Competitor traffic report skill

competitor-traffic-report

Three metrics, three different questions. Do not average them or treat one as a proxy for another:

Metric	Source	Answers
Monthly visits	SimilarWeb	total demand, every channel
Organic traffic	Ahrefs	how much of it search is producing
Domain Rating	Ahrefs	backlink authority — the structural position

Visits and organic move on a single viral page. DR does not, which is why a gap in DR says more about where a company actually sits than a good traffic month does.

Run

export AHREFS_API_KEY=...   # required
export BRANDDEV_API_KEY=...  # optional, for logos
S=skills/competitor-traffic-report/scripts

node $S/fetch-traffic.js acme.com rival1.com,rival2.com,rival3.com --out /tmp/data.json
node $S/build-report.js /tmp/data.json "AI presentation tools, ranked" --out /tmp/report.html

Open the HTML, or publish it with the Artifact tool. It is self-contained — logos inlined as data URIs, no external requests — so it survives a strict CSP.

The focus domain is coloured orange and ranked in every chart; competitors take a five-step blue ramp ordered by rank, so colour carries magnitude instead of being decoration.

Choosing the competitor set

Whoever asked usually names three or four. Fill the set out to 20–40 before running: a five-company chart cannot show where anyone sits, and the interesting findings (how much of the market one player holds, how many are shrinking) need the tail. Sources for names, cheapest first:

• the client's own "alternatives to X" and comparison pages;
• an X alternatives / best X tools SERP, read deep enough to reach the top 100;
• Ahrefs competitors_overview for domains ranking on the same keywords.

Drop anything that is not a real peer: a general-purpose giant (Adobe, Google) dominates every chart while competing for none of the same buyers, and its presence compresses everyone else into the bottom decile. Same for a friendly company the requester does not consider a rival — ask.

Subdomains are already counted. SimilarWeb's root-domain figure includes them, so listing both acme.com and app.acme.com double-counts and buries the real number. (Verified: wikipedia.org 3.53B vs en.wikipedia.org 0.88B.) List the root only.

Constraints worth knowing before the first run

• SimilarWeb rate-limits by IP, roughly 15 calls per window, and the block lasts hours. The script paces at 15s and still misses on a long list — rerun to fill gaps, or spread a 40-domain set across two sessions. The X-Extension-Version header is mandatory; a browser User-Agent alone returns 403.
• A returned 0 means "below SimilarWeb's detection floor", not "no traffic". The template drops those from the charts rather than ranking a live site last. For a domain you own, read Search Console instead.
• Ahrefs organic uses mode=subdomains. Domain mode silently excludes the www. host, which reads as an empty site for anyone whose canonical is www..
• The free DR endpoint costs no API units. Organic history does — one row per month per domain.

Publishing the report

The Ahrefs licence (ahrefs.com/legal/domain-rating-license) grants a royalty-free right to display and publish DR data provided the attribution Domain Rating by Ahrefs appears next to it with a live link. The template ships that line; do not remove it. The same licence prohibits reselling the data or using it to build something that competes with Ahrefs.

Reading the result

The report shows position. The story is usually in a gap between two of the three metrics:

• Organic rank far above visits rank — search is carrying more than its share; paid, social or direct is the thin part.
• DR far above organic — the authority is there and the content is not converting it. This is the most actionable gap, because it is a content problem, not a link-building one.
• DR far below peers at similar traffic — the traffic is bought or borrowed, and it stops the moment spend stops.
• A near-identical name at a fraction of the DR (brand.app at DR 87 vs brand.com.ai at 38) is a copycat domain, not a competitor.

State the denominator and the window beside any figure taken from this report. Both sources are modelled estimates, least reliable exactly where it matters most — the small end.