Maya Chen
Head of Market IntelligenceMaya leads Heroku's market intelligence practice.
Articles by MayaAbout Heroku
We watch the public web so your decisions don't fly blind.
Heroku is a web data and market intelligence company. We turn the largest open record of commercial activity — the public web — into structured, quality-measured intelligence that teams use to price, position and plan with confidence.
Heroku analysis floor — market monitoring, San Francisco.
Mission
Turn public-web signals into structured market intelligence.
Markets used to be observed in snapshots — quarterly reports, annual studies, gut feel. The web changed that: every price, listing and positioning statement is now published, dated and observable. Our mission is to give decision-makers access to that observation layer with the rigor of an engineering discipline and the honesty of a good analyst. Not more data — better-earned confidence.
Operating Principles
How we work when nobody is checking
Six principles that govern every program, dataset and deliverable — and the order in which they conflict-resolve: ethics, then quality, then speed.
Transparency first
Methodology, quality metrics and limitations are documented for every dataset. If we can't show our work, we don't ship the number.
Public data, public rules
We collect only publicly available information, at respectful rates, within documented boundaries. No exceptions, no creative readings.
Quality is measured
Quality gates with three outcomes — pass, degrade visibly, quarantine. Silent degradation is the only unforgivable failure.
Decisions before data
Programs are designed backwards from the decisions they serve. Data without a decision attached is decoration.
Analyst-grade synthesis
Machines watch, humans explain. Every deliverable has a named analyst accountable for its caveats.
No lock-in
Documented schemas, export options and migration support. Clients stay because the data is good, not because the door is locked.
Data Philosophy
Honest about what data can answer
Public-web data excels at observable behavior — prices, assortments, availability, public language. It cannot read minds. We are blunt with clients about the boundary: monitoring tells you what the market did and when; primary research and judgment explain why. Programs that blur that boundary produce confident nonsense, which is worse than honest uncertainty.
The second commitment is point-in-time honesty. Every record carries the moment it was observed, and historical views are rebuilt under the rules that governed them — so backtests and trend lines mean what they claim to mean.
Technology Approach
Boring on purpose
Our stack favors dependable over clever: collection profiles versioned like code, fingerprint-based drift detection, canonical schemas with deprecation windows, and delivery layers that fail loudly. We adopt new tooling only when it reduces operational risk — the excitement budget goes into the intelligence, not the infrastructure.
Pipeline engineers who have all been paged by a source redesign at 3 a.m.
Analysts who write the caveats they'd want to read.
A quality team with veto power over every delivery.
Solutions architects who say "no" in feasibility studies when the data can't support the ask.
Company Timeline
Eight years of watching the web work
Founded in a basement datacenter
Three engineers, one rack, and a conviction that public-web data deserved production-grade engineering.
First monitoring platform
Price and assortment monitoring for six retail clients. The quarantine-and-alert principle is born during a source redesign.
Normalization layer ships
Canonical schemas and entity resolution go live — cross-source comparison stops being handcrafted.
Data APIs launch
The platform's internal datasets become a product, with versioned contracts and a sandbox tier.
Market intelligence practice
Analyst synthesis joins data delivery: decision-first programs replace ad-hoc reporting.
Custom solutions practice
Feasibility-first engagements for questions the catalog can't answer — regulated markets, niche supplier landscapes.
Quality framework published
Our four-dimension quality framework and gate outcomes are documented publicly.
One system, every surface
Our full site experience — web, WordPress and Blogger — now ships from one canonical design system, keeping brand and quality identical everywhere we publish.
2019founded — San Francisco
6practices across data & intelligence
40+data professionals across three offices
24/7pipeline monitoring coverage
Team
The people behind the datasets
Authors of most of what you'll read in Insights — and the humans on the other end of technical inquiries.
Daniel Brooks
Principal Data EngineerDaniel designs the collection and normalization pipelines behind Heroku's datasets.
Articles by DanielPriya Shah
Lead Analyst, PricingPriya studies pricing dynamics across digital commerce.
Articles by PriyaMarcus Lee
Director of Data QualityMarcus owns the quality gates that decide whether a dataset ships.
Articles by MarcusElena Rossi
Search Intelligence LeadElena researches search behavior and category demand.
Articles by ElenaArjun Mehta
Solutions ArchitectArjun works with enterprise clients to design data programs that survive contact with reality — legacy systems, procurement cycles, and all.
Articles by ArjunWant the people behind the data in your corner?
Tell us the decision you're trying to improve. We'll tell you honestly whether public-web data can support it — and how.
Hannah Kim
The 2019–2026 timeline is a nice touch. Out of curiosity — what was the single biggest architecture decision that made multi-region monitoring feasible?
Derek Wong
Came for the services page, stayed for the data philosophy. The transparency-first framing is why we shortlisted you.
4 comments
“Measure quality, publish the numbers, quarantine what fails” — more companies should state their data philosophy this explicitly. The operating principles section reads like it was written by people who’ve actually run pipelines.
Every line of that section has a scar behind it, Samuel. Appreciate the note.