Define the customer and the spending window
Count new customers, not orders, if the question is customer acquisition cost. Decide how repeat buyers, canceled orders, and refunds are treated. Use the same policy across tests. A campaign report and an accounting export may use different definitions, so reconcile them before presenting a confident result.
List included costs: media, creative production, samples, commissions, and relevant outside help. You may also show a narrower media-only figure, but label it. Founder time can be shown separately for an early experiment. Hiding all non-media effort makes a labor-intensive channel appear easier to scale than it really is.
Calculate a hypothetical test
Assume $600 media spending and $200 creative cost produce 20 new customers under the chosen attribution method. Media-only acquisition cost is $30; the combined figure is $40. If each first order produces $28 contribution before acquisition, the first-order result is negative by $12 per customer on the combined basis.
Gross revenue of $1,400 would not change that conclusion by itself. The product and delivery obligations consume part of it. If five orders later cancel or qualify for exclusion under the stated policy, recalculate with the appropriate denominator and cash effects. Explain the change rather than quietly keeping the earlier favorable figure.
Handle attribution without pretending certainty
A buyer may see an advertisement, read a review, and return directly later. Different systems may credit that purchase differently. State the window and method, and use tagged links or other proportionate measurement where appropriate. Follow applicable privacy and platform requirements.
For a small test, compare platform-reported purchases with actual completed orders and customer information you legitimately have. Do not add results from two platforms if both claim the same sale. Where attribution remains uncertain, present a range or combined business-level result instead of a falsely precise channel ranking.
Set the next experiment from the constraint
If acquisition exceeds first-order contribution, investigate audience relevance, offer clarity, page performance, price, and repeat behavior. Change a specific element and record the test. Lowering the product price can increase conversion while reducing contribution, so evaluate both together.
Do not justify losses with imagined lifetime value. Repeat purchases need a plausible product cycle and observed evidence. An occasional durable purchase may not support frequent repeat revenue. Expansion is more defensible when acquisition, fulfillment, and customer satisfaction remain workable as spending increases, not merely when one small campaign produces an attractive screenshot.
Distinguish a learning test from a scalable channel
A first campaign may include one-time creative work and unusually intensive founder involvement. Show those costs rather than hiding them, but also identify which are expected to recur. If a $200 demonstration video can be reused, its allocation across later tests may change; the cash was still spent upfront. Keep cash cost and analytical allocation distinct so neither view is mistaken for the other.
Watch how performance changes when spending increases. A small audience of highly interested people may be exhausted quickly, leaving less relevant prospects. A $20 acquisition cost on ten customers does not prove the next 1,000 can be acquired for the same amount. Expand in controlled steps and compare cohorts under similar measurement rules.
Set a test loss ceiling before starting. If the experiment's purpose is to learn which message reaches a relevant audience, specify the learning and budget rather than pretending every test must be profitable immediately. At the end, decide whether the information justifies another test. Repeatedly relabeling an unprofitable channel as learning without a new question creates open-ended spending. The acquisition model should make the next decision clearer: improve the offer, change the audience, reduce a cost, test repeat behavior, or stop that channel.
Reconcile campaign spending to the dates when it occurred and customer results to the chosen cohort. Do not combine one week's spending with a different week's purchases simply because the totals look favorable. Note delayed conversions and incomplete return windows. If the sample is too small for a stable estimate, report the counts and continue only under the predetermined test budget. An honest provisional result is more useful than a precise acquisition number that changes dramatically when one additional order is canceled.
Decision checklist
- Are new customers distinguished from orders?
- Are included costs and attribution rules explicit?
- Is acquisition compared with contribution rather than revenue?
- Are repeat-purchase assumptions supported by observed cohorts?