What does it cost to build reconciliation?

Operations today. Today’s cost splits into automation engineers (writing rules and integrations) and exception handlers (the ops team picking up what the automation doesn’t). Automation can be toggled off if no one is doing this on an ongoing basis.

Exception volume. Distinct exceptions per year = monthly volume × 12 × (1 − automation rate) × (1 − same-cause rate), spread over 22 working days a month. The same-cause slider captures the share of unmatched records that cluster under a common root cause and can be batch-resolved — only the remainder needs distinct human review.

Build effort. Person-week estimates reflect what we actually shipped at End Close with a senior fintech team. They cover design, implementation, tests, evals, and rollout — but not discovery, hiring, or scope creep.

Data ingestion. A base of 2 person-weeks for shared connector infrastructure plus 1.5 person-weeks per source for auth, schema mapping, incremental sync, backfill, retry, dead-letter, and drift handling. Adjust the source count inside the data ingestion card to see how it scales.

Maintenance. Adjustable on a slider, default 30% of build cost per year. Reconciliation is integration-heavy — bank feeds break, payment processors deprecate fields, accounting systems version their APIs. 30% matches industry norms for finance infrastructure.

Automation lift. Going from your current rate to 99.5% eliminates the matching fraction of remaining work. We apply that reduction to exception handler hours and floor at 0.5 FTE because someone always reviews exceptions. Automation engineer cost goes to zero in both the build and buy paths — they redeploy to product work.

What this misses. See the “Doesn’t include” notes on each verdict column. Linear cost scaling with transaction growth, infrastructure and observability, scope creep, and the engineering opportunity cost of putting your team on plumbing instead of product typically dwarf the numbers above.