
• Quarterly refresh, or on a trigger: their pricing changes, a release lands, a lost deal surprises you. Pricing pages move faster than a quarter, so re-verify a price before anyone quotes it. Treat a battlecard older than a quarter as expired.
• A battlecard nobody opens is worth zero, whatever its quality. The only honest measure of this board is how often its artifacts get used in a live deal or a live campaign.

What moves all of these: how crowded the category is, how public the competitors are (a private, sales-led rival leaves far less surface than a self-serve one), and whether your buyers write reviews at all. In categories where they do not, review mining is replaced by sales-call notes and the evidence bar gets harder, not softer.

What has to be true before a card is trusted

Competitor profiles. Claims quoted with a URL and a date, never paraphrased. Entry price read off the pricing page this month rather than remembered. A last-verified date on the card, and an explicit "unverified since" line for any fact that could not be checked, because a confident-looking card with one stale row is worse than a card with a gap in it.

Review mining. Themes counted, not impressions. A verbatim quote attached to every theme. The recency window written on the card, because a reader six months from now cannot infer it.

Findings. One falsifiable sentence, the supporting cards linked, a so-what, and the artifact it feeds. A finding that feeds nothing is trivia: archive it rather than letting it decorate the board into noise.

Battlecards. Honest in both directions. An explicit when-we-lose section, landmines grounded in complaints their real customers actually wrote, and a genuine when-they-are-the-right-choice clause. A last-verified date on the front, because a wrong battlecard loses deals quietly and nobody reports it.

All of it. Public information only: pricing pages, docs, reviews, changelogs, job posts, filings, their own marketing. That is more than enough. No pretexting, no fake trials, no demo booked under an invented identity. One discovered incident costs more trust than every insight it ever produced.

The seed cards are the first instance, not the ceiling

Seven cards show the shape. A working board is denser than that, and the density lives in Gather.

A healthy Scope holds the set card plus, usually, a light-watch card naming everyone outside the deep set so the decision to exclude them is visible. Gather holds two cards per deep-set competitor, a profile and a their-customers card, so four competitors is eight cards, plus a profile for the non-consumption alternative that most boards skip and most categories lose to. Analyze holds five to twelve findings, including the cross-competitor patterns no single profile can show: the pricing rung everyone left empty, the objection nobody answers, the segment they all ignore. Arm holds one battlecard per deep-set member, one for the alternative, and the price map.

Two competitor cards on a board whose sales team names six rivals is not a scope decision. It is an unfinished phase.

The layout is one arrangement of many

The columns run by stage of the loop: scope, gather, analyze, arm. That is the right shape the first time through. Once several refreshes exist, one column per competitor reads better for the people using the board daily, and arranging by refresh cycle reads better for anyone trying to see how the landscape moved over a year. Rearrange it freely; the dated-evidence rule and the review-mining discipline survive any layout.

When this is the wrong playbook

If you have no customers yet there is nothing real to compare, and the honest exercise is a positioning sprint built from alternatives rather than from rivals. If the thing beating you is non-consumption, most of this board is busywork and one battlecard is the entire job. If the question you actually have is why your own customers bought, go and ask them: a rival's reviews are a poor substitute for your own. And if the analysis is being run to settle an internal argument rather than to change a decision, it will find whichever answer the loudest person already wanted.