How We Built a Proprietary 66,000-School Database When No Vendor Had the List

You sell to a niche that does not map cleanly to any vendor's filters. Maybe it is K-12 athletic departments, independent wine importers, or regional solar installers. You type your criteria into Apollo or ZoomInfo, and the result is either a few hundred mismatched records or a list padded with companies that have nothing to do with your buyer. The data is decayed, the bounce rate is ugly, and the segment you actually care about is invisible.

This is the exact situation we hit with GearLocker, a client selling into school sports programs. No vendor had a clean list of the right schools with the right contacts. So we built one: a proprietary database of 66,000 schools, assembled from scratch. Here is how, and how to decide when custom data is worth the effort versus buying a list.

When off-the-shelf data is the wrong tool

Apollo and ZoomInfo are built around firmographics that fit common B2B segments: software companies by employee count, agencies by revenue, SaaS by tech stack. They are genuinely useful when your buyer sits inside those categories. They fall apart when your segment is defined by an attribute the vendor does not track.

Schools are a good example. A vendor might know a district exists, but not which schools run varsity programs, how many students they enroll, or who runs athletics. The records that do exist are often role-based catch-alls (info@, principal@) that bounce or never reach a buyer. We see this constantly: bad and decayed Apollo data driving bounce rates that wreck sending reputation before a campaign even gets going.

The test is simple. If your ideal customer is defined by an attribute no vendor filters on, off-the-shelf data will always be a compromise. That is the signal to build. If your buyer maps to standard firmographics, do not build a database, just clean and enrich what you can pull.

Start with the authoritative source, not a vendor

The first move in building a custom database is to find the source of truth for your segment and work outward from it. For schools, that meant public education directories, athletic association rosters, and enrollment records: the data that actually defines the universe, not a sales tool's guess at it.

You are looking for a spine: a complete, authoritative list of the entities you sell to, even if it has no contact information attached yet. For GearLocker that spine was 66,000 schools. The point is coverage first. A list that is complete but thin beats a list that is rich but missing most of your market. You can always layer contacts on later. You cannot conjure the schools that a vendor simply never recorded.

Layer enrichment in waterfalls, not single lookups

Once you have the spine, the job becomes attaching the data that makes each record reachable and qualifiable: the right contact, a verified email, role, enrollment size, program details. This is where most teams go wrong by relying on one provider. One source covers maybe 40 to 60 percent of a niche list. The rest comes back blank, and you ship a half-empty campaign.

We run waterfall enrichment: query one provider, and for every record that comes back empty, fall through to the next, then the next, until coverage is maximized. We have built more than 1,800 production Clay tables doing exactly this, and enriched over 950,000 contacts. On a custom-built spine, waterfalls are what turn a directory of names into a workable target list. Verify every email before it enters a sequence: this is how we hold bounce rates between 0.15 and 0.9 percent instead of the double-digit bounce that decayed vendor data produces. High bounce is also the fastest way to land in spam, so clean data and deliverability are the same problem.

What the custom list actually produced

A proprietary database is only worth the work if it converts better than the list you could have bought. For GearLocker, the 66,000-school database produced 194 interested replies, from a segment no vendor could have targeted at all. That is the real argument for custom data: it is not that it is marginally better, it is that the campaign was not possible otherwise.

The same logic held for other niche segments we have built for. Chateau Constellation reached 177 interested wine importers by combining a custom importer list with trade-fair timing. LeverageRx pulled 143 interested physicians from a single campaign at a 46 percent positive reply share. In every case, the edge came from owning a dataset the competition could not pull off a shelf. You can see the full breakdowns on our results hub and the GearLocker case study.