What it is
Every service business already has a pile of intelligence sitting in public view — its Google reviews, and its competitors’. Nobody reads 2,500 reviews. Signal does, and turns them into a decision: exactly what to fix, who’s beating you at it, and where the whole market is weak.
It has two modes, because two very different people need this data:
- Provider view — a contractor sees their own review performance benchmarked against the market: a scorecard, rating trend, a theme matrix (pricing, punctuality, communication, cleanup…), a named-technician leaderboard, pricing failure modes, a negative-review autopsy, and a ranked list of what to fix first. A dropdown re-benchmarks every chart against any competitor.
- Network view — a retailer or franchisor managing many providers sees all of them ranked, a momentum quadrant that flags providers declining before their star rating shows it, the category’s shared weak spots, and recruitment targets.
The proof
Signal isn’t a mockup. The live demo is a real build: 10 plumbing companies in Atlanta, 2,547 real Google reviews across 24 months, every one classified and benchmarked. The same pipeline runs on any city and any service category — HVAC in Denver, electricians in Phoenix, roofers anywhere — by swapping the review export and re-running.
How it works
Every review is classified by an LLM against a fixed taxonomy — 12 operational themes plus sentiment, named employees, job type, pricing signals, and emergency flags — then cached, so the expensive step runs once. From there it’s pure aggregation: each provider scored against a computed market baseline, because the insight is always the gap, never the number in isolation. The whole thing ships as one self-contained HTML file that renders offline — no dashboard to log into, no data leaving the building.
What makes it trustworthy
Most review dashboards quietly launder model guesses into confident-looking charts. Signal is built to survive being read in front of a client:
- Every aggregate links back to the verbatim reviews behind it — any classification can be spot-checked.
- Every percentage shows its denominator, and thin samples are flagged low-confidence and excluded from ranked comparisons rather than smoothed over.
- It reads reviews as reviews, not jobs — Google reviews are self-selected and bimodal, and the methodology says so plainly.
- Where the data can’t support a metric, it shows an honest empty state and explains why, instead of inventing a number.
What you could do with it
The same data answers very different questions depending on who’s asking. A few examples:
- Find the one theme dragging you below the market — and spend your next dollar exactly there.
- Catch a location or competitor declining before their star rating shows it.
- Steal the exact words happy customers use in your category, and put them in your copy.
- Rank a whole network of locations or providers on one honest, comparable scale.
That’s four of twenty. Read all 20 ways to use review data → — written for any industry, from single shops to franchise networks.
Who it’s for
Signal is a close cousin of an idea I’ve written about — the Local Services Ranker — and it shares DNA with Demand Engine: both turn messy public data into decisions. It’s built for the businesses I most want to see win: independent service and repair shops, and the retailers and franchisors whose networks depend on them. Custom builds for your category are a consulting conversation.