Competitor research is not a report you read once. It is a weekly job nobody has time to do properly, which is exactly the shape of thing that should not be done by a person.

how it runs

The cost guard sits ahead of the actors and the vision calls, so a runaway session is stopped before it bills rather than reported after.

the workflows

0
Orchestrator
Intake form and weekly cron. Initialises the session and enforces the cost guard before anything expensive runs.
1
Discovery
Maps and SERP harvest for the niche and city, then a Claude pass that filters the list down to genuine competitors.
2
Harvest
Multi-platform ad pull with dedupe. Resolves brand names to ad-library page IDs by fuzzy match, writes the resolved IDs back, and pulls by ID from the second run onward. Matches below the confidence floor are dropped rather than guessed.
3
Scoring
Winner scoring from run duration, variant count and spend signals. Pure compute, so it is free to re-run against different weights.
4
Analysis
Claude vision teardown of the winning creatives, plus offer extraction from the landing pages. Pages are crawled in one batched run started before the vision work, and offers are extracted once per unique page rather than once per ad.
5
Synthesis
Pattern detection, then generation, then a critic pass. Produces the brief and the alerts.
E
Error handler
Marks failed sessions and sends failure alerts, so a silent break is not mistaken for a quiet week.

how it holds up

01
Precision improves session over session
Discovered page IDs are written back to the competitor sheet, so run two pulls by ID instead of by name. The system gets more accurate the longer it runs, rather than being fixed on day one.
02
Polling, not webhooks
Long actor runs are polled on a ninety-second interval with a hard ceiling rather than resumed by webhook. n8n offloads any wait past sixty-five seconds, so the execution is not holding a slot either way, and it ports to any instance with zero external config.
03
Cost guarded before the expensive part
The budget check runs in the orchestrator, ahead of the actors and the vision calls. A runaway session is stopped before it bills, not reported after.

what it will not do

Ad library coverage varies by platform and region. Some niches simply have thin data, and the system reports that rather than inventing competitors.
Runs weekly by design. Running it daily costs roughly seven times more for data that moves slowly.
It tells you what competitors are doing. Deciding what to do about it is still your job.

you supply

running cost

Roughly $1.50 to $5.00 per session at 20 competitors and 30 ads each.

Billed by your own providers, not by us. There is no platform fee and no seat count.

InstallCompetitor?

Install from $2.5K, about a week from kickoff to first session. Tell us about your setup and we will confirm it fits before taking anything.