Measuring whether exposure frequency improves recall

To test whether repeated game ad exposure improves recall, assign comparable eligible units to lower and higher exposure policies, then ask the same recall question at a planned time.

Higher recall among people with more recorded impressions is an association; it does not show that the extra impressions caused it.

Treat this as a frequency-policy test: compare groups by the policy they were assigned, not by impressions recorded after delivery. The comparison estimates the result of the policies under the tested delivery conditions; it does not isolate the effect of each additional impression.

Define the dose

State the period and whether frequency counts impressions per person, account or device. If you report an average frequency, name the reach unit it is based on. Ask for the distribution of observed impressions too: an average can hide people or devices with very different exposure.

Check whether repeated contact across games and devices can be identified. If the system identifies only devices, describe device frequency. Decide whether the policy controls served impressions or another measurable exposure.

Assign policies before delivery

Set lower and higher exposure policies before launch and include a no-ad control if feasible for the decision. Randomly assign eligible units where the buying path permits. Keep the creative, games, Australian eligibility, recall wording and survey timing consistent. Confirm which parts of the proposed supply path can enforce the policies.

Report both assigned policy and actual observed frequency. A unit assigned to the higher policy may have few ad opportunities, while exposure on another device may be invisible. Preserve the assigned-group comparison. Regrouping respondents after delivery by the impressions recorded for them can mix the effect of repetition with their activity and availability.

If random assignment is unavailable, treat a comparison of observed frequency groups as an association, not a causal result. Activity and availability may differ between those groups.

Read the recall result

Choose aided or unaided recall in advance, with identical wording and response options across groups. Record survey timing, group sizes, response rates, coverage and uncertainty where available. Uneven survey response or missing exposure data can weaken the comparison.

For a numeric recall outcome, compare the assigned groups with analysis of variance. Present descriptive statistics and confidence intervals alongside the comparison so the size of the result and its uncertainty are visible.

Interpret the finding as recall under the tested policies and delivery conditions. If the observed exposure distributions overlap heavily, the test has learned little about extra repetition even if the policies differed. If recall differs, consider uncertainty, additional media cost and any audience or placement changes before altering a cap. The result does not establish a universal ideal frequency.