Why do two sales reports show different numbers?
Two sales reports built from the same data usually differ because they made different choices, not because one of them is wrong. Write down four things for each (which sites count, whether returns are subtracted, which days are in the period, and the level it is added up at) and you can walk from one number to the other, line by line.
| Brand report | Territory rollup | |
|---|---|---|
| Label on the page | Units, last 13 weeks | Units, last 13 weeks |
| Sites | All ship-tos, mapped or not | Mapped accounts only |
| Returns | Subtracted | Not subtracted |
| Days | 12 closed weeks, Jul 18 to Oct 9 (the 13th is still open) | The 91 days to the last day of data, Jul 14 to Oct 12 |
| Added up at | One company total | 60 territory views, summed |
| Units | 17,248 | 19,214 |
The brand report is the kind a brand team sends out each week. The territory rollup is what you get when someone adds up every territory's dashboard in a spreadsheet. Both are right for what they measure.
They are 1,966 units apart, 11% of the brand report. A gap that size is easy to blame on bad data, but nothing here is wrong. The worked example below finds five separate differences between the two reports. Four push the rollup up and one pulls it down, so together they move 4,010 units, about twice the gap. The size of a gap tells you little about how many differences sit behind it.
Every number in this guide comes from a synthetic dataset. It models a small biotech with one injectable brand, 60 territories, 800 accounts and three years of distributor sales data. There are no real patients, prescribers or accounts in it. Both reports are run on Monday, October 12, 2026, with sales data through that day (one distributor sends a weekly file, so its data stops at Friday, October 9). Weeks run Saturday to Friday and are named by the Friday they end, so the week ending October 9 has just closed and the week ending October 16 is open, with three days of data in it.
The reconciliation, line by line
| Line | What changes | Units |
|---|---|---|
| Brand report | Net units, all ship-tos, Jul 18 to Oct 9, company total | 17,248 |
| 1. Unmapped ship-tos | Leave out sites not yet mapped to an account | -1,022 |
| 2. Returns | Stop subtracting returns at mapped accounts | +198 |
| 3. Earlier start | Add July 14 to 17 | +1,027 |
| 4. Open week | Add October 10 to 12 | +59 |
| 5. Accounts in two territories | Count each account in every territory it is aligned to | +1,704 |
| Territory rollup | Shipped units, mapped accounts, Jul 14 to Oct 12, sum of territory views | 19,214 |
Each line is measured after the ones above it. After lines 1 and 2 the count is shipped units at mapped accounts, so lines 3 to 5 are in those terms. A different order changes the size of some lines, but not the two end points. This order works well for your own reports too: what is counted, how it is counted, over which days, and at what level.
Line 1: unmapped ship-tos
The brand report keeps every line the distributors sent. A territory is a list of accounts, so the rollup can only show sales at ship-tos that someone has mapped to an account. In the brand report's 12 weeks, 34 of the 687 ship-tos that bought were not mapped yet. Their 1,022 net units are 5.9% of the brand report. They are real sales and belong in the company total, but no territory gets them until they are mapped.
Line 2: returns
Returns arrive as their own lines. The brand report subtracts them. The rollup was built from sales lines only, so returns never reduce it. In these 12 weeks returns were 1.1% of shipped units, and 198 units came back from mapped accounts. Shipped units can never be lower than net units, so a report that skips returns always reads high. When a report just says "units", ask whether it means net or shipped.
Lines 3 and 4: which 13 weeks
Both reports say "last 13 weeks", but they count different days. The brand report counts back 13 weeks, starting with the week that holds the last day of data, and keeps only the weeks that have closed. The data runs through Monday, October 12, the third day of the week ending October 16. That week is still open, so the brand report holds 12 full weeks: July 18 to October 9. When the data reaches Friday, October 16, that week joins and the report holds 13 again.
The rollup is a trailing window: the 91 days that end on the last day of data. It starts four days earlier, on Tuesday, July 14, which adds 1,027 units from the last four days of the week ending July 17. It also ends three days later, which adds the open week so far: 59 units. That line is small only because of the calendar. The three days are a weekend and a Monday that was a federal holiday (44 units, against 279 the Monday before), and the distributor with a weekly file hasn't sent those days yet. Run on Thursday, the open week would already hold four of its five weekdays, and still be missing that distributor.
An open week causes trouble beyond totals: on a trend chart, or next to last week, it reads as a drop. Which time period should each sales report use? works through that, and which period suits which report.
Line 5: accounts in two territories
Of the 800 accounts, 88 are aligned to two territories: a primary territory that gets the sales credit, and a second territory that also covers the account. In this example each territory's dashboard shows every account it covers, which is what each rep needs to see. Add the 60 dashboards together, though, and those accounts count twice. The 69 of them with shipments in the window shipped 1,704 units, so the sum of territory views comes out 1,704 above the company total for mapped accounts. It is the largest of the five lines, from about one account in nine. To total across territories, add up credited sales only, one territory per account.
Four things to state on every number
Put these next to any sales figure that someone might hold up against another report:
- Sites included: all ship-tos, or mapped accounts only. If unmapped sites are in, say how many units they hold.
- Returns: net of returns, or shipped units before returns.
- Period: the exact dates, and whether the last week or month is closed or still open.
- Level: a company total that counts each sale once, or a sum of territory or rep views that can count a shared account twice.
Fully labeled, the brand report's figure reads: net units, all ship-tos including 1,022 unmapped units, 12 closed weeks from July 18 to October 9, 2026 (the 13th, from October 10, still open), company total: 17,248. Anyone holding a different number can now find the line where the two part ways.
James Waterman builds commercial analytics for small biotech teams. hello@aimlitics.com
Ardlow uses the brand report's 13-week rule for its rising and declining accounts, keeps unmapped units in every total, credits each account's sales to one territory, and marks the open week as partial. Its open demo runs on the same kind of synthetic data.