A budget is not a strategy
Most media plans begin with a number.
Finance approves $250,000 for the month. The performance team distributes that amount across Google, Meta, TikTok and other channels. The plan then becomes a pacing exercise: spend the budget while trying to hold ROAS or CPA near target.
That workflow answers an administrative question:
How do we spend the money that was assigned?
It does not necessarily answer the economic question:
How much should we spend, and where should the next dollar go?
Those are different problems.
Paid media rarely scales linearly. The first $20,000 in a channel can reach the highest-intent or easiest-to-acquire demand. Additional spend pushes the system farther into the market. Auction costs change, audience quality changes, creative gets stretched and the return on the next block of spend can fall even while average ROAS remains attractive.
This workbook is built around that difference.
It uses current channel performance as a baseline, then asks the operator to model marginal tranches: what the next spend increment is expected to produce after incrementality and pre-ad contribution margin are considered.
What the workbook contains
The downloadable workbook has five operating layers.
Inputs captures current spend, attributed revenue, incrementality factor, pre-ad contribution margin, maximum planned spend, spend step and cash constraint for each channel.
Channel Model translates those inputs into platform ROAS, incremental revenue, contribution before ads, contribution after ads, incremental ROAS and break-even ROAS.
Marginal Tranches asks the operator to estimate what each additional block of spend is likely to return.
Allocation compares current contribution, the best positive next tranche, remaining channel headroom and cash headroom.
Dashboard summarizes the portfolio.
The model is intentionally simple enough to audit.
A workbook that produces a precise recommendation from fifty hidden assumptions is often less useful than a transparent model that makes the assumptions visible.
Start by separating attributed revenue from incremental revenue
A platform can report $200,000 in revenue without having caused the full $200,000.
Branded Search is the obvious example. Some customers would have navigated to the site anyway. Retargeting can capture people who were already close to buying. Upper-funnel channels can have the opposite problem: they may influence revenue later attributed to Search or direct traffic.
The workbook therefore includes an incrementality factor.
A factor of 0.75 means the planning model treats 75% of the attributed revenue as incremental for the purpose of the scenario.
This is not a substitute for experimentation.
It is a way to make existing causal evidence operational.
The factor should come from the best evidence available: lift tests, geo experiments, MMM calibration, documented channel studies or a conservative planning assumption.
Do not set every channel to 1.0 merely because the platform reports the conversion.
Contribution margin changes the meaning of ROAS
Revenue ROAS does not know what the business keeps.
Two channels can both report 3.0x and create very different contribution.
If one channel sells high-margin products and the other pushes discounted, expensive-to-fulfill products, equal revenue efficiency is not equal business efficiency.
The workbook therefore applies a pre-ad contribution margin after incrementality.
That value should represent the share of incremental revenue available to pay for media after the variable costs the business chooses to include.
The exact accounting definition should be agreed with finance.
The model is not improved by pretending there is one universal definition of contribution margin.
It is improved when media and finance use the same one.
The critical sheet is Marginal Tranches
The current channel model describes money already spent.
The Marginal Tranches sheet is where the resource becomes a planning tool.
For each channel, it provides a sequence of additional spend blocks and asks for a predicted marginal ROAS.
For example:
- next $10,000 at 3.0x;
- next $10,000 at 2.6x;
- next $10,000 at 2.2x;
- next $10,000 at 1.9x.
The workbook then adjusts those returns for incrementality and contribution margin and calculates the expected marginal contribution after ads.
A tranche is economically attractive in the model when the incremental contribution before ads exceeds the media cost.
This is a simplified representation of what Common Thread Collective calls the marginal frontier: the point at which the next dollar stops improving the chosen contribution outcome. CTC is a commercial agency and its specific models are proprietary, but the underlying diminishing-return principle is broadly useful.
Do not confuse a positive tranche with permission to spend infinitely
A positive marginal contribution result means the assumptions make that tranche look economically attractive.
It does not mean the estimate is true.
The model can be wrong because:
- creative performance deteriorates faster than expected;
- the channel saturates;
- competition changes;
- promotion conditions change;
- incrementality is overstated;
- margin changes;
- inventory constrains demand;
- tracking changes.
That is why the workbook also includes Max Planned Spend and Cash Constraint.
Long-term economics can be attractive while short-term cash makes the plan unsafe.
A subscription company can afford a high CAC eventually and still create a liquidity problem if payback takes twelve months.
How to estimate marginal ROAS without inventing numbers
The most difficult input is predicted marginal ROAS.
Several evidence sources can help.
Historical scaling periods. Examine what happened when spend moved materially before.
Controlled budget tests. Increase spend in a defined treatment and compare the incremental outcome.
Platform simulations. Useful as directional signals, not independent truth.
MMM response curves. Useful when the model is credible and calibrated.
Geo experiments. Stronger for causal return when the channel supports the design.
Operator ranges. When evidence is weak, use a downside/base/upside range rather than false precision.
The workbook contains example values only to demonstrate the mechanics.
Replace them.
The Allocation view should start a conversation, not end it
The Allocation sheet flags channels that have:
- positive modeled next-tranche contribution;
- remaining maximum-spend headroom;
- remaining cash headroom.
That creates a shortlist of candidate channels for incremental spend.
It does not automatically rank strategic value.
A channel with slightly lower modeled contribution may still deserve investment because it is learning a new market, generating new customers, building creative evidence or reducing concentration risk.
Professional allocation needs both exploitation and exploration.
If every dollar goes to the channel with the highest known short-term return, the portfolio can become dependent on one mature source of demand.
A recommended weekly workflow
Use the workbook in a recurring allocation meeting.
First, update current spend and attributed revenue.
Second, update incrementality assumptions only when evidence changes. Do not use them as knobs to justify a desired budget.
Third, refresh contribution margins when product mix or promotion changes.
Fourth, review the next marginal tranches.
Fifth, identify which assumptions are weakest.
Sixth, turn the weakest high-value assumption into a test.
For example:
Meta looks like the best next-dollar opportunity, but the marginal ROAS estimate is based on an old scaling period. Run a controlled spend test before allocating the full modeled headroom.
That is better than treating the workbook as an answer machine.
Failure mode: allocating from average ROAS
If Google averages 4.0x and Meta averages 3.0x, it is tempting to send the next dollar to Google.
That decision can be wrong.
Google may be at the edge of its efficient demand while Meta has unused headroom.
Average performance ranks history.
Marginal performance ranks opportunity.
Failure mode: using one incrementality factor forever
Incrementality changes.
A brand becomes better known. Search demand changes. Retargeting pools change. Creative expands the reachable market. Competitors enter.
A 0.70 factor from one test should not silently survive for three years.
Attach the evidence date to every major causal assumption and schedule revalidation.
Failure mode: optimizing the model instead of the business
A spreadsheet can always be made to approve more spend.
Lower the margin requirement. Increase incrementality. Raise marginal ROAS.
That is why planning assumptions need ownership.
Finance should own or approve contribution definitions. Measurement should own causal evidence. Media should own spend-response assumptions. Leadership should own the risk appetite.
The model becomes useful when no single team can quietly change every input.
Download the workbook
The workbook included with this resource is designed as a starting operating model, not a proprietary black box.
Use it to make the economics of allocation visible.
Then replace every example assumption with your own evidence.
The best outcome is not that the workbook tells you where to spend.
It is that the team can explain why the next dollar belongs there.
Download the Paid Media Budget & Marginal Return Workbook (XLSX)
