Once you have many stores, HQ’s easiest mistakes are spreading every line on one sheet, or watching only the total. The first cannot be read. The second buries a bad store in the average. A roll-up tree should make outlier stores surface by themselves.
-
Shenzhen12 stores · margin 29.1%
- Nanshan SZ-NSSales on plan · stock healthy
- Futian SZ-FTMargin -6.1pt · SKU-8841 stockout alert
- Baoan SZ-BATicket size stable
-
Guangzhou8 stores · margin 28.4%
- Tianhe GZ-TH3 days below regional median · free-goods tickets pending check
- Haizhu GZ-HZNormal
Illustration: HQ looks at red nodes first, then drills that store’s documents. The total layer does not show the collections definition, so you do not reopen the finance fight.
Watching outlier stores requires one store code. Write Nanshan four ways and the tree forks. HQ praises a store that does not exist. Why missing master data lies, see Without master data, visualization lies. Roll-up formulas must match the store definition, or the region never ties; see After one definition, weekly meetings argue less. How chain roll-up avoids forked definitions, see How chain roll-up avoids forked definitions.
HQ should watch deviation, not every POS line. Margin outside a threshold, stockout SKUs, and stores that did not open go to an alert list and a supervisor; see Alerts pushed to people, so exceptions rarely sit overnight. A store-manager dashboard shows only that store, so other stores’ numbers do not distract and permissions do not leak. If that store’s documents open, a supervisor does not call for a screenshot; see If a number cannot open a document, what is the dashboard for.
Do not mix franchise and owned stores without saying so. Tax, stored value, and delivery often have two rules. Unclear writing turns roll-up into politics. POS, membership, and inventory each calculating their own, see When every system calculates its own numbers, what breaks. When more systems connect, standardize APIs first; see More systems connecting: standardize APIs or draw charts first.
When DaXi builds a chain roll-up, we draw this tree first: how regions cut, how an outlier is defined, who receives alerts. Phase one can be sales, margin, and stockouts. No platform required. See the Data service page. Visible efficiency is supervisors who stop touring by feel; see Visible efficiency a data capability actually delivers.
Before a visit, a supervisor opens the red node and checks those documents in the store, instead of recopying POS flow. The tree is trimmed by org permission: a district manager does not see another district’s detail. After APIs are reused, adding a region is a slice, not a new extract; see Once APIs are standardized, the next chart is faster.
Keep new stores off the ranking during the observation window
A new store still in observation should not enter the ranking, or it is treated as a bad store. Closed stores retire the code. Do not delete them from the tree and evaporate history.
Roll-up frequency follows management rhythm: supervisors daily, HQ operations weekly. Do not force every layer real-time; see Does it have to be real-time, or is T+1 enough.
HQ watching stores should not use a single ranking as shame. Rankings mix season, trade area, and store age. A “last place” store is not always last at operations. Deviation and stockout lists are actionable. Dashboards nobody opens often have shame and no next action; see Nobody looks at the dashboard. Is it still worth building. Phase one can be the tree and outliers. No platform required; see What dashboards or a data platform cost, and how long they take. Why you need an operations layer you can challenge first, see Why an operations dashboard is not optional.