Why business data goes stale, and how we keep ours fresh

Data freshness has a reputation for being dark magic. It is not. It is a decay problem, and once you see it that way, the rules stop feeling arbitrary. Every business record is a claim about the world at a moment in time, and the world keeps moving after you write it down.

Everything below is just ways of noticing that movement before your user does.

Re-check on a schedule, not on a hunch

This is the foundation, and it is why Leadiy re-runs saved searches on a cadence rather than caching a snapshot forever. A record that was verified this week carries a different weight to one verified last spring, and we would rather show you the date than pretend the distinction does not exist.

• Store a verified-at timestamp on every field, not just every record - phone numbers and opening hours decay at completely different rates.
• Re-check the volatile fields far more often than the stable ones. A category almost never changes; a rating changes weekly.
• Expire loudly. A record that quietly ages into fiction is worse than one that admits it has not been checked in ninety days.

Coverage still matters, just not how you think

You will not lose a user because one phone number was wrong. You will lose them because the dataset looked complete and was not: plausible-looking gaps, silently dropped fields, a category that is 40% covered but presented like it is 100%. The fix is the same as good research generally: say what you know, say when you learned it, and say what you do not know.

The unsexy checklist

De-duplicate aggressively so you are not re-verifying the same business four times under four spellings. Keep a change log, because a rating that dropped 0.4 in a month is often more useful than the rating itself. And never present an estimate as an observation - label it, or do not ship it.

None of this is glamorous, and that is the point. Data quality rewards the team that behaves like a careful librarian at scale. Do the boring things and the dataset stays trustworthy. Skip them and no clever model will save you.