The CGPI, week by week

These are week-over-week moves in the published index: the change from the prior weekly release, not a year-over-year rate. The middle number is each sub-index's level. Category sub-indices share a fixed base (the week of 13 July 2026 = 100), so a level reads as the change since that week and category levels are comparable with each other; the ranking here is still by the week's move. All CGPI series →

Each cell is that category's week-over-week percentage change in the published national sub-index; shading runs to the largest absolute move in this grid (±5.6%), so the scale is relative to the weeks shown and not to any fixed range. Rises are orange, falls are blue, and a dashed cell is a week with no published change (a series' first week has no prior week to compare against, and a week published after a gap carries no week-over-week figure). Category sub-index levels share a fixed base (the week of 13 July 2026 = 100), so levels are comparable across categories and each reads as the change since that week; the grid still shades weekly moves because that is the question it answers. There is no year-over-year figure here because the published record starts in July 2026.

The bottom row is the top rows. The headline CGPI is not a separate measurement: it is these nine sub-indices combined, so it can be rebuilt from them exactly. Take each category level, raise it to the power of the share of the basket its matched items carry (the percentage printed beside its name), and multiply: CGPI = ∏_c (I_c/100)^ω_c × 100. Every series here is the same estimator, a matched-model Jevons index against the week of 13 July 2026 formed within one store, which is what makes the arithmetic close to the cent rather than approximately. The percentages above are that week's matched-weight shares, the exact ω the identity uses; they are published per week as matched_weight on every category row. All CGPI series →