Competitor Teardown Template: What to Score and How to Act

Most competitor research fails in the same way. Somebody spends a day on three rival sites, writes an impressive document, and nothing changes, because the document records observations rather than reaching a decision. Omniconvert has scored stores against real competitors in the same category and country across the CROBenchmark dataset of 7,000+ ecommerce sites, measured against 248+ audit criteria over 13 years in eCommerce and 70,000+ experiments, and the teardowns that change behaviour share one property: the structure is fixed in advance and the output is a commitment. This template sets out the six dimensions, the evidence each one needs, how to convert six scores into one decision, and the specific mistakes that turn the exercise into theatre.

What a competitor teardown is, and why structure is the whole trick

It is worth being precise about what the structure buys, because the instinct is to treat it as bureaucratic overhead.

Without fixed dimensions, two people tearing down the same competitor produce different documents, and the same person tearing down two competitors produces findings that cannot be set beside each other. Worse, a teardown repeated next quarter records whatever is salient that day, so you cannot tell whether anything changed.

With fixed dimensions you get three things at once. Competitors become comparable, because every one is scored on the same axes. Quarters become comparable, because the axes did not move. And disagreements become tractable, because an argument about whether a rival is ahead on creative becomes an argument about a specific score with specific evidence behind it.

The cost is that you have to choose the dimensions once and then stop changing them, which is harder than it sounds when something new and interesting appears. Add a dimension and you have broken the comparison with every previous run. The wider toolkit for gathering the evidence is in the DTC competitor research stack.

The six dimensions, and the evidence each one needs

Work through these in order. Each one takes twenty to forty minutes per competitor once you have done it twice.

1. Creative and ads. Which concepts are running, how long each has been live, and which format dominates. Longevity is the signal worth recording: an ad that has run for months is one somebody keeps paying for, which is the closest thing to performance data you can see from outside. Meta's Ad Library is the primary source and it is free.
2. Reviews and user content. Review coverage on their best sellers, how recent the newest reviews are, and whether customer photographs appear on product pages. Coverage and recency matter more than the star average, because a product with nothing recent to read converts worse regardless of its rating.
3. AI visibility. Ask an answer engine the five questions your buyers actually ask and record who gets named and what is said about them. Do it with the same five questions every quarter. This is the dimension where positions move fastest and where most stores have no idea where they stand.
4. Agentic readiness. Fetch one of their product pages without rendering and record which buying facts survive: price with currency, per-variant stock, delivery cost and timing, return terms. This is invisible from the rendered page and it is increasingly what decides inclusion in a machine-assembled shortlist.
5. Competitor synthesis. What is their actual position, stated in one sentence, and what are they deliberately not doing. The omissions are more informative than the features, because a gap a competent rival has left open is usually a decision rather than an oversight.
6. Conversion experience. Walk the funnel to the payment step on a phone, and record every point where a fact you needed was missing. Do not score how it looks. Score what you could not find out.

Two of these are differentiators in the sense that almost nobody includes them, and they are the third and fourth. AI visibility and agentic readiness are both measurable from outside, both moving quickly, and both absent from every teardown template written before this year.

How to score without inventing precision

The scale matters more than it should, because a scale finer than your evidence turns a teardown into a negotiation.

You are looking at a competitor from outside, with no access to their numbers. What you can honestly say is whether they appear clearly better, roughly comparable or clearly worse on a dimension, with a note on what you saw. That is three points, and pretending to ten makes the exercise feel rigorous while making it less reliable.

Record the evidence next to the score, always, in one line. "Ahead: four concepts running over ninety days, ours longest is three weeks." Now a colleague who disagrees has to argue with the observation rather than with your judgement, which is a much more productive conversation and occasionally reveals that you were looking at the wrong thing.

The table sets out the six dimensions with what you record and what the common error is on each.

The last row is the most common failure in the whole exercise. A teardown that concludes a rival has a nicer checkout has recorded an aesthetic preference. A teardown that records that their delivery estimate appears before the cart and yours does not has found something you can act on this week.

Turning six scores into one decision

This is the step that separates research from a decision, and it is the step most templates omit entirely.

You now have six scores with evidence. Two filters reduce them to a choice. First, gap width: where are they clearly ahead rather than roughly level. Second, closability: could you meaningfully move this inside a quarter with the people you have. A wide gap you cannot close this quarter is a strategy conversation, not a teardown output.

Usually one or two dimensions survive both filters, and in practice it is often agentic readiness or AI visibility, because both are wide gaps for most stores and both are cheap to close relative to competing on creative volume.

Then write the sentence. "They state delivery cost and timing on the product page and we do not, so this quarter we are putting delivery cost and estimated arrival above the add button on our top fifty products." That sentence is the output. The scores are working notes.

Where the dimension you pick concerns data completeness, the groundwork is in the Omniconvert Explore discipline of testing the change rather than assuming it, and where you want the resulting queue of fixes ordered by profit rather than by which competitor annoyed you most, Nexus by Omniconvert is the AI for eCommerce growth engine that ranks the next experiment by True Profit and generates the variants you approve before they go live. Once you know where you stand, Ecommerce Benchmark scores your store across the same six dimensions against real competitors in your category and country, with a free leaderboard entry and a paid audit report for the detail.

The four mistakes that make a teardown useless

Taste over evidence is the first. A teardown full of judgements about whether a brand feels premium cannot be checked, cannot be compared and cannot be acted on. Every score needs an observation beside it, and if you cannot write the observation you do not have a score.

Changing dimensions between runs is the second and the most tempting, because something new always looks important. The discipline is to keep the six fixed for at least a year and keep a separate list of candidate dimensions for the next revision, which is also when you accept that you are resetting the comparison.

Tearing down only your closest competitor is the third. The closest rival is the one you already understand, so the yield is low. The volume winner teaches you what scale changes, and the structurally different store, often from an adjacent category or another country, is where the ideas you had not considered come from. Baymard Institute's usability research is useful alongside this for the same reason: it describes what shoppers need in general, which is a check on conclusions drawn from a sample of three.

Ending in observation is the fourth and the fatal one. Six paragraphs of findings with no named commitment is a document that will be praised in the meeting where it is presented and never opened again. One sentence beats six paragraphs because one sentence can be put on a roadmap.

FAQ: competitor teardowns

What is a competitor teardown?

A competitor teardown is a structured assessment of a rival store across a fixed set of dimensions, scored the same way every time so that results are comparable between competitors and across quarters. The fixed structure is what separates it from browsing a competitor site, which produces impressions rather than a record you can act on or revisit.

How many competitors should you tear down?

Three is usually right: the one closest to you in positioning, the one winning on volume, and one outside your immediate set that is doing something structurally different. More than four and the exercise stops being finished. The third choice matters most, because the useful ideas rarely come from the competitor you already watch.

Which teardown dimension do most stores skip?

Agentic readiness, meaning whether a competitor publishes price, stock, delivery and return terms as data that something other than a browser can read. It is skipped because it is invisible from the rendered page, and it is increasingly the dimension that decides which products appear in a machine-assembled shortlist.

How do you keep a teardown from becoming a document nobody uses?

End it with one decision rather than a summary. The output is a single sentence naming the dimension with the widest closable gap and what you will do about it this quarter. A teardown that produces six paragraphs of observation and no commitment is research theatre, and the usual cause is scoring everything without ranking anything.

How often should a competitor teardown be repeated?

Quarterly is right for most categories. Creative changes faster than that and everything else changes slower, so a monthly teardown mostly re-records the same findings at four times the cost. Keep the dimensions and the scoring identical between runs, because the comparison over time is where most of the value turns out to be.