How a UA team reconciled spend across 5 ad platforms automatically

An illustrative scenario — a composite of UA reconciliation setups, with realistic numbers.

The setup

An in-house UA team spending ~$400k/month across Meta, Google, TikTok, Snapchat and Apple Search Ads, AppsFlyer as referee. Platform-claimed conversions exceeded MMP-attributed by 30–80% depending on channel — normal physics, but unexamined: reconciliation was a quarterly spreadsheet exercise, and one quarter it surfaced a TikTok postback gap that had quietly mispriced a channel for six weeks.

What was set up

• The reconciliation view — per channel, per week: spend, platform-claimed, AppsFlyer-attributed, variance % — appended into the team's existing report
• Pinned pulls — Meta at 7-day-click/1-day-view stated in the header; Google conversions trailing-window re-pulled; Snap swipe vs view labeled; SKAN in its own columns
• Baselines + thresholds — four weeks of history per channel, then flags on variance movement beyond each channel's band
• The schedule — weekly with the main report; flags into the UA channel

The numbers

Reconciliation time went from a quarterly half-day to zero marginal effort; detection latency for tracking breaks went from weeks to the same run.

What changed operationally

Variance became a monitored series instead of a quarterly argument. A changed attribution setting on one platform tripped the band within a week and was reverted before it contaminated a month of CAC. Budget meetings switched denominators: MMP-attributed for decisions, platform-claimed labeled as the platforms' own optimization signal.

What stayed human

Deciding what each variance jump meant — the flags name the cell, the team names the cause.