Changes with filters
Both years at a glance
Start here
Four things worth knowing from this check
What to do next
Three actions to test, each with its own follow-up group
The at-risk list, split by what each customer needs
Changes with filters
Revenue held steady while new buyers made up for lost ones
Net revenue each month, after cancellations, split by who paid. Peak season (September to November) is shaded.
Returning customersNew customers (first month)Guest checkout (no account)
December 2009 is the first month in the data, so every customer counts as new that month. In that month, cancellations from new customers outweighed their sales by £3,578. That small negative amount is not drawn, but it is included in the month’s net total. Hover a month for exact figures.
Whole business
What moved revenue from year 1 to year 2
Each customer is counted once: lost (bought in year 1 only), returning (both years), new (year 2 only). Guests have no account, so they're shown as their own line. A few accounts only ever cancelled, and they get their own small line so the movements add up exactly.
Whole business
Every customer type spent slightly less in year 2
Net revenue by customer type. Type is set by each customer's typical order size.
Small: typical order under £150. Mid: £150 to £400. Trade: over £400. Growth came only from new customers and guests.
Whole business
Customers won in the autumn rush come back less often
Share of new customers who placed a second order within 180 days. Average across all months: .
Return rate by the month a customer first bought
Below averageAbove averageAverage
First orders from March 2010 to early June 2011, so every customer had a full 180 days to come back. Earlier customers may have bought before the data starts, so they're left out.
It holds inside every customer type
First bought Dec to AugFirst bought Sep to Nov
Bigger first orders go with higher return rates
180-day return rate by the value of the first order.
Some first products go with higher return rates than others
Return rate for customers whose first order included the product, with the range it could reasonably sit in (95%). Products bought first by at least 60 customers.
Whole business
Cohorts: share of each month's new customers who bought again
Rows are the month a customer first bought. Columns are months since. Darker means more of that group came back.
Whole business, as at 10 Dec 2011
The simple rule ranks customers as well as the 12-signal model
Customers sorted into ten equal groups by predicted risk. The line shows how many really stopped buying in the next 180 days.
Gradient boosting, 12 signalsSimple rule: days since last order and order count
The lost regular
Customer 13093 ordered 55 times, usually every 9 days. Then nothing for 275 days. A monthly sales total would never show this.
This is what the at-risk list is for: spotting a regular who has gone quiet while there is still time to ask why.
At-risk customers, biggest spenders first
Changes with filters
Top 60 products, by net revenue
Whole business
Saturday sales are very low in the records
Sales by day of the week, both years, before cancellations
Saturday took £9,803 in two years. Sunday took £1.79m. The order records can't say why. Website visits and Saturday enquiries would show whether customers try to order and give up.
Orders by hour the order was placed
Orders peak at midday and almost stop after 6pm. Almost all trade happens in office hours, which fits a shop that sells mostly to other businesses.
How we did this
Every step, so you can check it
The same code runs on your data. Every number here comes from the scripts listed in the report folder, and they can be rerun from the raw file.
Cleaning the data
Definitions we used
- Net revenue is sales minus cancellations. A large order that was later cancelled counts as nothing.
- Year 1 is December 2009 to November 2010. Year 2 is December 2010 to November 2011. The partial December 2011 is left out.
- Returning means the customer had bought in an earlier month. Guest means no customer number on the order.
- Customer type comes from each customer's typical (median) order size.
- Came back means a second order within 180 days of the first.
- Ranges on rates are 95% Wilson intervals. A difference counts only when the ranges don't overlap.
Testing the model