Does agent matching work if my collectors are all similar?

The short answer

Agent matching (predictive pairing) changes which collector works each account. It can only add something if the collectors on your floor get different results on the same kinds of account. If they do, sending each account to the collector who does better with accounts like it is where the gain comes from. If they do not, there is little for matching to use.

So the honest answer to "my floor is very uniform" is: it depends on whether the floor is uniform in its results, or only in how it was built. On this page, "similar" means similar in results, approach, experience and skill.

Alike on paper is not alike in results

Floors are often built to be alike. Collectors are hired in classes, trained on the same script and the same rules, and coached to the same scorecard. That makes them look interchangeable, and many routing set-ups treat them that way.

Research outside collections suggests looking at results before deciding. A 2012 study by Mehrotra, Ross, Ryder and Zhou in Manufacturing & Service Operations Management, working from data on a medium-sized financial services firm’s customer service call centres, opens with the observation that agents trained to handle all calls perform very differently on the same type of call, measured by handling time and by whether the call was resolved. That was an inbound service floor, not a collection floor, but the point carries: the same training does not produce the same results.

In our view, the differences that matter sit in how each person works the call: how they open, how they listen, how they handle a stall, when they ask for the payment, and what they do with a consumer who is angry or embarrassed. Two collectors from the same hiring class can do these things differently, and consumers respond to the person asking. A floor-wide leaderboard averages those differences away, which is why ranking collectors by an overall average hides them. Why collectors convert at different rates sets out the idea.

When results really are close

Some floors are close in results as well as in background. In our view the likelier reasons are a book made of one narrow kind of account, where no collector has much room to do better or worse than the rest; a call so tightly scripted that the collector’s own approach rarely shows; or conversion that is already high. BestPair’s fit rules exclude floors where 60% or more of right-party contacts already end in a payment or an arrangement.

If your collectors really do convert about the same on every kind of account, predictive pairing has little to use, and live pairing would add little. That is a real outcome and a useful one to learn before committing to anything larger. It is better found from history than from months of live testing.

How to check your own floor before buying anything

You can get a first view from your own reports. In our view these steps keep the check fair:

1. Use enough history. A single month is mostly noise. Use as much call and payment history as you keep, up to 24 months if you have it.
2. Compare collectors on comparable accounts. Look within the same campaign, in the same weeks, on the kinds of account your floor already reports on. If accounts were not handed out at random, a gap can come from the accounts a collector was given rather than the collector. Fair collector scorecards covers this.
3. Score on the right measure. Use dollars collected per account contacted, or the share of right-party contacts that end in a payment or an arrangement, rather than raw totals that follow volume.
4. Separate skill from luck. A collector who leads on a kind of account month after month is more likely showing skill. Leads that come and go are more likely luck.
5. Count your collectors. BestPair works with floors of 25 or more agents because, below that, too few agents work each kind of account to tell skill from luck.

If the order of your collectors stays the same on every kind of account and the gaps are small and unstable, your floor is uniform in the sense that matters. If the order changes from one kind of account to the next, and holds over time, there is something for matching to use.

What the Floor Assessment shows

The Floor Assessment (Stage 1) does this check on your own history, with nothing changed on the floor. It estimates from your own history what pairing would have been worth against your current routing, and if it cannot show a gain, the report says so in plain words.

Its agent strengths page shows where each collector does better or worse than the rest of the floor, with actions you can take now, with or without BestPair: who should coach whom, who to hire and where to start new collectors. On a floor that turns out to be close, that page is the evidence. A sample report shows the format.

The Floor Assessment is a $15,000 fixed fee, paid before any data moves, and not refundable. You keep the report whatever it finds. Stage 2 starts on the first of the month after the readout unless you opt out in writing within ten business days, and if you opt out, nothing further is owed. What you pay if the report shows no gain has the detail.

In either stage, people make every call. BestPair never contacts consumers and does not change offers, scripts, contact rules or call frequency; FDCPA, Regulation F and TCPA controls stay with the agency.

Sources

• Routing to Manage Resolution and Waiting Time in Call Centers with Heterogeneous Servers: Mehrotra, Ross, Ryder and Zhou, Manufacturing & Service Operations Management, 2012; author copy of the paper, on agents who handle the same call types with very different results at a financial services firm’s customer service call centres.