Choose a completed outcome
Define success before the campaign begins. For a pickup offer, distinguish an order started, payment collected and order collected. For an event, distinguish an inquiry, deposit and fulfilled booking. Count cancellations and refunds. Select the same reporting window for every channel you compare so one campaign does not receive a longer opportunity to convert.
Record the offer, dates, customer group and capacity. If the kitchen could only fulfill twenty additional orders, that limit belongs in the plan. Selling out can be a successful operational result without proving the campaign would scale to a larger audience.
Calculate contribution before claiming a return
Start with net collected sales after discounts and refunds. Subtract the product cost, incremental labor, payment charges and other variable costs caused by serving those orders. This gives a contribution estimate before campaign spending. Use the merchant’s actual cost records where possible and label missing inputs as estimates.
Illustrative arithmetic makes the distinction clear. Forty genuinely additional $20 orders create $800 in sales. At an assumed 40% contribution rate, they provide $320 before marketing. A $250 campaign leaves $70 of incremental profit. On these assumptions, campaign ROI is 28%, calculated as $70 divided by $250. These are not customer results.
Include the work behind the campaign
Add creator fees, hosted product cost, staff coordination, editing, usage rights, paid media and messaging. Include the relevant subscription cost consistently. If a monthly service supports several campaigns, choose an allocation method before reviewing results. Also keep a whole-month view so allocations cannot make every campaign look profitable while the total loses money.
Record setup separately from ongoing work. An initial catalog cleanup may benefit future campaigns, but it still consumes resources. Show both current cash spending and labor cost, including owner time, so the decision reflects the complete operating burden.
Separate tracked purchases from additional purchases
A campaign link or code helps connect an order to a touchpoint. It does not prove that the message caused a purchase. Regular customers may use a code for a visit they already planned, and several channels may claim the same order. Deduplicate using transaction records and keep attribution separate from your estimate of additional demand.
Google’s Conversion Lift guidance makes the same conceptual distinction through treatment and control groups. That is a measurement method, not a promise that a particular restaurant qualifies for Google’s product or that a small campaign can produce a precise causal estimate.
Use the strongest practical comparison
For a discretionary follow-up campaign, a randomly selected comparison group can help estimate what eligible customers would have bought without that message. Keep pricing, observation time and other promotions comparable. Preserve normal service and transactional communication. Watch for customers appearing in both groups or seeing the same creator post outside the test.
When randomization is impractical, compare similar weekdays and record weather, holidays, menu changes and other promotions. Explain that this baseline is less conclusive. A small difference across a few orders should lead to more observation, not a confident percentage lift claim.
Decide what the next dollar should do
Review reach, relevant visits, checkout completion, fulfillment, contribution and repeat purchases as separate stages. A campaign with good local interest and weak checkout may need clearer terms or a better purchase path. A campaign with strong first orders and no return visits may have found an occasional purchase rather than a repeat habit.
Compare the next proposed campaign with doing nothing, a merchant-owned alert or another affordable channel. Increase spending only when the expected additional contribution and delivery capacity justify it. Save the assumptions alongside the result so the next review can challenge them.
Common questions
Is return on ad spend the same as profit?
No. Revenue divided by ad spend omits product costs and often other campaign costs. Use contribution and complete marketing spend when deciding whether the activity improved the business.
Should I count future customer value?
Show observed repeat purchases separately. Future value can be a planning scenario, but assumed months of spending should not be reported as money already earned.
What if transaction data is incomplete?
Report the coverage and keep unknown sales unknown. Directional evidence can still guide a small experiment, but it should not become a precise ROI claim.
Sources and further reading
Reviewed October 23, 2026