Dunwell Insights · Returns & Margin Lab · Synthetic demonstration. Not a client result.
Where the money went, last 12 months
AddsTakes awayTotal
Summary for the owner · Synthetic data
What we found
What we'd do
Open the live dashboardSet your own fulfilment, handling and stock recovery costs and see each category's cost of returns and break-even cut change. Look up any of the top 300 products and download the full review list.1 · Data checks · Synthetic data
Order status by how long ago the order was placed
Share of items in each status. In a real shop, recent orders would still be processing or in transit and old ones would be settled.
What we checked
Who counts
2 · Money · Simulated, assumption-based
Last 12 months under three cost cases
How much each assumption matters
Cost of returns over the last 12 months as one assumption moves and the others stay at the central case
Formula check
3 · Return rates · Synthetic data
Return rate by category, every year
Delivered items past the 3-day return window. Lines are 95% ranges. The dashed line is the whole shop.
Every grouping we tried
Could the differences between groups be chance? Lower p means less likely. Adjusted for testing ten groupings at once.
"Serial returner" products are what chance gives
Share of products with every item returned, and with none, against what a 28% coin toss would give
4 · Predicting a return when the order is placed · Synthetic data
Every model on the test period
5 · Where returns cost most · Simulated, assumption-based
Cost of returns a year, by category
Last 12 months. Bars use the central costs; the line runs from the low-cost to the high-cost case.
What a $2,000-a-year fix needs to do
Product review list, top 12
Ranked by expected cost of returns a year, central costs. The full list is in the dashboard downloads.
6 · What to test · Synthetic data
Test brief: clearer size guidance
Weeks needed to see a cut in returns
80% power, 5% two-sided. Products are randomised, so the count allows for items of the same product being alike.
7 · How we did this · Synthetic data
The plan was locked before any return rate was compared
The models
Checks we ran
Data dictionary
Decisions log
What this data can't tell us
This is a demonstration of method on made-up data. Every money figure is a scenario built on stated assumptions. It is not a promise of any result, and it is not financial, legal or tax advice.
Every job. Done well.