Most competitor analyses are thorough, well-presented, and change nothing. They get built over three weeks, shown once, complimented, and filed. Six months later somebody asks a strategic question and nobody opens the document, because the document doesn't answer it.
The reason they don't land
The failure isn't rigour. Plenty of these documents are meticulous. It's that they were produced without a decision waiting for them.
"Let's do a competitor analysis" is an activity, not a question. It produces a comprehensive description of five companies across ten dimensions — which is genuinely informative and has nowhere to go, because information only becomes useful at the moment it changes what someone does.
The inversion this article runs on Don't start with the competitor. Start with the decision you're stuck on. Then work out which single fact about which competitor would resolve it.
That's a much smaller piece of work and a far more useful one. "Should we raise prices?" is a question with a shape — it tells you exactly what to look at, when to stop looking, and what would count as an answer. "Let's analyse our competitors" tells you none of those things, which is why the resulting document is comprehensive and directionless.
So before anything else, write down the decision. If you can't name one, you don't need a competitor analysis — you need to work out what you're actually trying to decide, which is usually a symptom that your goals aren't specific enough to generate real questions.
You're probably analysing the wrong companies
The second failure, and it invalidates everything downstream.
Most competitor lists are assembled internally from memory — who the sales team mentions, who turns up at the same events, who the founder is annoyed by. That list is reliably wrong in two directions: it includes companies buyers never consider, and it omits the ones they actually choose instead.
Your real competitive set is defined by your buyers, not by your industry. Which means you can find it out rather than guessing.
Ask the people who chose. Win-loss conversations are the highest-quality source available, and most organisations either don't run them or run them only on losses. Ask both. "Who else did you look at, and what nearly made you pick them?"
Read your sales calls. Whoever gets named unprompted is in your competitive set, whatever your positioning deck says. If you record calls, this is a search away — and it pairs naturally with using AI to analyse customer evidence you already have.
Check the comparison queries. Search Console will show you which "X vs Y" queries you appear for. Some of those Ys will surprise you.
Ask the machines. A 2026 addition and a genuinely useful one. Ask several AI assistants for alternatives to your product, for your kind of customer. Whatever comes back is increasingly the shortlist your buyers are seeing — the mechanism behind generative engine optimisation, applied here as research rather than as a visibility tactic.
Read the threads. Community discussions comparing options in your category name your real alternatives in public, with reasoning attached — one of several reasons forum threads are worth mining even when you never post in them.
SurveyMonkey
Online survey platform for feedback, market research and forms with AI assistance and analytics.
Best for: Surveys, customer feedback & market research
Why SWOT is where analysis goes to die
Nearly every guide on this topic recommends a SWOT. It's worth being direct about the problem.
SWOT is a sorting mechanism, not a research method. It gives you four boxes to put findings in — it doesn't generate findings. So a SWOT built from assumptions produces neatly organised assumptions, which feel more credible than they were because they're now in a grid.
It also invites exactly the vagueness that makes analysis useless. "Strong brand" is a category, not a finding. "Better funded" is a fact with no implication attached. Four boxes of adjectives about a competitor tell you nothing about what to do on Monday.
If you already have specific evidence, you don't need the boxes. If you don't, the boxes won't create it. Spend the time on the underlying research instead.
What's actually worth looking at
The test for any data point: if this turned out one way rather than another, would we do something different? If no, it's interesting rather than useful, and it belongs in the appendix or nowhere.
| Source | What it reveals | Effort |
|---|---|---|
| Pricing page, tracked over time | Strategy shifts, packaging bets, what they think they're worth | Low |
| Their reviews, especially the middling ones | Real weaknesses in customers' words — your best messaging material | Low |
| Job listings | What they're building next, six months before it ships | Low |
| Homepage copy changes | Repositioning, usually before it's announced | Low |
| Win-loss interviews | The actual decision criteria, including uncomfortable ones | High — and worth it |
| Their content and search footprint | Which topics they're investing in, which they've abandoned | Medium |
| Follower counts, posting frequency | Almost nothing about commercial performance | Low — and usually wasted |
Two rows deserve emphasis. Job listings are the most underused free intelligence available — a competitor hiring three people for a capability they don't currently offer is telling you their roadmap without meaning to. And middling reviews beat the extremes: one-star reviews are often about a bad support experience, five-stars are often incentivised, but three-star reviews from people who broadly like the product are where the honest limitations live.
The last row is the one to actually cut. Social follower counts correlate poorly with revenue, are easily inflated, and consume analysis time because they're trivially available. Availability is not relevance.
The trap: analysis produces convergence
Something worth naming because it undermines the whole exercise if left unchecked.
The instinctive response to competitor research is to close gaps. They have a feature you don't; add it. They rank for a topic you don't; write it. They post daily; post daily. Each decision is locally sensible.
Do it consistently and you converge on the category average. Worse, so does everyone else running the same exercise — which produces a market where every product has the same features, every website makes the same claims, and no buyer can tell anyone apart. That's the strategic version of the sameness problem, and competitor benchmarking is one of its main causes.
The useful question isn't "what are they doing that we're not." It's "what are they all doing that we could deliberately not do?" Uniform behaviour across a category is usually inherited assumption rather than tested truth, and the gap it leaves is the most defensible position available — because closing it would require your competitors to contradict themselves.
So run the analysis looking for the shape of the consensus, then decide whether you want to join it or exploit it. Both are legitimate; only one of them differentiates.
Turning findings into decisions
The step that separates useful analysis from a nice document, and it's mechanical.
Every finding gets written as a sentence with a consequence attached. Not "Competitor X has cheaper entry pricing" but "Competitor X's entry price is 30% below ours, which is why we lose small deals — so either we add a lower tier or we stop pursuing that segment."
Findings without consequences are trivia. Try this format:
- What we found: the specific, evidenced observation
- Why it matters: the mechanism connecting it to your business
- So we should: the action, or explicitly "nothing, but watch it"
- How we'd know we were wrong: the signal that would overturn the conclusion
That last line is the one people skip and shouldn't. Competitor analysis is unusually prone to confirmation bias — you generally start with a hypothesis about a rival and then find evidence for it. Writing down what would disprove your conclusion is the cheapest available guard against building a strategy on a story you liked.
Aim for five findings in this format rather than fifty observations in a grid. Five things somebody can act on beat fifty things nobody reads.
Cadence: monitoring beats reports
The annual competitor deep-dive is a poor format for a market that moves continuously. By the time it's compiled, presented and circulated, parts of it are stale — and it arrives detached from any decision, which returns us to the first problem.
A better rhythm has two speeds:
Continuous, lightweight. Under an hour a month. Check pricing pages, homepage messaging, job listings and new review themes for your top three or four real competitors. Log changes with dates. Most months nothing happens; the value is in catching the month something does, while it's early enough to respond.
Deep, decision-triggered. Full analysis only when a real decision needs it — a pricing change, a repositioning, a launch, a budget reallocation. Scoped to the decision, and finished when the decision is made.
That combination gives you early warning without the ritual, and depth without the drift. It's also considerably less work in aggregate than the annual report most teams produce and nobody uses.
Where competitor analysis genuinely can't help
Four situations worth recognising, because in each of them the analysis will produce output that misleads:
- You don't know why you win. Competitor research can't tell you your own value. Talk to your happy customers first — otherwise you're comparing an unknown quantity against known ones.
- The category is genuinely new. There's nobody to analyse, and the closest analogues will point you toward a market that doesn't exist yet.
- The real problem is internal. Slow delivery, weak onboarding, an unclear offer — none of these is caused by a competitor and none is fixed by studying one. That's a funnel audit, not a competitive one.
- You're competing on trust, not features. In categories where buyers weight peer evidence over vendor claims — increasingly most of them, as B2B buying research keeps showing — a feature matrix explains very little about why one vendor wins.
A workable version, in a week
- Write the decision. One sentence. If you can't, stop here — that's the actual finding.
- Establish the real competitive set. Win-loss notes, sales call mentions, comparison queries, AI assistant alternatives, community threads. Expect surprises, and include the do-nothing option.
- Gather only what bears on the decision. Three or four sources per competitor, not everything available.
- Look for the consensus. What are they all doing? Decide whether to join it or break it.
- Write five findings with consequences and disproof conditions. Not a grid.
- Set up the monthly check and put it in someone's calendar. This is what makes it a capability rather than a project.
If the output points at repositioning rather than tactics — you've discovered the whole category says the same thing and you need to say something else — that's a different and larger piece of work than a competitive audit, and the point at which a brand and positioning partner is more useful than another spreadsheet.
The short version
Start from the decision, not the competitor — an analysis with nowhere to land doesn't land. Find your real competitive set from buyers rather than assumptions, and remember that doing nothing is often the winning alternative. Skip the SWOT, which organises findings without producing them. Prioritise pricing pages, job listings and middling reviews over follower counts. Watch for the convergence trap: the useful question is what everyone in your category is doing that you could deliberately not do. Write five findings with consequences attached, and replace the annual report with an hour a month.
Found out your whole category says the same thing?
We build positioning and brand strategy that gives buyers a reason to choose you specifically.
Explore Strategy & Content →Frequently asked questions
How do you identify who your real competitors are?
Ask your buyers rather than guessing. The most reliable sources are win-loss conversations, the alternatives named in sales calls, and the comparison queries people search before finding you. A quick modern check is to ask several AI assistants to list alternatives to your product for your type of customer, since that increasingly reflects the shortlist buyers actually see. The set that comes back is usually broader than the one on your internal list and frequently includes doing nothing, building it in-house, or a spreadsheet.
Is SWOT analysis useful for competitor research?
Rarely, in the form most teams use it. SWOT is a way to sort observations into four boxes, not a method for producing them, so a SWOT built from assumptions simply organises the assumptions. It also invites vague entries — describing a competitor as having a strong brand is a category rather than a finding. If you already have specific evidence, the SWOT adds little; if you do not, it will not generate any. Time is better spent on the underlying research.
How often should you run a competitor analysis?
A full analysis belongs to a decision rather than a calendar — before a pricing change, a positioning shift, a launch, or a budget reallocation. Between those, lightweight continuous monitoring works better than an annual deep dive: a short monthly check of pricing pages, messaging changes, hiring patterns and new review themes takes under an hour and catches movement while it is still actionable. Annual reports tend to describe a market that has already moved on by the time anyone reads them.
What should you actually look at when analysing competitors?
Prioritise evidence that reveals decisions rather than outputs. Pricing pages and how they change, positioning language on the homepage, what their job listings suggest they are building, complaint themes in their reviews, and what customers say when comparing them to you. Skip vanity metrics such as follower counts and posting frequency, which correlate poorly with commercial performance and are easy to mistake for momentum. The test for any data point is whether a plausible finding would change something you do.
Why do competitor analyses often fail to change anything?
Because they usually answer questions nobody asked. A document describing five competitors across ten dimensions is comprehensive and directionless — there is no decision waiting for it, so nothing happens when it arrives. Starting from a specific decision, such as whether to raise prices or which objection to address in messaging, produces a much smaller piece of work that has somewhere to land. The other common failure is that benchmarking against competitors pulls a company toward the category average rather than away from it.
