Your media partner reports a positive lift. Your production partner sends a proposal for the next batch of videos. For a brand sourcing products or content from China, those documents may arrive from different teams, in different currencies, and with very different ideas of success. A statistically detectable effect does not, by itself, connect them.
ONCE proposes a short approval document: compare an incremental-order interval with a matching break-even requirement, then state exactly what the next purchase is expected to change. This is an evergreen decision method grounded in current source checks, not a claim about a new platform launch or an ONCE client result.
Start with the decision the report can actually support
The American Statistical Association’s March 7, 2016 release on statistical significance and p-values states that significance does not measure effect size or importance, and that business decisions should not rest solely on a threshold. We rechecked it on November 8, 2026.
Google Ads Help’s About certainty of lift, checked the same day and carrying no publication date, describes uncertainty for Brand Lift, Search Lift and user-based Conversion Lift. It notes that low certainty can arise from limited data or small lift, and provides intervals around estimates. That is guidance for the specified Google studies, not a universal reporting convention.
The useful question is therefore larger than “Did it pass?” How much changed, how uncertain is that estimate, and what contribution could the change generate? This article addresses performance spending evaluated through incremental orders. Awareness or recall cannot be converted into immediate purchases without a defensible model.
Build one matching economic denominator
Before doing arithmetic, have the analyst identify the market, campaigns, comparison condition, outcome and time window. Confirm whether the reported quantity is orders, people, revenue or relative lift. If only a percentage is available, ask for an order estimate and interval under the study’s own scope; multiplying it by unrelated total-store sales will not fix the gap.
Break-even incremental orders = costs to be covered in the matched scope / contribution per incremental order.
Use the brand’s own contribution after relevant product, fulfilment, payment, marketplace, commission, discount and expected return costs, but before the media and production expenses being evaluated. Do not deduct the same expense twice. If contribution is zero or negative, this division cannot produce a usable order hurdle.
A blended margin can also mislead if the treatment changes the mix of products, markets or customer types. Segment where necessary, or estimate incremental contribution directly. Match the accounting window to the conversion and refund delays under review. The analyst and finance owner should agree on the scope before the result is used to approve spending.
ONCE’s interval-versus-break-even review
Here is a fully hypothetical retrospective exercise, not a rate card, benchmark or measured result. A completed campaign has $12,000 in matched costs and a fixed $40 contribution per incremental order. It needs 300 incremental orders to cover those costs. Assume the analyst has produced valid 95% confidence intervals for the same campaign scope and period.
| Illustrative order interval | Position against 300 orders | Decision note |
|---|---|---|
| 30 to 90 | Above zero, entirely below break-even | Detectable positive lift does not support covering the matched costs. Review economics or strategy before authorizing more production. |
| -50 to 500 | Crosses zero and break-even | The economic conclusion remains unresolved. Establish whether a better study could change the spending choice before buying more research. |
| 350 to 800 | Entirely above break-even | Stronger support for covering this historical outlay under the stated assumptions. It does not establish returns at a higher spend level or with new creative. |
An interval’s endpoints are not guaranteed best and worst cases. Read them with the confidence level, study design, possible errors and business risk. Do not rank intervals with different confidence levels as though they were comparable success probabilities. This is an economic screening document, not a substitute for statistical review.
Keep the historical account and the next approval separate
A shoot already paid for belongs in the retrospective project account. But if tomorrow’s choice is to keep running existing assets or stop, the same nonrecoverable production expense is not a new cash outlay tomorrow. Keep two columns: the historical account records the completed period’s full costs and corresponding incremental contribution; the forward approval records avoidable costs from now on, expected additional contribution and uncertainty.
This distinction must work both ways. A new shoot, usage extension, edit, study or media commitment caused by the proposed choice belongs in that choice’s cost. Calling old costs sunk must not make a new supplier invoice disappear. Explain any change in accounting scope rather than moving the hurdle to get an approval.
There is a second, easily missed distinction: advertising versus no advertising is not new creative versus existing creative. A campaign lift study may support the campaign package. It does not isolate the value of ten additional videos. For that procurement decision, compare continuing with the current asset set against the proposed production plan. The relevant question is the contribution added by the new plan relative to its added cost, not whether the existing campaign generated any lift at all.
Specify what the next supplier is being asked to resolve
When uncertainty is the main obstacle, a research brief should identify the uncertainty to reduce, why the answer could alter the choice, the spending cap and the stopping rule. Qualified analysts should define the follow-up design in advance; repeatedly checking results and stopping whenever a preferred label appears is not a sound approval process.
When a specific creative hypothesis is actionable, start with a comparable edit using existing footage if possible. For example, change the order of a product demonstration while documenting the main variable, fixed conditions and version mapping. Commission pickup shots only when essential actions or angles are missing. A clearer edit is a hypothesis to test, not a promised order increase.
When even the more favourable part of the interval struggles to cover matched costs, discuss pricing, product mix, channel strategy or production method. A more expensive film has not automatically solved the problem. When the interval straddles the economic hurdle, holding the expansion decision is a legitimate recorded outcome.
Ask for six fields on the approval page: the two choices being compared; the experimental comparison and scope; the order or contribution interval; the matching break-even requirement; the additional purchase; and the condition for the next review. Teams coordinating Chinese suppliers with overseas media partners can use the same page across both handoffs. Our video marketing articles address related planning questions; this review belongs specifically between a lift result and a new commitment.
For ONCE, the durable service asset is the record of the decision: assumptions, costs, creative versions and subsequent outcomes that a client can revisit. A larger content delivery is not itself evidence of growth. Each new production commitment needs a question of its own.
Source and method note: We also read DoubleVerify’s August 10, 2026 article, Hot Take: In Incrementality Testing, Statistical Significance Doesn’t Matter. (Hear Us Out.), on November 8, 2026 as a peer perspective. We do not adopt its example numbers or treat its statistical wording as a standard. The break-even review, two-column account and procurement recommendations are ONCE’s proposed method. No client accounts were inspected and no experiment was run. The cover is an original conceptual image.
