Most competitor analysis produces a comparison table: their price, rating, review count and estimated sales, laid beside yours.
That table almost never produces an action. It answers "how are they doing" when the question you need answered is "what should I do". Their price is $3 lower — and therefore? Cut yours?
Useful competitor analysis is not looking for comparison. It looks for the difference: what they have that you do not. A difference maps to an action.
Four differences that produce actions
| Angle | What to find | The action it maps to |
|---|---|---|
| Keyword difference | Terms they show up for and you do not | Add terms or advertise |
| Dual-placement waste | Terms where they hold organic and paid at once | Read their ad strategy |
| Unmet demand in negatives | Complaints they keep getting and have not fixed | Product differentiation |
| Variation structure | Variations they carry and you do not | Add configurations |
The first two come from traffic data, the third from reviews, the fourth from product detail.
One: the keyword difference
Three steps, on fixed semantics:
- Pull traffic terms for three to five comparable competitors (same marketplace, same month)
- Pull traffic terms for your own ASIN, under the same conditions
- Take the difference: terms at least two competitors have and you do not
"The same conditions" in step two is the load-bearing part. Query only competitors and you get a table proving competitors are strong. Query the two sides on different months or marketplaces and the difference is noise.
The difference is a candidate list, not a to-do list — they may carry variations, sizes or certifications you do not, making some terms relevant to them and not to you.
How to read the lookup's fields and where its pagination stops: ASIN reverse keyword lookup.
Two: a number you can compute directly
The traffic keyword stat endpoint returns three counts per ASIN: keywords (all traffic terms), ranks (organic) and ads (advertising).
The documented examples are 2685, 1848 and 1414. Note that 1848 + 1414 = 3262, 577 more than the total.
The excess is not an error. It is the set of terms holding an organic and a paid position at once:
dual-placement terms ≈ ranks + ads − keywordsRun that on a competitor and you can read their advertising style: a high dual-placement share means they keep advertising on terms they already rank for organically (defensive); a low share means advertising is mostly opening new terms (offensive).
That is far more concrete than "do they spend a lot on ads", and it takes three numbers. Applying it to yourself: Monitoring keyword rank.
Three: unmet demand inside negative reviews
A competitor's negative reviews are worth more than your own, because they are demand buyers have already stated out loud that the competitor has not met.
The method: pull their 1 and 2 star reviews, sort by recurrence, and find what keeps coming up. Where your product happens not to have that problem, you have a differentiator that goes straight into your bullets.
Two boundaries that matter:
Separate the review sources first. Reviews carry verified, vine, free and experience; someone who received the item free is missing the "was it worth the price" judgment. Use actually-purchased reviews for the baseline — The four sources of a review.
Separate "fix the product" from "fix the description." A review saying "doesn't look like the photo" is a photo problem, not a product problem. Treating the second as the first is the most expensive misreading in competitor work. Full method: How to analyse reviews.
Four: variation structure
The product detail endpoint returns parent and variationList / variations. A competitor carrying eight colors against your three is a concrete, actionable gap.
There is a trap here too: parent and child data are not the same thing. A parent's reviews are pooled across variations and its sales estimate covers the whole family. Comparing their parent figures against your child figures can be off several times over, and nothing errors out. Full explanation in What is an ASIN.
Getting it into a sheet
Keep the output of the four angles separate: what you observed, what it might mean, what you intend to do — three columns. Welding them together is the well-worn route to shipping a guess as a finding.
The sheet's full structure and field mapping is in The Amazon competitor analysis sheet. To hand the whole flow to a model, see Competitor analysis with an AI agent.
Three things competitor analysis cannot do
It cannot give you their backend search term report. That is authorized seller account data. The lookup returns terms that actually appeared in public search results — different source, no correspondence.
It cannot give you their real sales or ad spend. Sales are derived from BSR and ad spend is entirely invisible. Anything claiming exact figures is, underneath, an estimate.
It cannot tell you why they did something. The data shows they added spend on a term; it does not show whether that is a strategy shift or clearing stock.
Questions
How many competitors should I pick? Three to five comparable ones. Below three the difference is full of one-offs; above five it dilutes into "the whole category's terms" and loses its edge.
How do I know who my real competitors are? Not by who sells well, but by who competes for the same search terms. Overlap in the reverse lookup beats impressions.
How often? Monthly. Their terms and reviews change slowly, weekly runs mostly show noise, and your own changes need time to take effect.
Can their negative reviews serve as a requirements document? Not directly. They are demand signals that still have to pass your own cost and supply chain judgment — sometimes a competitor has not solved something precisely because it is not worth solving.