A CMO I advise called last month in a mild panic. Organic search had "collapsed" 18% quarter-over-quarter, and the board wanted answers. We pulled the GA4 audit trail. The drop traced to a single Admin change: someone had switched Reporting Identity from Device-based to Blended in early April. Traffic hadn't moved. Demand hadn't shifted. A settings toggle had rewritten the executive dashboard.
This is not an edge case. According to Splinternet Marketing's 2026 audit findings, many GA4 properties show "drops" or "spikes" that trace back to Admin configurations rather than market signals. Before you reallocate budget based on channel performance, you need to audit the measurement layer itself.
The Three Levers That Rewrite Your Numbers
GA4's Admin panel contains three settings that can materially shift reported revenue by channel without touching traffic, spend, or creative. If your finance partner is reviewing marketing ROI and these settings changed mid-quarter, you're comparing apples to aircraft carriers.
Reporting Identity determines how GA4 counts users. Under Device-based, each browser or device is a separate user. Switch to Blended, and GA4 merges cross-device sessions using User-ID, Google Signals, and device identifiers. The result: fewer reported users, higher conversion rates, and a redistribution of credit across channels. For B2B accounts with logged-in portals or multi-device buying committees, this single toggle can move executive dashboards by double digits.
Blended identity also depends on Google Signals availability and user consent. If consent rates shift (say, after a cookie banner redesign), your user and conversion metrics can change even when campaigns stay flat. The CFO sees volatility; the cause is consent infrastructure, not demand.
Attribution Model controls how credit distributes across touchpoints. GA4 offers a cross-channel data-driven model alongside rule-based alternatives. Google's documentation confirms that data-driven attribution distributes credit based on observed conversion paths rather than assigning everything to the last click.
The practical effect: branded Paid Search often loses credit when earlier Organic, Paid Social, or Referral touches get recognized. In accounts with strong retargeting, branded campaigns may gain partial credit if they frequently appear in high-probability paths. Changing the model mid-quarter creates executive confusion unless you document the switch and restate prior periods.
Lookback Windows determine how far back GA4 looks for touchpoints. Longer windows (60 to 90 days for considered B2B purchases) increase the chance that upper-funnel SEO or prospecting campaigns receive credit. Shorten the window, and those same campaigns appear to decline overnight. No demand change occurred; you just stopped counting the touches that mattered.
Data-Driven Attribution Has Volume Requirements
The new Conversion Attribution Analysis Report (beta since February 2026) is a genuine improvement: it surfaces assisted conversions and categorizes touchpoints into Early, Mid, and Late stages. For the first time, GA4 shows which campaigns create demand versus which campaigns close it.
But data-driven attribution requires volume. The same source notes that GA4 needs 400+ conversions per key event and 20,000 total events to qualify for data-driven modeling. Below those thresholds, the model can fluctuate or fall back to rule-based logic. If your property sits near the boundary, you may see attribution credit shift week to week based on volume, not performance.
This matters for board reporting. Volatility in a low-volume account is not necessarily a tracking failure; it may reflect limited path data. Document the threshold and flag it in your executive summary.
GA4 and Ad Platforms Will Never Match
A recurring source of CFO frustration: GA4 says one thing, Meta says another, LinkedIn says a third. This is not a bug. GA4 uses session-scoped attribution, while Meta uses user-based attribution with view-through credit. The methodologies are structurally different.
Cometly's 2026 attribution guide frames the problem clearly: most teams are "piecing together fragments from five different platforms, hoping the answer sounds confident." The solution is not to pick a winner but to document the methodology behind each number and present a unified view that acknowledges the gaps.
For board-grade reporting, I recommend a three-column format: GA4 attributed revenue, platform-reported revenue, and a blended estimate with assumptions stated. Finance can then apply their own haircut rather than discovering the discrepancy in a pipeline review.
The Audit Checklist Before You Reallocate
Before you cut a channel or double down on a winner, run this five-point audit:
- Reporting Identity: Confirm the current setting (Admin → Reporting identity). Document any changes in the past two quarters. If you switched from Device-based to Blended, restate prior periods or flag the discontinuity.
- Attribution Model: Confirm the model in Admin → Attribution settings. If you changed models mid-quarter, note the date and expected directional impact on branded versus prospecting channels.
- Lookback Windows: Check acquisition and key-event lookback windows. For B2B cycles longer than 30 days, a 30-day window will systematically undercount upper-funnel contributions.
- Key Event Definitions: Verify that key events (formerly conversions) are defined consistently. A renamed or duplicated event can split credit and distort channel comparisons.
- Volume Thresholds: Confirm your property meets data-driven attribution minimums. If not, document the fallback model and flag volatility risk.
This audit takes 30 minutes. It can save you from a budget reallocation that solves the wrong problem.
The Board Memo Format
When presenting channel performance to the board, lead with assumptions. A one-paragraph methodology note at the top of the memo prevents the "but Meta says something different" derailment:
"Channel revenue is attributed using GA4's data-driven model with a 60-day lookback window. Reporting identity is set to Blended, which merges cross-device users and may reduce reported user counts relative to prior periods. Platform-reported figures (Meta, LinkedIn) use different methodologies and are shown separately for reference."
This is not defensive hedging. It is CFO-safe communication. Finance teams respect marketers who state their assumptions up front and acknowledge measurement limitations. They distrust marketers who present a single number as ground truth.
Model or It Didn't Happen
Attribution is not a set-and-forget configuration. It is a measurement system that requires governance. Every quarter, someone on your team should verify that Admin settings match your documented methodology, that key events are defined consistently, and that volume thresholds are met.
The alternative is reacting to settings artifacts as if they were market signals. You cut a channel that was working. You double down on a channel that was borrowing credit. You present a board deck that falls apart under CFO scrutiny.
Audit the measurement layer first. Then reallocate budget.
