Pop-up Ads: Testing Design Hypotheses

Our hypotheses favored reviews in both categories. Yet specifications outranked reviews in the reported pairwise tests for both electronics and apparel. In apparel, 54.95% ranked images first, compared with 11.72% for reviews.

How we arrived here

Within-category Friedman tests, then Wilcoxon signed-rank comparisons with Bonferroni correction.

Evidence

Specs > reviews: adjusted p = 0.00911 in each category (Tables 3–4).

Interpretation

For these mock-ups, social proof was not a reliable default for the primary information block. The evidence supports testing information hierarchy before prioritizing review content.

The paired comparison found a stronger preference for specifications in electronics than apparel. Price and review rankings did not show a detectable cross-category difference in the reported tests.

How we arrived here

Two-tailed paired Wilcoxon signed-rank tests for each information emphasis across product categories.

Evidence

Specifications: z = −3.41, p < 0.001, 189 non-zero pairs. Price: p = 0.321; reviews: p = 0.168 (Table 5).

Interpretation

A design hypothesis is to adjust the prominence of specifications by category. These results do not establish that every information block needs a different layout, or that non-significant differences imply equivalence.

Less frequent shoppers ranked electronics specifications more favorably in the two-group analysis (2.12 vs. 2.46; lower is better). The original six-level frequency analysis did not reach significance.

How we arrived here

Compared a Mann–Whitney U two-group result with the Kruskal–Wallis result across the original six frequency levels.

Evidence

Two groups: n = 52 / 221, p = 0.0342; six levels: H = 9.64, p = 0.086 (Tables 31–32). The subgroup p-values are reported without a multiple-testing adjustment.

Interpretation

Treat shopping frequency as an exploratory follow-up question. This grouping-sensitive association is not sufficient evidence for a fixed personalization rule.