How to Write a Competitive Analysis: Steps, Examples and Checklist

Choose the competitor set from your customer's point of view — include the alternatives they actually weigh, which may be spreadsheets and doing nothing, not just rival products. Agree your comparison criteria before gathering evidence, so the analysis cannot quietly bend towards a flattering result. Collect dated, sourced evidence, compare like for like, and mark what you could not verify. Then do the work most analyses skip: identify where you can genuinely differentiate, where you are exposed, and what decisions follow.

By Gensudo Team · Updated 23 July 2026

When you need this document

Write a competitive analysis during Discover, when you are establishing market context before committing to a direction — and refresh it whenever positioning, pricing or roadmap decisions depend on knowing where rivals stand. It is also the document that stops competitor anxiety being settled by whoever speaks loudest. If the question is broader — how big the market is, where it is heading — start with a market research report and let this analysis go deep on the competitive slice.

Gather these first

The decision at stake: positioning, pricing, a build/kill call, or an investment case
The customer's real consideration set — from win/loss notes, sales calls and churn interviews
Public evidence per competitor: sites, pricing pages, docs, release notes, reviews, filings
Your own product's honest capability picture, from the people who know its weaknesses
The comparison criteria that matter to your target buyer, not to your roadmap
Dates and sources for everything, because competitive facts decay quickly

Step by step

Build the competitor set from the customer's shortlist, not yours
List the alternatives your target customers actually consider — pulled from win/loss data, sales conversations and churn interviews, not from your own mental map of the category. This usually adds unglamorous entries: the incumbent spreadsheet, the internal tool, the decision to do nothing. It may also demote a rival you obsess over but customers rarely mention. Analysing the wrong set makes every subsequent page precise about the wrong question.
What good looks like: The set matches what buyers name in real deals, including at least one non-product alternative like the status quo.
Fix the comparison criteria before you gather a single fact
Decide what you will compare on — the capabilities, price points and qualities your target buyer weighs — and write them down before research begins. Criteria chosen after the evidence is in drift, unconsciously, towards dimensions where you win. Draw them from what customers evaluate in purchase decisions, and include criteria you expect to lose on; those are the ones the analysis most needs to be honest about.
What good looks like: The criteria list was fixed before research started and includes dimensions where you expect competitors to beat you.
Gather evidence with sources and dates, and mark the gaps
Work through public material systematically: pricing pages, documentation, release notes, review sites, job postings, filings. Attach a source and date to every claim, and distinguish three grades of knowledge — verified, reported, inferred. Where you could not find out, record 'unknown' rather than guessing; a visible gap is useful, a confident guess is a landmine someone will build a strategy on.
What good looks like: Every cell of the comparison can answer 'says who, and when?' — and the unknowns are marked as unknowns.
Compare like for like, at the same altitude
Assemble the comparison so each competitor is assessed on the same criteria at the same depth. The classic distortion is granular knowledge of your own product against brochure-level knowledge of rivals — which manufactures false gaps in both directions. Compare their shipped reality with your shipped reality, not their marketing with your roadmap, and note where a competitor's weakness is actually a deliberate choice serving a different segment.
What good looks like: No competitor is judged on marketing copy while you are judged on shipped product — or the reverse.
Locate the differentiation and the exposure
Now interpret: where can you genuinely win, and where are you vulnerable? Real differentiation must pass three tests — customers care about it, competitors cannot quickly copy it, and you can evidence it. Be equally rigorous about exposure: the gaps a competitor could exploit, the segment drifting towards a rival, the pricing pressure building. An analysis that finds only strengths has been written for comfort.
What good looks like: Claimed differentiators pass the care/copy/evidence tests, and the exposure list contains something genuinely uncomfortable.
Anticipate movement, not just position
A snapshot ages the day it is written. From release cadence, hiring patterns, funding and stated strategy, make explicit — and clearly labelled — inferences about where each key competitor is heading. Separate observation from speculation ruthlessly, but do speculate: 'they have hired four ML engineers and their last three releases point at automation' is exactly the forward view that makes the analysis strategic rather than archival.
What good looks like: Each major competitor has a labelled trajectory inference, with the evidence behind it, distinct from verified fact.
End with decisions, owners and a refresh trigger
Close by converting the analysis into consequences: the positioning, pricing or roadmap decisions it should drive, who owns each, and what you are explicitly choosing not to react to. Then set a refresh trigger — a time interval or events, such as a rival's major release or funding round, that would make the analysis stale. A competitive analysis without decisions attached is competitor tourism.
What good looks like: The reader leaves knowing what you will do differently, who owns it, and when the analysis stops being trustworthy.

Common mistakes

Comparing your roadmap against competitors' shipped products — or their marketing claims against your engineering reality — producing gaps that do not exist. — Hold the altitude constant: shipped versus shipped, claimed versus claimed, and label which is which. Mixed altitudes are the single most common way competitive analyses mislead the teams that trust them.
Criteria chosen after the research, quietly weighted towards dimensions where you win — so the analysis flatters and the market later corrects it. — Fix criteria before gathering evidence, source them from what buyers actually evaluate, and deliberately include your weak dimensions. The uncomfortable rows are the ones leadership most needs.
Undated, unsourced claims — a pricing figure from an old screenshot, a capability rumour from a sales call — presented alongside verified fact as equals. — Grade every claim as verified, reported or inferred, with source and date. Competitive facts decay in months; ungraded evidence makes the whole document only as trustworthy as its weakest guess.
The analysis describes competitors thoroughly and concludes nothing — no differentiation call, no exposure named, no decision proposed. — Force the last mile: what should we do differently, who owns it, what are we choosing to ignore? Description is research; the decision section is the analysis.
The customer's real alternatives — spreadsheets, internal tools, doing nothing — are missing, because they are not 'proper' competitors. — Include the status quo as a competitor row with the same criteria. In many markets it wins more deals than any named rival, and an analysis that ignores it misdiagnoses why you lose.

Before you call it done

The competitor set reflects buyers' actual consideration sets, including the status quo
Comparison criteria were fixed before research and include your weak dimensions
Every claim is graded verified, reported or inferred, with source and date
Comparisons hold altitude constant — shipped against shipped, claimed against claimed
Differentiators pass the care/copy/evidence tests; exposures are named honestly
Key competitors have labelled trajectory inferences, not just current positions
The document ends with decisions, owners, non-reactions and a refresh trigger

Frequently asked questions

How many competitors should a competitive analysis cover?

Deeply, usually three to five — the alternatives that appear most often in real buying decisions — with a lighter watchlist for the rest. Depth on the set that decides your deals beats shallow coverage of twenty names, because the analysis has to support specific positioning and roadmap choices. Let win/loss frequency, not category directories, decide who earns the deep treatment.

How often should a competitive analysis be refreshed?

Set both a clock and a tripwire. A periodic review — quarterly in fast-moving categories, twice a year in slower ones — catches gradual drift, while named trigger events (a rival's major release, funding round, acquisition or pricing change) catch the sudden moves. Write the triggers into the document itself, so staleness is detected by design rather than by an embarrassing meeting.

How is a competitive analysis different from a market research report?

The market research report surveys the whole territory — size, growth, segments, trends, external forces — with competition as one section among several. The competitive analysis is the deep dive into that section: named rivals, compared on fixed criteria, ending in differentiation and exposure decisions. Sequence them that way when you can; knowing the territory first tells you which competitors are worth the depth.

Can competitive research be automated?

Parts of it. Gensudo's shared research pool includes a competitor-intelligence category: findings arrive with source URLs, anything without a usable source is dropped, and you can verify sources and flag conflicts between them before documents cite the evidence. What stays human is the judgement — choosing the criteria, weighing the trade-offs, and deciding what you will do about it. The analysis is the thinking, not the collection.

Start from the structured template

See the full Competitive Analysis for Product Managers template and structure, or draft it in Gensudo with cited evidence.

Get started