Datanamo
All systems

Retained client, Datanamo contract. 2018 to present.

Property enrichment platform for US real estate

The data backbone of a US real-estate data company. Around 200 automated county scrapers and four data vendors feed one clean property database of close to 800,000 tracked auctions and tax liens, which the client sells through its own platform and API. Owned end to end for eight years, from requirements to production support, with every upgrade done live.

~500K
auctions tracked
300K
tax liens tracked
~200
county scrapers, 8 years in daily production

How data moves through it

SourcesCollectCoreDeliverCounty auctions~500K tracked, ~200 scrapersTax lien records300K trackedProperty data providersthird-party, per recordParcel and insurance datathird-party providersScrapersproxies, retriesVendor enrichmentquotas enforced in codeValidate and repairaddresses, parcelsProperty database~800K records, 8 years liveClient platformPublic APIDaily exportsidempotent, retriedMeasure, every runSource coveragea quiet county is an alertVendor healtherrors, quotas, spendStalled enrichmentreapers and deadlinesalerts before a customer notices

Relevant to you if

Your product is data collected from many sources that change without warning, you pay vendors per record, and you need the pipeline to run for years with one accountable engineer rather than a department.

What it changed

The client sells data that automation collects instead of a research team. Vendor spend is capped in code so enrichment never overruns the budget, and alerts catch a broken source before a customer does.

How it works

  • Around 200 county scrapers collect auctions and tax liens, close to 500,000 and 300,000 respectively, and four data vendors enrich them into one property model, with a cleanup layer that fixes addresses and catches bad records before they overwrite good ones.
  • Enrichment runs as queued jobs with vendor quotas and retries built in, so a vendor outage or a bad night never stalls the pipeline or blows the budget.
  • Alerts fire when a source goes quiet or a vendor degrades, and slow pages get fixed at the query, so the platform has stayed fast while the data grew for eight years.