A respondent watches an explanatory product ad, then a shorter demonstration. They find both easy to understand. Before cutting the explanation from your launch film, ask whether the second version benefited from information supplied by the first.
For international teams commissioning production in China, the research brief should specify whether the decision concerns an ad encountered alone or reactions after several alternatives. Both are useful tasks. They should not be treated as interchangeable ways to obtain the same answer.
Define the viewing context before collecting scores
SurveyMonkey’s monadic and sequential monadic design guide, dated September 23, 2026, distinguishes one concept per respondent from several concepts evaluated in sequence. It identifies order bias as a risk of sequential designs and recommends randomized order to limit, rather than completely remove, that influence.
Ipsos’s Understanding the Order Effects in Sequential Monadic Product Tests, dated May 15, 2013, explains how evaluation of a second product can occur in the context created by the first. Its subject is product testing; ONCE is applying the contextual question to video planning, not transferring an effect size to advertising.
These are established methods, not a newly completed industry transition. Faster AI-assisted production increases the number of candidate videos a team can prepare. It does not decide how respondents should encounter them.
ONCE’s viewing-task brief separates two decisions
A first-exposure task asks whether a video communicates on its own. Have the research lead design appropriate groups in which respondents encounter only the assigned version before answering. Recruitment, allocation, sample requirements and analysis still need an actual research plan; two convenient groups of colleagues do not automatically support wider inference.
A sequential evaluation task collects reactions to multiple options. The researcher specifies an appropriate randomized or rotated order, with the same questions answered after each video. Keep the actual order and completion record. A final preference question is a separate comparison; it should not overwrite the earlier evaluations.
Specify the intended viewers, versions per person, presentation order, replay rules, question timing and permitted interpretation of each answer. If a study needs both initial reactions and later comparisons, distinguish them in the design. Do not give one set of responses two different labels after the research is finished.
Clearing a form does not erase exposure. Changing order can address parts of the design, but does not guarantee a return to independent first viewing. The analyst must interpret the responses under the method actually used.
Commission research-ready versions
Consider a hypothetical home-care brand. Version A explains the job; B starts with the product in action. Showing A first to everyone does not, by itself, answer whether B communicates adequately to someone who has seen neither ad.
Sequential feedback may still help identify reactions to pacing or expression. If the purchasing decision is whether B can serve as a standalone launch asset, ask the research lead whether fresh respondents or another suitable design are needed. Neither a compulsory second round nor a simple order reversal is a universal fix.
The production handover should identify anonymous version codes, exact playback files, technical specifications, changes between versions and submission dates. When the intended variable is the opening structure, maintain the other agreed elements. If one version receives extra explanatory captions or a different mix, do not attribute every difference to the opening.
Whole-concept comparisons can legitimately contain multiple changes, provided the brief acknowledges their limits. Deliver separate playback files so the research design controls order. A single A-then-B compilation should not be the only asset available.
Before fielding, test playback, replay, questions and routing using a test account. That verifies implementation, not audience representativeness or advertising effectiveness.
Compare the cost of the research question
Budget for recruitment and screening, participant time and incentives, stimulus preparation, questionnaire and order configuration, and data review and analysis. Multiple videos per person may reduce recruitment requirements while increasing each person’s viewing and response burden. Use real quotes and completion assumptions; there is no universal saving percentage.
Responses to several versions from the same person remain related responses. Extra spreadsheet rows are not the same as extra independent participants. Ask how the proposed analysis handles the design, and whether competing quotes answer the same business question.
The research partner owns method and interpretation, the production partner supplies accurate stimuli and change records, and the brand defines the decision and budget. Survey preference should not be relabeled as observed purchase behavior. Where understanding on first contact remains unresolved, carry that limitation into the next production decision.
Explore related planning in ONCE’s video production articles. For ongoing content work, the reusable record connects a version to its viewing conditions, feedback and subsequent edits. Its commercial value requires actual project evidence.
Begin the next brief with a concrete question: what has each respondent already seen before answering?
Scope: Official materials checked November 8, 2026. The brand example is hypothetical. The viewing-task brief is ONCE’s recommendation; no customer study or ad experiment was conducted. The original AI cover is conceptual.
