Work out a brand/performance split from how a category actually buys — instead of starting from a ratio.
Splitting a growth budget
Derive the split from how your category buys — don’t start from a ratio.
Rules like 60/40 summarise patterns observed across many brands. They can’t tell an individual brand its best allocation without knowing how its category works. This page builds the allocation up from the purchase system itself: how many purchases there are, which ones you already expect to win, which of the rest each kind of media can actually change — and what that costs.
01
Define the purchase system
How many category purchase occasions actually exist? Pick a category type to start, then adjust.
Buyer groups
Subscribers · 50% of buyers · 12 purchases/yr
Top-up buyers · 50% of buyers · 6 purchases/yr
Expected category purchase occasions this year9M
Buying frequency is not always usage frequency: someone who replaces something 12 times a year but buys 4 at a time makes 3 purchases, not 12.
The model starts with an expected pool of category purchase occasions over the planning horizon. Media can change which brand wins them; growing the category itself is outside this simplified model.
02
Map the flow of demand
Split this year’s purchase occasions by what would happen without any new media.
Expected to choose you 1.4M
“Already yours” does not mean guaranteed. It means expected under the no-new-media counterfactual. Performance cannot claim incremental value merely for capturing purchases that would have happened anyway.
An economically live opportunity can plausibly be identified and treated near the purchase. That does not mean performance will cause the purchase.
Expected elsewhere · not directly addressable 3.5M
Not directly addressable through the performance path defined here. Changing prior choice probability becomes the primary media mechanism.
Direct-response-addressable purchase occasions — derived from how each buyer group buys, not chosen because it feels reasonable. Some purchase occasions are easier to identify and intervene in through direct response than others.
54%
03
Two different media jobs
Brand changes the odds before the purchase occasion — the stock. Performance intervenes when an addressable buying window goes live — the flow.
Brand · changes the odds before people shop
money → productive reach → response → changed chance of being chosen → extra purchases
“Productive reach” already contains the things that make a reach count: the right audience, enough attention, enough frequency, and a message able to change how people think about you. They’re collapsed into one number here on purpose. Reach and delivery aren’t outcomes in themselves — the response to the exposure is what matters.
Performance · steps in on live buying windows
money → coverage of addressable windows → share of them you’d otherwise lose → switch rate → extra purchases
Performance pays to treat addressable windows — including ones you’d have won anyway, because it can’t tell them apart in advance. The windows replenish as people come back to buy. Spending more covers more of them, then saturates; it doesn’t make each treatment work better. Brand does, a little: buyers it has warmed up are easier to switch.
Brand and performance interact in two ways. Brand wins purchases outright — sales that need no performance at all, which leaves slightly fewer undecided buyers in the pool. At the same time, it warms up the buyers it doesn’t win outright, making them easier for performance to switch. Most of brand’s payoff still comes from the sales it wins on its own, but on balance it also leaves performance more productive, not less.
Brand’s effect carries forward (65% a year over 3 years); a performance-won purchase is counted once. Both are assumptions you can change in the mechanics.
ATTRIBUTED ≠ INCREMENTAL
Attribution can tie a transaction to an ad. The economic question is whether the purchase would have happened without it. This model values performance on incremental lift only.
Purchase windows treated 2.5Mof which you’d otherwise lose 2Mincremental purchases 163K
Treatment does not mean causation. Only the modelled incremental difference counts as economic impact.
Sales reported as “organic” or not attributed to direct response are a residual reporting bucket, not a cause. They can contain the effects of brand equity, habit, distribution, price, past media and more.
04
Two things that improve the system
Neither creates purchases on its own. Data improves what you can identify and target. Testing improves what you know about response.
Audience & data · improves what you can identify and target
Performance faces three questions: can I find the buying window? · can I find the person worth treating? · does treating them change the choice? Data helps with the first two. Lift decides the third.
More buying windows identifiable
Better known-audience data helps recognise who is approaching a purchase.
Addressable: 54% → 54%
Less treatment wasted on likely buyers
Suppression and better signals tilt treatment toward people you’d otherwise lose — but never beyond how many there are.
Treated windows you’d lose: 81% → 83%
More productive brand targeting
Better seeds and lookalikes concentrate brand spend on valuable prospects.
Productive reach cost: −1%
First-party data has value when it makes future media more productive — not because collecting records is valuable in itself.
Learning · improves what you know about response
How certain are these assumptions?
If plausible response ranges materially change the answer, the model may recommend testing before scaling.
05
Derive the allocation
Start from the work that needs doing, and let the split fall out. Or start from a fixed budget and see how much of it is worth spending.
The work
Target500K purchases
Expected category purchase occasions9M
Expected to go to you anyway1.4M
Expected to go elsewhere7.7M
Addressable buying windows4.9M
… treated by performance2.5M
… of which you’d otherwise lose2M
Brand: incremental purchases, year 1337K
Brand: carried into later years (discounted)317K
Performance: incremental purchases163K
… of which possible only because brand warmed buyers up20K
Gain vs the same media plan without the data investment5.5K
Learninguncertainty not shown
Recommended investment
Brand$630K
Performance$270K
Audience & data$13K
Learning$0
Total$910K
Value after investment (3-yr)$2.36M
At the minimum-cost solution, brand accounts for 337K incremental purchases in year 1 and performance for 163K — 20K of which happen only because brand warmed those buyers up. The two come from their joint economics — neither gets first call on demand.
A reporting grouping for Finance. It is not the brand/performance media split that 60/40 describes.
06
Compare with 60/40
60/40 describes the media split: brand ÷ (brand + performance). To test it fairly, hold audience & data and learning at the derived amounts, and change only that split.
Three different views — don’t mix them
Media split — compare this with 60/40
Brand ÷ (brand + performance)70% / 30%
Total growth allocation
Brand69%
Performance30%
Audience & data1%
Learning0%
Finance view
Immediate demand capture30%
Future-growth capital70%
Derived media split vs forced 60/40
Held fixed: audience & data, learning$13K
Forced 60/40 — total cost to reach the target$930K
Derived split — total cost$910K
Saving$12K
Try the five category types — the derived split moves, because the purchase systems differ. Here 60/40 lands close to the derived plan — for this system, not as a rule.
Lowest total cost to reach the target at each media split│ 60/40 ● derived
Audience & data and learning are held at the derived amounts at every point. Gaps are splits that can’t reach the target at any budget.
A brand/performance split should not be chosen. It should emerge from:
how many purchase occasions exist, and how often they replenish;
which ones you already expect to win;
which lost occasions direct response can actually intercept;
how often intervention genuinely changes the outcome;
how efficiently brand can change future choice;
and whether better data, or better knowledge, changes those economics.
60/40 gives you a ratio. This shows the system from which a ratio should emerge.
That isn’t a case against brand, performance or 60/40 as an empirical average. It’s that one heuristic can’t give the best allocation for every category, because the demand systems underneath differ.
Numbers illustrative
Response curves, every assumption, what moves the split, and what’s arithmetic vs assumed.
Try this
1Switch between the five category types and watch the derived media split move.
2In Bulk replenishment, raise the pack size. What happens to performance’s opportunity?
3Switch to a fixed budget, turn on uncertainty, and lower the test cost until the model says test first.
Takeaway
The split is downstream of the demand system. 60/40 gives you a ratio; this shows the system a ratio should come from.