GA4 Cross-Channel Budgeting: Forecast Spend Before You Commit

Most media plans get written in a spreadsheet, agreed in a meeting, and then left alone for three months. By the time anyone can say whether the split was right, the quarter is over and the money is gone. The feedback loop is too slow to change the decision it was meant to inform.

Google put cross-channel budgeting into Google Analytics on 16 January 2026 to shorten that loop. You build a plan inside GA4, watch how it paces against the number you committed to, and model what a different allocation would return before you spend it. Two tools sit inside the feature: projection plans for the activity already running, scenario plans for the budget you have not committed yet.

It is in beta and not every property can open it. Even so, the second half of this article matters, because the models inherit whatever cost data and conversion settings you feed them. A confident forecast built on sloppy inputs is worse than no forecast at all when a client is signing off budget against it.

What cross-channel budgeting is

Cross-channel budgeting is a planning workspace in the Advertising section of Google Analytics. Google describes it as a way to optimise paid channel investment and track performance across channels. It pools spend, conversions and revenue from every channel it can see into a single model, then lets you ask two questions: where is this heading, and where should the money go instead.

The reports live under Advertising, in a Budgeting drop-down, and they only work on desktop. Before anything opens you need the Advertising section configured: a linked Google Ads account and conversions set up. No link, no conversions, no model.

Google's own summary of what the tool is for could have been written by anyone who has sat in a budget review. Evaluate cross-channel budgets so investment decisions rest on evidence rather than on who argues best. Monitor pacing weekly so you adjust inside the quarter. Build annual or quarterly plans from historical performance rather than from last year's spreadsheet plus an optimistic multiplier.

Projection plans: is the plan on track

A projection plan takes a target and tells you whether your channels are pacing toward it. Google lists five questions it answers:

• Am I on track to spend the planned budget?
• Am I on track to drive the target number of conversions?
• Am I on track to hit the target revenue?
• Which channels are overperforming or underperforming?
• Where should budget move to improve overall performance?

Creating one takes four inputs. Path: Advertising, then Budgeting, then Projection, then Create plan. You give the plan a name, a planning period such as Q4 2026, a target KPI (budget, conversions or revenue), and a target value for that KPI. Your property has to be eligible and hold enough historical data, and GA4 shows which conversions qualify for the model.

The chart is where people misread the output, so learn the three elements first. The solid line is actual performance to date. The dotted line is the projection. The shaded area is the model's confidence bounds. When the bounds are wide, the model is telling you it is unsure, and the usual reason is that something changed in a way seasonality and holiday effects cannot explain. That is not a forecast you take to a board meeting. It is a prompt to go and find out what moved.

Above the chart, a recommendations panel names the most and least efficient channels, based on historical data and model estimates. Google's worked example makes the workflow concrete: a 12-week initiative with a $100,000 revenue target, checked five weeks in, shows a significant under-pace on both spend and revenue. The recommendations name one channel as the most efficient and another as the weakest, so $5,000 moves from paid social into paid search, and the projection updates to show the plan back on track. Whether that channel detail is trustworthy in your account is a separate question.

Scenario plans: where the money should go instead

A scenario plan starts from a budget you have not spent. You give it a future period, a KPI (conversions or revenue), a budget level and a single conversion to optimise for, and it returns a response curve: the allocation across channels that should produce the highest return for that money. You can move along the curve to compare return on investment at different budget levels, and you also see what happens if you change nothing.

The requirements are stricter than for projections. A scenario plan needs data-driven attribution active in the GA4 property, plus campaign data arriving through linked product integrations or data imports. Offline data joins the model if you have offline imports set up. Only one conversion can be selected at a time, so a property optimising for both leads and purchases needs two plans.

Two mechanical rules catch people out. Scenario plans can only start on a future date, so you cannot point one backwards at a quarter that already ran. And the response curve is frozen at the moment of creation: it reflects market conditions on that date, and it will not update unless you change a plan setting such as budget, date range or target metric. Want a fresh curve? Build a new plan. The discipline is to rebuild plans regularly rather than trust one saved six weeks ago.

One detail from the documentation is worth acting on now even if you are not eligible yet. The budgeting models consider all click-through conversions, YouTube engaged-view conversions and all cost data. Include impressions in your data import where you have them, because that prepares the same import for future model updates. Adding a column later is easy. Backfilling three years of it is not.

What it will not do

Four limits, stated by Google, and each one matters when someone asks you to defend a number.

These tools are for planning only. They do not change budgets or spend in your connected accounts. Anyone expecting a projection plan to touch a campaign is looking at the wrong product.

The outputs are modelled estimates, not guarantees. Google repeats that in both the release note and the help documentation. A forecast is an argument about the future built from the past, and it is only as good as the past you fed it.

Availability is limited. Every page in the documentation carries the same banner: the feature may not be available to your property and the team is expanding it. Behind that, two model methodologies exist. The DDA-based budgeting model is in beta and available to Google Analytics and Google Analytics 360 customers. The Meridian-based model, built on Google's open source Meridian library, is in alpha and only available to selected 360 customers.

The Overview page lists your saved plans, your product links and a cost import tile showing usage, the date of the last upload and a file match rate. Its manual integrations panel for third-party platforms is still static text today. Non-Google channels therefore reach the model through cost imports alone.

The unglamorous work that decides whether the numbers hold up

A forecast model is a mirror. Point it at a property with broken cost data and it will hand back a confident, well formatted reflection of the mess. Before you trust a projection, run these checks.

Cost imports must carry a currency. Since 28 July 2026 GA4 refuses to process a cost import that does not state its currency, either mapped from a column or set as a fixed value for the whole dataset. That requirement exists precisely because budgeting projections inherit whatever you feed them. If cost figures were imported before the change, projections built on them inherit the old distortion. Add the currency on the next edit of each source, then rebuild any plan that read the old data.

The file match rate on the Overview page is a quality signal. A rate well below full means part of your uploaded campaign data never matched a campaign in GA4, so a slice of your spend is invisible to the model. Chase that number before you chase the forecast.

Conversions have to be right, and aligned. The model reads click-through conversions, so a conversion action with a lookback window that does not match your sales cycle produces a forecast built on the wrong attribution. GA4 lets you set custom windows per conversion, and the settings are mirrored in Google Ads conversion management, which is exactly why the two can now be compared side by side under Conversion management.

Campaign parameters need to be intact. If aggregate identifiers are missing from campaign URLs, GA4 raises a diagnostic warning that campaign data accuracy is affected. A model trained on incomplete campaign data will still produce a curve.

Source grouping changes what the model reports. Google added Source Group and Source Platform dimensions to make budgeting plans easier to action. Those consolidated values are what a recommendation panel names, so a channel that looks efficient may be three platforms bundled under one label. If you need to know which, that is a report you run yourself.

How to judge a projection someone else built

Ask four questions before you repeat a forecast in a meeting.

Which conversion is the plan optimising for? A model pointed at newsletter signups will happily tell you to move budget away from the channels that drive revenue.

What attribution is in play? Scenario plans require data-driven attribution, so a plan built on last click comparisons is not comparable to a scenario output.

How much cost data does the model actually have? Check the import dates and match rates, not the chart.

How wide are the confidence bounds, and when were they drawn? A narrow band around a curve created six weeks ago is still a six week old opinion.

Keep the review rhythm simple. Fifteen minutes a week: check pacing against target, read the recommendation line, and write down any change you make. That log is the difference between a forecast that adapts and one that goes stale.

A sensible first month

1. Check eligibility by looking for the Budgeting entry under Advertising, and confirm you are signed in on desktop with Admin or Editor access.
2. Fix cost imports before anything else: currency on every source, then rebuild the reports that read them.
3. Confirm data-driven attribution is active and that offline imports are connected if a chunk of your conversions happen off site.
4. Pick one conversion that maps to money, not to engagement, and build a projection plan for the quarter you are in.
5. Build one scenario plan for the next quarter with two budget levels, and compare the response curve rather than the headline number.
6. Revisit weekly. If the confidence band widens, stop optimising and start investigating.

The short version

Cross-channel budgeting turns a budget conversation into a model you can interrogate, and the model is only as honest as the data underneath it. For many UK properties the curve is not yet available anyway. The preparation is: currency-tagged cost imports, conversions that match the sales cycle, attribution set deliberately, and campaign data complete enough to be worth forecasting from.

If a forecast keeps disagreeing with reality, the disagreement is usually an upstream measurement problem wearing a budget costume. That is when an audit of the collection layer pays for itself.