How a Marketing Agency Cut 40 Hours of Weekly Prospecting

The client is a marketing agency whose customers are real estate brokerages. Its reps spent about 40 hours a week manually finding leads across listing sites, spotting intent signals and updating the CRM before a single outreach went out. That manual research capped how much outbound prospecting the team could do, and left reps little time for the outreach itself.

Atherial built an overnight scraping and enrichment pipeline. It scrapes realtor.com through ScraperAPI, using a custom adaptive-concurrency and checkpointing layer so long runs can scale up and resume cleanly, and processes about 12,000 listings across 300 ZIP codes each night. Most agent contact details come from the scrape itself. Where they are missing, Perplexity Sonar fills in phone numbers and emails. Qualification is rule-based: listing price, agent sales history and other data points sync to the CRM as filterable fields, and tags and pipeline stages are set deterministically in code, so reps start each morning with a scored, prioritized list.

The four-week build started with solution ideation and CRM setup, defining which listing and agent data points the agency's reps actually used to qualify a lead. Atherial then built the scraper on ScraperAPI with its own adaptive-concurrency and checkpointing layer, so overnight runs across 300 ZIP codes could throttle themselves and pick up where they left off after a failure. Contact data comes mostly from the scrape, with Perplexity Sonar used to enrich agent profiles that were missing a phone number or email. Qualification was kept in code: price, agent sales history and other signals are written to the CRM as filterable fields, and tags and pipeline stages are set by deterministic rules, which keeps every lead's priority explainable to the reps working it.

Marketing agencies, brokerages and sales teams that prospect from public listing or directory data at volume, where reps currently build lists by hand. It fits best when qualification rules are clear enough to encode, so the pipeline can tag and stage leads deterministically and use AI only to fill missing contact details.