Plan 2023 budgets as ranges when attribution is partial
Attribution will be incomplete again next year. Build the budget around ranges, reconciliation to your own revenue data and pre-agreed rules for moving money, not around one channel report.
Budget planning usually starts with a spreadsheet that has one cost per acquisition for each channel. That spreadsheet assumes each channel's reporting is accurate enough to compare. For 2023 that assumption is weaker than it was a few years ago, and the plan should say so openly.
What is changing, and what is not
The signal environment is a mix of new tools and continuing limits. Consider what has happened this year:
- Apple released SKAdNetwork 4.0 with iOS 16.1 on October 24, with up to three postbacks, hierarchical source identifiers and coarse conversion values that appear when privacy thresholds are not met.
- Meta's third-quarter results, released October 26, showed revenue of $27.7 billion, down 4 percent year over year, with ad impressions up 17 percent and the average price per ad down 18 percent, according to one analyst's summary. That analyst attributes the lower price partly to Apple's tracking prompt; Meta's own commentary pointed to ad challenges "coming from Apple." For buyers the practical reading is that signal quality on iOS remains a factor in what platforms can deliver.
- Google has said standard Universal Analytics properties stop processing hits on July 1, 2023, so any plan that relies on year-over-year comparisons in your analytics tool needs a GA4 baseline.
- Google announced in February that the Privacy Sandbox is coming to Android, with the existing advertising ID supported for at least two years from that announcement, so Android signals are stable for now but not guaranteed.
None of this means attribution disappears. It means each platform's view is partial in a different way, and the gaps move over time.
Budget in ranges, not point estimates
Instead of "Meta iOS will deliver a $30 cost per trial," write "between $25 and $40, depending on how much is under-reported." The width of the range should reflect how much you trust the channel's reporting. A range changes how decisions get made: you don't cut a campaign because it landed at $38 when its plausible range includes $28.
Reconcile to a source you own
Each month, compare the sum of platform-reported conversions with the number your backend records for the same period. The gap is not an error to fix but a number to track. Hypothetically, if platforms together report 800 purchases and your backend shows 1,000 paid by new customers, then about 80 percent of conversions are attributed somewhere. The remaining 20 percent still came from somewhere, and your plan should say who gets credit until you know better, for example proportionally to reported share.
Keep that ratio by platform and by operating system, since the gaps differ. Consistency matters more than precision: a stable ratio lets you trust trends even when the level is off.
Fund a measurement line
Set aside part of the budget, even a small one, for tests that answer questions attribution can't: holding out a region from a channel for a few weeks, pausing a campaign type and watching total conversions, or running a brand lift survey. Hypothetically, 5 percent of paid spend reserved for these tests is cheaper than discovering in the fourth quarter that your biggest channel was taking credit for sales that would have happened anyway.
Write the rules for moving money in advance
The point of planning under uncertainty is to avoid reacting to noise. Agree on the triggers now:
- Cost per result must be outside its range for a set number of weeks, not days, before budget moves.
- No more than a fixed share of a channel's budget moves in a single step.
- Any cut to a channel is checked against total conversions in your backend, not just the platform report.
- A named person owns the call, and the reasoning is written down.
Split the budget by role
A simple split helps: a core of spend on channels and campaigns you understand and can measure well, a growth portion for new placements and formats with a set test duration, and the measurement reserve above. Apple's new Today tab and product page placements, and Google's move toward Performance Max, are the kind of opportunities that belong in the growth portion, with a stated question and an end date.
Plan the calendar around the calendar
Put the July 1 analytics change into the plan as an event, with the GA4 setup and a parallel run completed well ahead. Put a review point after each quarter to re-measure the reconciliation ratio. If it moves, the plan should move with it.
The takeaway: a good 2023 budget doesn't pretend to know each channel's true cost per result. It states the range, tracks the gap to your own revenue data, reserves money for testing, and decides in advance how money will move.