How to count it
- take paid orders, not booked ones
- the same period for the numerator and the denominator
- the month of an order is set by the date of first payment
- calculate the median next to the mean
We once saw a median deal size up 57% and nearly put it in a case study. It turned out we were counting by record creation date rather than payment date. After the recalculation the median was flat. How you count changes the conclusion completely here.
Three ways to lift it
- A product mix. Smaller products stand next to the flagship: extended access, a single module, ongoing support. The rep has something to offer whoever cannot stretch to the main thing.
- Bundles. The product plus what goes with it, as one decision. In furniture that is covers and care products, in training a workshop next to the course.
- Work with people who have already bought. The cheapest route: the advertising for these people is paid for already.
What goes wrong
- Upsell before need. The rep offers something dearer before understanding the job, and loses both.
- No list of who to offer it to. With no segments in the base, upselling happens at random.
- Comp that cannot see the difference. If the percentage is identical, the rep sells whatever sells more easily.
How to build an upsell list
- Split every order by what exactly was bought.
- For each group, find what people most often buy on top of it.
- Calculate the window: how many months usually pass before a second purchase.
- Pull a list of the people whose window falls in the next few weeks.
- Write a conversation script for each group separately.