Background
The client ran a service focused on fixing and removing negative Google reviews, but prospecting meant hours of manual searching across Google Maps.
I built an automation that scrapes local businesses through Apify, reads review patterns, detects low ratings and unanswered negative feedback, ranks the highest-risk profiles, and stores everything in Google Sheets for outreach. Instead of guessing, the team now works only on businesses clearly struggling with their reputation.
The problem with the data
Google Maps data is massive and noisy. The system had to detect only recent, unreplied negative reviews, avoid duplicates across runs, keep the sheets clean enough for a non-technical sales team, and filter cost-efficiently so it never burned budget on profiles that had already been handled.
What I mapped first
Before building anything I mapped the lifecycle: prospect discovery, review scoring, detection of unreplied negatives, and export to the outreach sheets.
Then I studied the fields Apify’s Google Maps scraper actually returns, totalScore, reviews, reviewsDistribution, placeId, googleMapsUrl and the rest, to define the filters, the thresholds, and exactly which data had to be stored for later decisions.
What I built
The qualification logic pulls only businesses with ten or more reviews, isolates one to three star feedback, removes cases where the owner already replied, and sorts what is left by the newest negative review.
It outputs two synchronised datasets. A Business sheet with name, location, rating, total reviews, link, a summary of the worst recent review, price band and contact data. A Reviews sheet with the business, the review text, the stars, the link, the date and the city. Businesses are upserted so they never duplicate; reviews are appended so the history stays intact.
Handover
I delivered step-by-step documentation, explained each node, showed how to change locations, categories and niches, covered the cost-control settings, and trained the sales team on using the output.
The client can modify and run the workflow without engineering support. That was the point.
What changed
The business moved from guesswork to targeted outreach. Instead of wondering who might need help, the team sees which businesses have multiple fresh one-star reviews with no replies, which makes the sales conversation relevant before it starts.