How accurate is the forecast?
For agency staff, how Fabrik records the forecast's accuracy each month and how to read the accuracy line on a brand's Targets page.
3 min readUpdated 7 October 2026
Every projection Fabrik Analytics makes is checked against what really happened. Once a month closes, the projection made on each day of that month is compared with the month's final figure, and the difference is recorded. Agency staff see a summary on each brand's Targets page, so you know how much weight to give the projection when talking to a client.
Before you start
- Only agency staff see forecast accuracy.
- The brand needs at least one closed month with a recorded projection.
Step by step: find the accuracy line
- Open the brand and choose Trading, then Targets.
- Scroll to How attainment is read at the foot of the page.
- The last line reads, for example: "Last 6 months, the forecast at day 10 was within 14% on average (net revenue)." Net revenue here means net sales plus shipping.
What is recorded
For each brand, each closed month and each day of that month, the record holds:
- the projection made that day for the month's finish;
- the month's actual finish;
- the difference between them.
It is kept for net sales plus shipping, net sales and total ad spend.
The Targets page summarises the last six closed months at day 10 of the month: the average distance, in percent, between the day 10 projection and the final figure.
A worked example
On 10 October Fellwick's projection was 97,000. October finished at 105,000. The projection was 8,000 short, which is about 7.6% of the final figure. Averaged with the five months before, the line might read "within 9% on average".
Reading it
- Early in the month the projection is less certain. It improves as more days arrive, so a day 10 figure is a fair test of an early call.
- A lower percentage is better. A brand with steady sales will usually sit lower than one driven by a few big launches.
- "Replayed with today's method" means some months before October 2026 were worked out again with the current method, using only what was known before each month.
- The method itself is the same for every brand. See How the forecast works.
Troubleshooting
- No accuracy line shows. Nothing is recorded yet for this brand, or you are signed in as a client user.
- Accuracy looks poor for one month. A one-off event (a sale moved, a stock-out, a launch) will show up as a large miss. Check Intelligence's timeline for what changed that month.