Choose What to Buy, Configure, Connect, or Build

Facing three broken capabilities at once, unreliable deduplication, ungoverned enrichment spend, and no sequence-orchestration safety net, GTM Lab's first instinct is to build custom software for all three. That instinct is expensive and usually premature.

Four paths, considered in order, and never a free pass

Buy when a mature vendor already solves the problem well. Configure when native tooling covers the requirement once properly set up. Connect when a tool you already have covers the gap through integration. Build only when none of the above meets the requirement, and the gap is costly enough to justify ongoing maintenance. Even then, a build decision needs a named owner, a maintenance expectation, a way to tell if it's silently broken, and a condition under which you'd retire it.

Scoring three capabilities, honestly

Deduplication resolves to configure: native platform matching plus the review-gated merge process from Chapter 3 already meets the requirement without new tooling. Enrichment eligibility and cost governance resolves to build, but bounded: a mature provider's spend cap doesn't defer individual over-cap requests or gate per-segment eligibility (buy, ruled out); native workflow automation can trigger the call but can't atomically check a live spend ledger before each concurrent call (configure, ruled out for the gate itself); a middleware connector can orchestrate the trigger but doesn't expose an approval-queue view (connect, ruled out). So the atomic spend gate and approval queue get built, with the maintenance burden, owner, observability requirement, and sunset condition all stated up front. Sequence-orchestration reconciliation resolves to connect plus a thin rule layer: the sequencing tool's native CRM integration already handles unsubscribe and closed-won removal; it just doesn't recognize GTM Lab's own disqualification and customer-status fields, so a bounded custom rule layer checks those specific fields against sequence membership.

I applied this same discipline, ruling out buy, configure, and connect before building and keeping anything I did build small and bounded, across an entire GTM stack, not data quality alone. That effort eliminated $110,000 in tooling cost in a single fiscal year and cut overall GTM stack spend 40% year over year. Both figures describe the broader consolidation work, not the data-quality program on its own, and both are one company's result against its own contracts and prior baseline, not a guaranteed outcome of running this scoring exercise.

Carry this into your business

Score every capability against all four paths before writing a line of custom code, and for anything that does land on build, write down who owns it, how you'll know if it's silently broken, and when you'd retire it.