Choosing a market-data provider is one of those decisions that looks like a feature comparison and turns out to be an architecture decision. The wrong choice shows up months later as per-vendor adapters, licensing surprises and a latency number that does not survive contact with production. Here is the checklist worth actually working through.
Coverage and symbology
Start with breadth — every asset class you trade now and plausibly next — but weight symbology just as heavily. A provider that covers everything under inconsistent symbols still makes you write mapping code forever. One consistent symbology across stocks, crypto, forex and commodities is worth more than a slightly longer coverage list.
History depth and point-in-time
For anything research-driven, ask two questions: how far back does history go, and is it point-in-time? Fundamentals that have been silently restated will leak the future into your backtest. Deep, point-in-time history is the difference between a study you can trust and one you cannot.
Latency, measured your way
Treat latency claims as hypotheses to test, not facts. Ask where the number is measured, insist on percentiles, and verify in-region against your own clock. A feed you can benchmark honestly beats a bigger number you cannot reproduce.
Alt-data and total cost
The hidden cost of most stacks is vendor sprawl: one provider for prices, another for news, another for fundamentals, another for anything alternative — each with its own key, schema, bill and outage. Consolidating price feeds and alt-data behind one provider cuts integration work and incident surface at once. Count the total cost of the stack, not the sticker price of the feed.
The short version
Favour consistent symbology over raw coverage counts, point-in-time history over depth alone, latency you can measure over latency you are told, and one key over five. Pig Data was built to win on exactly those axes — the widest coverage, the lowest latency, and the alt-data others make you source elsewhere, behind a single key.