See also:
Our metrics for predicted purchasing behaviours are powered by statistical models that learn from your order data.
As well as looking at each individual customer's order behaviour, the models compare those behaviours against overall patterns across your entire customer base.
We use up to three years of transaction history to train our models.
If you have less than three years' data, we can still train once there are at least 180 days of transactions in your account.
The models analyse recency, frequency, and monetary value (RFM) data across all customers who have placed orders.
The models consider data points such as:
- Purchase history - number of orders, timing between purchases
- Spending behaviour - CLV, average order value
- Recency - how long since the last order
By combining individual behaviour with overall customer patterns, Ometria estimates:
- The likelihood a customer will order again in the next 12 months
- How much they’re likely to spend
- Their predicted average order value
Accounts set up by brand and region
Some accounts treat each brand and region separately. Note: if you're not sure whether yours does, ask your CSM.
In these accounts, Ometria trains a separate model for each brand and region, and makes its predictions inside each one.
A contact's predictions for a brand and region are based only on their orders there, compared with the patterns of other customers who ordered there. A contact who has ordered with more than one brand or region has a separate set of predictions for each.
Ometria measures the transaction history described above for each brand and region on its own, rather than across your whole account. So Ometria can have enough data to predict for one brand and region but not for another.
One-time buyers
Customers who have only placed one order will still get predictive labels - but as these are based on your account average they might be less effective.
If your account is set up by brand and region, a contact can be a one-time buyer with one brand even though they have ordered several times across your account as a whole. Their predictions for that brand and region are based on the average there, so they might be less effective.
Accuracy
Ometria's models are trained and continuously validated against real data.
That means we regularly check predictions against what customers actually do, and update the models to keep them accurate.
This approach is widely used in retail to forecast customer lifetime video (CLV) and future purchasing behaviour.
Comments
0 comments
Article is closed for comments.