Switching a cross-channel reporting vendor looks like a procurement project on the slide deck and like a demolition job in practice. The contract is the easy part. The hard part starts the morning after signature, when a new pixel is sitting next to the old one, two dashboards disagree about last week's ROAS, and the CFO wants to know which number to trust in the forecast.
Most switches fail here, not at selection. The buyer picked a reasonable tool, but the cutover was treated as an IT ticket instead of a measurement project. Historical data got stranded. UTM conventions quietly changed. Conversion definitions drifted. Six weeks in, nobody can reconcile the new ROAS against the old one, and finance stops believing either.
So what does a cutover look like when it actually holds?
Decide What Actually Travels With You
Not everything in your old tool is worth moving. The instinct is to export it all; the discipline is to decide what the new system has to answer that the old one did. Three categories belong on the plane:
- Raw event history, at the lowest grain the old vendor will give you (visit-level or user-level, with timestamps, source, UTM, revenue). Aggregated weekly rollups are almost useless for rebuilding attribution later.
- The conversion dictionary: every event the business counts as a conversion, its definition, its deduplication rule, and which campaigns are allowed to claim credit for it.
- Spend reconciliation: platform-reported spend by day, by campaign, with the currency and timezone the old tool used. This is what makes ROAS comparable at all.
What you leave behind is dashboards, saved views, and any attribution model you were only tolerating. Rebuilding a bad report in a new tool is how you import your old problems. If you have never written down which conversions your campaigns should be optimizing toward, the migration is the moment to do it — the exercise in choosing conversion goals is faster now than after the pixel is live.
Rebuild the Conversion Dictionary Before You Touch the Pixel
Attribution tools disagree about ROAS mostly because they disagree about what a conversion is. Deduplication windows, view-through credit, assisted conversions, offline uploads — each vendor ships defaults, and the defaults are rarely identical. Two concrete examples worth knowing before you cut over:
- GA4 switched its default attribution model to data-driven in November 2023, replacing last-click as the out-of-the-box setting, which means channel reports often show fractional conversions rather than whole ones. If your old tool was still on last-click, the switch itself will move numbers.
- GA4's default lookback window is 30 days for acquisition and 90 days for other conversions, and changes to that window are not retroactive. If your new vendor uses a different default, the historical gap is baked in from day one.
Write the dictionary down before you install anything. One row per conversion: the trigger, the dedup key, the lookback, the view-through rule, the model. Then make the new tool match it, not the other way around.
Fix the UTM Taxonomy While Nobody Is Looking
A new reporting tool inherits whatever discipline your campaign URLs already have. Different casing or formatting — utm_source=Facebook vs utm_source=facebook vs utm_source=facebook.com — causes GA4 to treat each variation as a separate source, fragmenting campaign data and distorting attribution. Every cross-channel tool has the same failure mode.
Before the parallel run starts, pick one casing, one naming convention, and one owner. Rewrite the live links in your highest-spend campaigns first. Build a lookup table that collapses the historical variants into their canonical form so year-over-year comparisons still work after the switch. This is unglamorous and it is the single highest-leverage hour of the whole project.
Run Both Tools in Parallel Long Enough to Trust the Delta
Nobody who has done this twice turns the old tool off on the 31st and the new one on the 1st. The parallel window is where you earn the right to defend the new numbers. How long depends on volume and sales-cycle length, not on how impatient the executive sponsor is.
Two to four weeks is the floor for a high-volume ecommerce account where a week delivers a statistically meaningful sample. B2B accounts with a 60-day consideration cycle need the parallel run to span at least one full cycle, or the new tool will look wrong simply because it has not yet seen conversions mature. The vendor questions worth asking before you sign include how the tool back-fills conversions during that window; if the answer is "it doesn't," plan a longer parallel period.
Reconcile the Numbers Before Finance Sees Them
During the parallel run, the new tool and the old tool will not agree. That is the point. The job is to explain the gap in categories a non-marketer can follow, so the delta becomes a known quantity instead of a credibility problem.
Most of the gap collapses into a handful of causes. Platform-reported numbers add up to more than reality because conversions get double- and triple-counted across walled gardens — any tool that deduplicates will show lower totals than the sum of in-platform reports, and that is a feature, not a bug. Meta's window changes compound this: after iOS 14.5, Meta reduced its default to a maximum of 7-day click, 1-day view, so conversions outside that window never make it back to the originating campaign in the first place. Model differences account for a chunk; iOS signal loss accounts for the rest. Apple's ATT reduced the share of trackable Apple traffic in the United States by 55 percentage points, from 73% to 18%, and that gap sits inside every tool's numbers.
Write the reconciliation as a one-page memo before anyone asks. Known delta, known causes, residual unexplained variance. If the residual is under a few percent, you have a defensible number.
Sequence the Cutover So Rollback Is Still Possible
The actual flip is a sequence, not an event. The order below is the one that keeps a rollback cheap if week six surprises you.
Keep the old pixel firing for the full parallel window and for a buffer after cutover. Rebuild integrations into the new tool one platform at a time — the Google Ads connection first, because it is the highest-spend and the best-documented, then the social platforms, then CRM and offline. A well-scoped cross-channel analytics rollout usually sequences the ad platforms before any CRM write-back, because CRM errors are expensive to unwind and the ad-side numbers stabilize faster. Your list of available integrations drives the order more than the project plan does; a connector that has to be custom-built belongs at the back of the queue, not the front.
Only decommission the old tool when three things are true: the reconciliation memo has been accepted by finance, the new tool has survived one full reporting cycle without a correction, and the historical export is stored somewhere you can still query a year from now.
Defend the New Number in the First Board Meeting
The first finance review after cutover is where migrations get undone. The ROAS number moved, the CFO wants to know why, and "the new model is more accurate" is not an answer that survives a follow-up question. What survives is a short written note that names the trade-off the new tool is making and why that trade is the right one for the business.
Have context ready for the obvious questions. Google deprecated first-click, linear, time decay, and position-based models in part because, at the time of deprecation, less than 3% of Google Ads web conversions used them — the industry had already moved on, and anyone still comparing to a 2022 last-click baseline is comparing to a model the platforms themselves abandoned. The signal environment also changed underneath everyone: Apple shipped iOS 14.5 with App Tracking Transparency on April 26, 2021, and the cross-app attribution infrastructure most tools were built on has not fully recovered. A new tool that shows lower reported conversions than the old one is often showing a more honest number, not a worse one. Say that out loud, with the dates, before anyone has to ask.
What a Clean Cutover Buys You
A migration done well is not just a new dashboard. It is a reset of the measurement discipline — a clean conversion dictionary, a consistent UTM taxonomy, a reconciliation memo finance has signed off on, and a historical archive you can still query. Teams that treat the switch as a measurement project rather than a software install come out the other side with better numbers than they had going in, which is the only outcome that justifies the disruption. Teams that treat it as a vendor swap tend to run the same reconciliation argument every quarter until somebody switches tools again.
The cutover is where the value is won or lost. Spend the budget there.