Is the sales dip seasonal?
We compared this year's late-May sales against the same calendar weeks in 2025 and 2024, using the Revenue Tracker sheet (actual order revenue, ground truth). If the dip were a normal seasonal lull, prior years should show a similar shape. They do not. Last year, late May was the strongest part of the season and still climbing. This year it turned down.
Seasonality does not explain the dip. The clean comparison is weeks 20–22: in 2025 the weekly average rose through late May ($8,533 → $9,100 → $9,571/day). In 2026 it fell over the exact same weeks ($8,929 → $7,381 → $6,612/day). 2026 had been running ahead of 2025 into mid-May, then diverged downward at the week of May 18.
Weekly average sales per day: 3-year overlay
Through the week of May 11 (wk 20), 2026 was at or above 2025. Then the two lines split: 2025 accelerated into its seasonal peak while 2026 turned down. The week of Jun 1 is shown for continuity, but it is a partial one-day point.
The ceiling weakened: strong days vs weak days
The dip is not the floor getting worse. Weak days (weekends) held steady all year. What disappeared is the strong-day ceiling. Last year's peak days kept climbing into late May; this year they fell sharply after the week of May 11.
2026 peak: $12,439 (wk of May 11) → $7,883 (wk of May 25), a 37% drop. Meanwhile the 2026 floor sat around $4,300–5,000 for the same period. In 2025 the peak ran the other way: $11,691 → $12,923 → $13,394 into late May. This argues against a full-site outage, but it does not rule out a localized desktop, paid-search, category, or AOV issue.
What this suggests
- Seasonal lull: unlikely. Last year these weeks were the seasonal high and rising.
- Sitewide checkout break: unlikely from this view. A localized page, device, channel, category, or AOV problem is still possible.
- A/B tests: not proven either way from this page. The pause since May 29 without recovery weakens the case, but exposure-level variant data is needed to rule tests out.
- Something new starting mid-May (wk of May 18): consistent with the timing. Demand, competition, paid-search mix, category mix, or AOV mix are hypotheses to test next.
Caveats
- 2024 data starts at week 17 and predates the current designer, so treat it as a rough floor, not a clean baseline. The 2025-vs-2026 comparison is the reliable one (same site era).
- wk of Jun 1 is a single day so far and is not used for the main verdict; its peak, floor, and average are the same number.
- Figures are actual order revenue from the Revenue Tracker, not GA4, so they are not subject to analytics undercount.
Next layer: a Category × Channel matrix over time (from GA4 BigQuery) to localize where the missing weekday volume went. This page answers the seasonality question; the matrix is needed before making causal claims.