The media mix allocation calculator above takes your current spend and cost per acquisition on up to four channels and returns the split of a fixed budget that produces the most conversions, along with the conversion gain and the blended cost per acquisition that split implies. It assumes each channel gets gradually less efficient as you spend more into it, which is why it never simply dumps the whole budget into the cheapest channel.
Arb Digital's team runs this calculation at the start of every quarterly planning cycle, because most budget splits are inherited rather than chosen. Once a channel has a line in the plan, its number tends to survive from year to year regardless of what the performance data says — and the cost of that inertia is usually larger than any single optimisation inside the accounts.
What This Media Mix Allocation Calculator Does
Enter each channel's name, current spend and current cost per acquisition. The calculator converts those into a response curve for each channel, then solves for the allocation of your total budget that maximises total conversions across all four. The recommended spend per channel appears as a bar, with the current figure alongside it so you can see the size of each proposed move.
The headline number is the additional conversions the reallocation would produce at the same total budget. The supporting grid shows conversions before and after, and blended cost per acquisition before and after, so you can judge whether the change is worth the disruption. You can also change the total budget to model a larger or smaller pot — set it above the sum of your current spends to plan an increase, or below to plan a cut.
How to Use It
- Name your channels. Use whatever unit you make decisions at — platforms, campaign types, or individual campaigns. The maths does not care, but the comparability of the CPAs does.
- Enter current spend for each channel over a consistent recent period, ideally a full month or longer so short-term noise is smoothed out.
- Enter current cost per acquisition for each channel from the same period and the same measurement source for all four.
- Set the total budget to allocate. Leave it equal to the sum of your current spends to reallocate what you already have.
- Click Calculate and read the recommended split, then treat it as a direction of travel rather than an instruction.
The Formula / How It's Calculated
Each channel's conversions are modelled with a square-root response curve: conversions = k × √spend, where k is calibrated from your own data. Since current conversions are spend ÷ CPA, the constant works out as k = √spend ÷ CPA. That curve captures diminishing returns — doubling spend on a channel produces roughly 1.41 times the conversions, not twice as many.
Maximising total conversions across channels under a fixed budget then has a clean solution. Setting the marginal return equal across all channels gives an optimal spend for each that is proportional to k², which simplifies to spend ÷ CPA². Each channel's recommended share is its own spend ÷ CPA² divided by the sum of that quantity across all channels.
With the default figures, the shares work out at roughly 66%, 26%, 5% and 3%, which turns $50,000 into about $33,000 on Google Search, $13,100 on Meta, $2,600 on YouTube and $1,300 on LinkedIn. Projected conversions rise from about 920 to about 973 — a gain of roughly 53 conversions and a drop in blended cost per acquisition from about $54 to about $51, with no additional budget.
Why the Model Never Sends Everything to the Cheapest Channel
A naive optimiser told to minimise cost per acquisition would put the entire budget into whichever channel currently has the lowest one. Every experienced media buyer knows what happens next: the cheap channel gets expensive. Its cheapest inventory and highest-intent audience are already being bought, so incremental spend reaches progressively less responsive people at progressively higher prices.
The square-root curve encodes that behaviour directly. Because the marginal return of each channel falls as its spend rises, the optimum equalises marginal returns rather than average ones — every channel keeps a share, and the cheapest channel gets the largest but not the entirety. This is why the recommended split in the default example still leaves money on the two expensive channels. They are worse on average and still worth something at the margin.
The Assumption You Are Buying Into
Every media mix model is only as good as the response curve behind it, and this one uses a single fixed curve shape for all channels. That is a deliberate simplification. Real channels have different curve shapes: search is constrained by the number of people actually searching, so it saturates hard at a ceiling this model does not know about, while broad social prospecting can absorb far more budget before its efficiency really collapses.
The practical consequence is that the recommendation is reliable in direction and approximate in magnitude. If the tool suggests moving 60% more into search, the direction is well supported by your own data; the exact figure assumes search has room to absorb it. Check the ceiling separately — Google's impression share reporting shows how much search volume is genuinely left to buy, and audience size does the same job on social. Never let a model override a constraint you can observe directly.
Cost Per Acquisition Is Not Comparable Across Channels by Default
The single most common way this calculation goes wrong is upstream of the calculation. If your search CPA comes from a last-click model and your social CPA comes from a platform using a seven-day click and one-day view window, the two numbers are measuring different things, and optimising between them optimises the measurement rather than the media.
Before entering anything, make sure every CPA comes from one source with one attribution rule — usually your own analytics platform or a data warehouse, not each ad platform's self-report. The attribution model documentation in Google Analytics Help is a good place to confirm which rule your own reporting is applying before you trust the numbers. Upper-funnel channels are systematically penalised by last-click measurement, so a model fed last-click data will consistently recommend shifting budget away from awareness and towards capture, until there is no demand left to capture. If you have incrementality results, use incremental CPA figures here instead; they are the closest thing to a genuinely comparable number across channels.
Move in Stages and Re-Measure
The recommended allocation is calculated from the efficiency your channels have today, at the spend levels they have today. The moment you move budget, those efficiencies change, which means a single large reallocation lands you in territory the model has no data about. Large moves also disrupt platform bidding algorithms, which need time to re-learn after a significant budget change.
A safer approach is to move a portion of the gap — perhaps a third — hold for a full measurement period, then recalculate with the new figures. Two or three of those cycles typically get you most of the available gain with far less risk, and each cycle tells you whether the model's assumptions are holding in your account. Track the blended result with the CPA calculator between cycles, and use the ad budget calculator to plan what each new channel-level budget should buy in clicks and conversions.
Arb Digital builds media plans from one consistent measurement source, models the response curve of each channel against real spend history, and reallocates in controlled stages so performance is never gambled on a single move.
Paid Advertising Services Google Ads & PPC ServicesCommon Mistakes to Avoid
- Mixing attribution sources — comparing a last-click search CPA with a platform-reported social CPA optimises the reporting, not the media.
- Making the full move at once — efficiencies change as soon as spend moves, and bidding algorithms need time to re-learn.
- Ignoring channel ceilings — search cannot absorb unlimited budget because search volume is finite, whatever the model says.
- Using a single unusual month — a promotion, an outage or a seasonal spike will distort every CPA you enter.
- Zeroing out a channel entirely — a channel with no spend produces no data, so you lose the ability to re-evaluate it later.
Related Free Tools From Arb Digital
Check each channel's efficiency with the CPA calculator and the cost per lead calculator, plan the new channel budgets with the ad budget calculator and the marketing budget calculator, then judge the outcome with the ROAS calculator. Browse the full free online tools hub for more.
Frequently Asked Questions
It models each channel with a square-root response curve calibrated from your own spend and cost per acquisition, then allocates the budget so the marginal return of the last dollar is equal across channels. That solution makes each channel's share proportional to its spend divided by the square of its cost per acquisition.
Because efficiency falls as spend rises. The model assumes each additional dollar in a channel produces slightly fewer conversions than the last, so concentrating everything in one place would push its marginal return below what the other channels still offer.
Yes, as long as you use the same definition for every channel. The maths works with any cost-per-outcome measure, but mixing leads on one channel with sales on another will produce a recommendation that optimises for the wrong thing.
Cap it at what the channel can realistically spend, then rerun the calculation with the remaining budget across the other channels. Search in particular is limited by actual search volume, which no allocation model can create.
Use one source for all channels. Platform figures use each platform's own attribution rules and typically overlap with one another, so a single analytics or warehouse source usually gives a more comparable basis for allocation decisions.
After each measurement period in which spend levels changed materially, since the response curve is calibrated on current spend. Quarterly is a reasonable rhythm for stable accounts, with a recalculation after any large budget change.
No. Full media mix modelling uses regression across long spend histories with controls for seasonality, price and external factors. This tool is a single-period allocation model built on one simplifying assumption about diminishing returns, intended for planning rather than for econometric measurement.
Figures produced by this tool are planning estimates only — they depend on a simplified diminishing-returns assumption and on the comparability of the cost data you enter.
