Does lead matching work when agent turnover is high?

What turnover does to matching

Matching works from each agent’s own history: which kinds of leads they have closed and which they have not. Experienced agents carry most of that history. When one leaves, what the floor knew about their strengths leaves too, and the leads they did best with go to someone the model knows less about.

New agents have the opposite problem. They have too few calls for anyone to tell what they are good at, and early results swing. A new agent with a few lucky sales in their first week looks like a star on paper.

Matching can still work on a floor with high turnover, but it has to be built for it.

How new agents are handled

• They are treated as floor average. Until an agent has their own history, BestPair treats them as an average agent on your floor.
• They get a normal mix of leads, not the leftovers. A spread of lead types lets their own record build across the board, instead of only on the leads someone guessed they would suit.

How the model keeps up

The model is updated about once a month, and weekly in the first couple of weeks after matching starts. Each update learns from the latest results, so agents who have grown into a type of lead get more of it, and leavers drop out.

The data that keeps it current is your floor’s own results, with the current agent list, sent by encrypted SFTP to a folder BestPair provides, never by email. It is pseudonymised, and BestPair does not ask for customers’ names, phone numbers or email addresses.

What a wave of leavers does to results

When several experienced agents leave in the same month, the matched leads lose some of their best-known pairs, and the gain over the control group can dip until the next update rebuilds the picture. It helps to know that is coming. Log leaving dates, so a dip can be traced to a cause instead of argued about.

The comparison itself stays fair. Leads are split at random between matched and control, so turnover hits both groups in the same weeks. What the comparison shows is how much matching adds with the floor you actually have, churn included. How to prove a routing change worked.

Floors where agents barely overlap from one quarter to the next, or where the whole floor is new, have little history to match on. That is better found out at the start, from your own data, than once you are into a pilot.