Does holding back a control group cost us sales?

What happens to the control leads

Before anyone dials, incoming leads are split at random into two groups. One group is worked by the agents BestPair matches to each lead. The other, the control group, is routed the way your floor routes leads today. Both groups run in the same weeks, through the same dialer and CRM, with the same agents, scripts and offers.

So the control leads sell at whatever rate your current routing would have given them. Nobody works them less hard, and nothing is taken from them except the matching. People make every call in both groups. BestPair never contacts leads.

What it costs, and when

The cost of a control group is the extra those leads would have sold had they been matched. Two things limit it.

• It is a cost only if pairing beats your current routing. If it does not, the control leads gave up nothing, and the control group is how you learn that.
• It applies to the control leads alone. Every matched lead gets whatever gain pairing brings.

In return for some of an uncertain gain on part of your leads, you find out whether the gain is real, and how large it is on your own floor.

What you could not tell without it

Sales on an outbound floor move for many reasons in any month: a better lead list, a new offer or price, a campaign starting or ending, the season, a wave of new agents. Any one of them moves sales on its own. If every lead were matched, a good month would credit pairing with all of it and a bad month would blame pairing for all of it.

A control group in the same weeks takes in all of those changes equally. A better list lifts both groups and a slow season pulls both down. The gap left between them is what the routing did. Both groups are scored on sales per contact, by your floor’s own definition of a sale, applied the same way to each.

Why it protects the floor

Any gain BestPair reports is measured against the control group. A month in which the whole floor sold more because of a new offer shows up in both groups, so it is not counted as a gain from pairing. That keeps a better list or a good season from being credited to the routing, and it keeps the decision to carry on, or stop, tied to what the routing did.

The team behind BestPair has applied predictive pairing to outbound floors, collection campaigns among them, since 2011. BestPair does not guarantee a level of lift; the control group is what shows whether there is one. How to set up a fair test.