1. Lighthouse

Level 6 · December's suspiciously good return rate

The request

From Priya Shah, Head of Data

“The monthly dashboard says December 2025 had a 2.92% return rate, against 6.98% in November and 5.96% for the year. Operations want to celebrate. Before they do, can you check the definition?”

Your goal

Pick the definition change that makes December's return rate honest.

The tile shows refunds on orders placed in each month ÷ net sales of that month's booked orders. December 2025: $8,573.75 of refunds on $293,358.44 of net sales = 2.92%. Returns arrive 4 to 40 days after the order, and the data ends on 2025-12-31. Which fix makes the metric honest?

Pick one answer, then check it.

The lesson

A definition has four parts

Every KPI definition answers four questions: what is counted (refunds), divided by what (net sales of booked orders), over which window (orders placed in the month), and when is it final. Most dashboards answer the first three and forget the fourth.

Metrics with a delay

Some events happen long after the thing they belong to. A return arrives 4 to 40 days after its order. When the tile is calculated on December 31st, an order placed on December 20th has had 11 days to come back, and an order placed on the 31st none at all. The numbers show it:

order month   net sales    refunds     return rate
2025-10       184,241.76   12,123.51       6.58%
2025-11       305,172.51   21,306.68       6.98%
2025-12       293,358.44    8,573.75       2.92%

December is not a great month for returns. It is an unfinished one. The same thing is called right-censoring in statistics: for the newest items, the outcome has not happened yet.

Maturity windows

The fix is a rule in the definition: a month’s return rate is final once the longest return delay has passed after the month ends. Here that is 40 days, so December 2025 becomes final on 2026-02-09, and November 2025 on 2026-01-09. Until then the tile shows the number greyed out with “not final”, or leaves it off the trend line.

If waiting is not acceptable, compare like with like: “returns within 14 days of the order” can be computed for every month older than 14 days, and it is the same measure for every month.

Status filters

The denominator matters too. Net sales count booked orders only: completed and refunded. In December 2025 there were also 94 pending and 36 cancelled orders. Cancelled orders were never paid and pending ones are not settled, so neither can be refunded. Putting them in the denominator would lower the rate again, for no reason.

Common mistakes

  • Celebrating the newest period. Any delayed metric improves at the right edge of the chart.
  • Changing the denominator to fix a timing problem. It moves the number without making it true.
  • No “final” date in the definition. Write it down next to the formula.

Pick what may run on this site. You can change this any time from "Cookie settings" at the bottom of every page.