PLC Dashboard
How every product is performing — by region, by store type, and what to range where.
Go straight to
What this is for
- One place to decide what to buy, for which stores.
- Built on what actually sold, store by store — not on opinion.
- Every product carries its photograph, so a buyer sees the garment, not just a code.
Where the numbers come from
- Store PLC file — every product in every store: units sold, stock left, the store's own rating.
- PO file — what is on order.
- GRC file — what has already arrived.
- Store tagging — which region and which store group each store belongs to.
- The date on the header is the date of the data, not the day the page was built.
The five ideas everything is built on
- Sell-through — of what we gave the stores, how much sold. Sold ÷ (sold + stock).
- Confidence — the same 80% means less from 2 stores than from 200. We use a statistical floor: we are 95% sure the true figure is at least this.
- Rating by store count — a product's rating is the one the most stores give it. Not the one with the most units, so a few big stores cannot speak for everybody.
- Depth — how many units we actually handled: sold + closing stock. Used to put the big commitments first.
- Here vs elsewhere — every product is scored twice: how it did in this market, and how the same product did everywhere else. The comparison is what makes a number regional.
The four range decisions
- Range it — works here, does not work in half its other markets, and is carried in enough stores here to trust. The regional argument. Always the smallest group.
- Trial — same signal, too few stores to commit. Test it, do not bet on it.
- Core — works here and nationally. Buy it centrally, stock it here.
- Drop here — works nationally, fails here. Stop shipping it to this market.
The fit index
- Answers one question: is this department special to this market, or does it sell the same everywhere?
- 1.00 = no more regional than the market overall.
- 2.00+ = twice as regional. Build a range of its own.
- Under 0.70 = no regional case. Take the national range.
- Worked example — girls' pyjamas in J&K: 17 of 46 options work there but not elsewhere (37%), against 12.6% for the market as a whole. 37 ÷ 12.6 = 2.93.
How evidence is separated from a hint
- A product is only marked proven when it is carried in at least 60% of that market's stores.
- And when the statistical floor on its sell-through there is at least 35%.
- Everything else is shown as Trial — visible and labelled, never hidden. A thin result is a lead worth testing, not noise.
How the numbers are kept honest
- Before any refresh, the files are checked: columns, row counts, dates, and whether every trading store is tagged.
- If a check fails, nothing is published and the reason is reported. A late dashboard is better than a confident wrong one.
- The data date is locked to the file's contents, so copying a file cannot make old data look new.
- Refreshes happen in working hours only, so numbers never move under someone mid-meeting.
What it does not do
- It does not set unit quantities. It says what to range and why; depth per option needs the OTB budget and store grading.
- It does not know about price changes or markdowns planned after the data date.
- Around 44% of products have no photograph on file. They are still counted — never dropped from a decision just because a picture is missing.
- Stores not in the tagging file fall outside every regional and store-group view. The refresh reports how many.