UCI ONLINE RETAIL · REPORT 02 / 03
Growth arrived
as workload.
Autumn’s increase was mostly more orders, not dramatically larger baskets. A useful operating report separates throughput, timing and uncertainty instead of projecting one average into the future.
Real public retail data, not client data. Descriptive business intelligence, not causal inference. Positive-sale value is not settled revenue or profit. Business hypotheses and external research are labelled separately; no current-market benchmark is implied.
01 monthly orders
The autumn ramp is unmistakable.
All-country order groups. Amber highlights September–November 2011. Hatched December bars are partial source months, not like-for-like monthly comparisons.
Explore the exact figures
| Period | Orders | Coverage |
|---|---|---|
| Dec 10 | 1,560 | Partial source month |
| Jan 11 | 1,093 | Observed period |
| Feb 11 | 1,108 | Observed period |
| Mar 11 | 1,460 | Observed period |
| Apr 11 | 1,252 | Observed period |
| May 11 | 1,686 | Observed period |
| Jun 11 | 1,534 | Observed period |
| Jul 11 | 1,476 | Observed period |
| Aug 11 | 1,363 | Observed period |
| Sep 11 | 1,838 | Observed period |
| Oct 11 | 2,041 | Observed period |
| Nov 11 | 2,772 | Observed period |
| Dec 11 | 819 | Partial source month |
02 recorded value
More orders, plus a smaller basket lift.
Positive-sale value increased 66.2% over the summer-to-autumn comparison. This is recorded quantity × unit price, not settled revenue.
Explore the exact figures
| Period | Positive-sale value, GBP | Coverage |
|---|---|---|
| Dec 10 | £823,746.14 | Partial source month |
| Jan 11 | £691,364.56 | Observed period |
| Feb 11 | £523,631.89 | Observed period |
| Mar 11 | £717,639.36 | Observed period |
| Apr 11 | £537,808.62 | Observed period |
| May 11 | £770,536.02 | Observed period |
| Jun 11 | £761,739.90 | Observed period |
| Jul 11 | £719,221.19 | Observed period |
| Aug 11 | £759,138.38 | Observed period |
| Sep 11 | £1,058,590.17 | Observed period |
| Oct 11 | £1,154,979.30 | Observed period |
| Nov 11 | £1,509,496.33 | Observed period |
| Dec 11 | £638,792.68 | Partial source month |
03 weekly rhythm
Thursday leads. Saturday is empty.
Total observed orders by weekday, not orders per open day. No Saturday orders are recorded; that alone does not prove the retailer was closed.
Explore the exact figures
| Measure | Observed orders |
|---|---|
| Monday | 3,129 |
| Tuesday | 3,561 |
| Wednesday | 3,699 |
| Thursday | 4,252 |
| Friday | 3,149 |
| Saturday | 0 |
| Sunday | 2,212 |
04 clock rhythm
Six clock hours carry most orders.
All-country order timestamps, 10:00–15:59 highlighted. These are invoice creation times, not website visits, staff hours or authorization events; timezone is undocumented.
Explore the exact figures
| Period | Orders · hour of source clock | Coverage |
|---|---|---|
| 06 | 1 | Observed period |
| 07 | 29 | Observed period |
| 08 | 566 | Observed period |
| 09 | 1,486 | Observed period |
| 10 | 2,367 | Observed period |
| 11 | 2,398 | Observed period |
| 12 | 3,223 | Observed period |
| 13 | 2,756 | Observed period |
| 14 | 2,461 | Observed period |
| 15 | 2,348 | Observed period |
| 16 | 1,342 | Observed period |
| 17 | 669 | Observed period |
| 18 | 192 | Observed period |
| 19 | 146 | Observed period |
| 20 | 18 | Observed period |
These weekday and hour charts are separate totals. They cannot reveal a Thursday-at-noon peak without a joint breakdown.
05 model stress test
The narrow model misses. The wider one is not yet validated.
Separate UK arrival-model cohort: 273 training days and 99 later holdout days, fixed UTC calendar-day windows. This is not the all-country BI population above.
Explore the exact figures
| Measure | Share of holdout days inside interval |
|---|---|
| Poisson | 7.1% |
| Negative binomial | 100.0% |
| Nominal target | 90.0% |
Passing every day can still be uninformative.
The Poisson interval covered only 7 of 99 holdout days. The negative-binomial interval covered all 99; such broad coverage is not evidence of calibration. Compare width and coverage, not coverage alone.
Training variance was 16.6 times the mean, and holdout mean count was 1.64 times the training mean. Both dispersion and change over time matter.
No forecast is promoted. These results support testing time-varying count models, not deploying the wider interval as a calibrated capacity forecast.