How to allocate ad budget across multiple locations (without just splitting it evenly)

The short answer: give each location enough to cover its fixed basics, then put every additional dollar where it buys the most extra customers, and keep moving money until the cost of the next extra customer is roughly the same everywhere. That is the whole idea. The rest of this guide is how to do it with real data, and where it goes wrong.

Most multi-location brands do something simpler. They split the budget evenly, or they weight it by each location’s sales, and adjust when a regional manager complains loudly enough. Both habits are understandable. Both quietly overfund locations that are already saturated and starve ones that would have grown.

Three ways to split an ad budget across locations

Almost every allocation method in the wild is a version of one of these three. They differ in what they assume about how a location responds to spend.

• Even split. Every location gets the same amount. It assumes all locations respond identically, which they don’t. It is defensible in year one, when you have no evidence, and in franchise systems where fairness is part of the deal.
• Sales-weighted (or share-of-revenue). Each location gets a budget in proportion to its sales, often a fixed percentage like 3% of revenue. It assumes bigger locations get more from each dollar. Often the opposite is true: your busiest store is busy because demand there is already strong, and ads mostly reach people who were coming anyway.
• Response-based. Each location gets budget according to how many extra customers its spend actually produces, measured against what would have happened without the ads. It assumes nothing; it uses evidence. It also needs that evidence, which is why most brands never get here.

A sensible path is to start even or sales-weighted, then drift toward response-based as measurements come in. You don’t need perfect response curves for every location to beat a static split. You need to know which few locations are clearly over- or under-served.

Diminishing returns: why the last dollar matters more than the average

Every location has a response curve. The first few thousand dollars reach the most persuadable people in the trade area. The next few thousand reach people who are less interested, or the same people again. Eventually you are paying to show ads to someone who has seen them nine times and still prefers the place across the road.

This is why average cost per extra customer is a trap. A location might show a respectable $30 average while the last $2,000 you added bought customers at $80 each. The average blends the cheap early customers with the expensive recent ones. When you are deciding where the next dollar goes, only the marginal number matters.

The practical rule follows directly. If Location A’s next extra customer costs $80 and Location C’s costs $36, moving money from A to C gets you more customers for the same total spend. Keep moving until the marginal costs converge. In economics this is the equal-marginal principle; in a budget meeting it is just common sense with numbers attached.

How to measure each location’s response to spend

You cannot read a response curve off an ad platform. Platform-reported conversions tell you who clicked, not who came because of the ad, and they are least reliable exactly where demand is already high. (Our guide to why click attribution misses in-store sales covers the mechanics.) What you need is a comparison: what happened at a location when spend changed, against what similar locations did when it didn’t.

1. Use the spend changes you already made. Budgets move all the time: a regional push, a paused campaign, a new market. Each change is a natural experiment if you compare the affected market with markets that held steady over the same weeks.
2. Make deliberate steps. Raise spend 20% in a handful of markets and cut 20% in a handful of others for six to eight weeks. Two points on a curve tell you its slope, which is what you need for allocation.
3. Measure customers, not revenue alone. Count guests, members, patients or booked jobs by location, by week, from your POS or booking system. Ticket size wobbles for reasons unrelated to ads.
4. Keep the range. Every estimate comes with uncertainty. A location whose lift range runs from −50 to +400 extra customers has told you very little. Don’t move serious money on it.

One wrinkle: you usually buy media by market, not by store. If three locations share a metro, you allocate to the metro and then split within it by radius targeting or by location-level campaigns. Measure at the level you can actually control.

A worked reallocation across four locations

Here is an illustrative example. A four-location urgent care group spends $40,000 a month on paid social and search. It has run a couple of deliberate step tests, so it knows roughly what each location’s last $2,000 bought, and has a reasonable estimate of what the next $2,000 would buy.

Downtown has the biggest budget because it has the biggest revenue. On average it looks fine. At the margin it is the second-worst place to spend. Airport road is worse still: it is near capacity on weekday evenings, so extra demand mostly turns into longer waits and walkouts.

The group moves $4,000 out of A and $4,000 out of D, in two $2,000 steps each, and puts $4,000 into each of B and C. Using the step-test estimates (with each successive step worth a bit less than the one before):

Same budget, 82 more extra patient visits a month, and the blended cost per extra visit drops from $30.77 to $28.94, about 6%. That is not a dramatic number on its own. It compounds, though: every month, at no extra spend. And notice what happened to the averages. They converged, from a $26–$45 spread to $23–$34. That convergence is the signal you are getting closer to the right split.

Then the group stops and measures. The gains at B and C are estimates from a curve; the real response might be flatter. After six to eight weeks the new read either confirms the move or tells you to back off. Allocation is a loop, not a decision.

Capacity: the constraint most budget models forget

A location can only serve so many customers. A restaurant has a fixed number of seats on Friday at 7pm. A clinic has a fixed number of exam rooms. A gym floor gets unpleasant past a certain headcount at 6am. Advertising into a full location doesn’t create customers; it creates queues, and sometimes bad reviews.

Check peak utilization before you add budget anywhere. If a location runs above roughly 85–90% of its practical capacity at peak (your own operators know the real number better than any model), extra spend should be pointed at off-peak demand or taken elsewhere. Daypart targeting and offers aimed at quiet hours can turn a capacity problem into a growth opportunity. Blanket awareness spend can’t.

Set floors and ceilings before you optimize

Pure response-based allocation, run blindly, will eventually tell you to spend nothing in some markets. Sometimes that is right. Often it ignores things a model can’t see.

• A floor for every location. Brand search, map listings and a minimum presence on the main social channel protect demand you already have. Cutting them to zero can cost more than the model expects, because the lift you measured at higher spend says little about what happens at no spend at all.
• A ceiling per location. Cap any location at the spend level where you have evidence. If your step tests went up to $14,000, don’t extrapolate the curve to $30,000.
• Protected budgets. New openings, remodel relaunches and franchise co-op commitments sit outside the optimization. Measure them separately. Openings in particular will look like ad lift if you leave them in.
• A maximum step size. No location moves more than 20% in one cycle. This keeps reads clean and stops a single noisy estimate from wrecking a market.

Common budget allocation mistakes

• Rewarding the biggest store. High sales usually mean strong organic demand, which means ads there mostly reach people already on their way. Sales-weighting builds this mistake into the formula.
• Using platform ROAS to rank locations. Platform-reported returns are highest where branded search and retargeting intercept existing demand. They will consistently tell you to fund the locations that need it least.
• Reallocating on one month of data. A single month is mostly noise. Weather, a local event or a competitor’s promotion can swing a location 10% on its own. Wait for a read with a range narrow enough to act on.
• Moving everything at once. If every location’s budget changes in the same month, you have no steady markets to compare against and can’t tell what worked. Stagger your moves.
• Forgetting cannibalization. Two locations five minutes apart share customers. Extra spend at one can simply move customers from the other. Measure at the market level when trade areas overlap.
• Treating the curve as permanent. Response changes with creative, season and competition. A location that was saturated in July may have room in January. Re-test the biggest budgets at least twice a year.

When not to reallocate

Sometimes the right move is no move. If the ranges on your location reads overlap heavily, you don’t yet know which locations are better, and shuffling money will mostly add noise. If you have fewer than about ten markets, you may not have enough steady comparisons to read individual locations well; group them into clusters and allocate between clusters instead. And if a location’s problem is operational (slow service, short staffing, a broken sign), ad budget won’t fix it, and pulling budget from it won’t either.

One more case: a location in a test. If you are running a geo holdout, its budget is fixed until the test ends. Move it early and you have paid for an experiment you can no longer read.

A monthly allocation routine you can actually keep

1. Pull the latest lift reads by location: extra customers, cost per extra customer, and the range.
2. Drop any location whose range is too wide to act on, or that is inside a test or an opening window.
3. Rank the rest by marginal cost per extra customer. Check peak capacity for the top half.
4. Move 10–20% from the two or three most expensive locations to the two or three cheapest, respecting floors and ceilings.
5. Write down what you moved and why, with the date. Future you will need it to read the result.
6. Wait six to eight weeks. Repeat.

It is unglamorous. It is also how budgets get better every quarter instead of being reset from scratch every January.

Questions people ask about allocating ad budget by location

Written by The Countlift team at First Byte in The Woodlands, Texas. Questions or corrections? Email hello@countlift.com.