Why generic call center KPIs mislead a transfer floor
Most lists of call center KPIs are written for inbound support: average handle time, first call resolution, customer satisfaction. A transfer floor in Lahore or Mohali running a US Medicare or final expense campaign is paid on something else, billable transfers, and every number it tracks should lead back to that.
This post covers the eight that matter, how to calculate each, and what to watch for. Definitions vary between dialers and between buyers, so the first job is to write down your floor's definitions and use them the same way every night. A number that changes meaning between shift managers is worse than no number at all.
Two definitions cause most of the confusion. The first is what counts as a contact: some dialers count any answered call, some count only calls an agent spoke on. The second is what counts as a transfer: some floors count every conference started, others only transfers the closer answered. Decide both, write them at the top of the report, and use the same definitions when you compare your numbers with the buyer's. When a new manager joins, those definitions are the first thing to hand over, before the passwords.
The funnel: connect rate, contact rate and transfer rate
These three describe how dials turn into transfers. The numbers below are illustrative only. Say the floor makes 1,000 dials.
- Connect rate = answered calls divided by dials. Answered includes humans and machines. If 300 of 1,000 dials are answered, connect rate is 30 percent. Watch for a drop on one list (bad data) or on all lists at once (caller IDs flagged as spam, or carrier trouble).
- Contact rate = live human contacts divided by dials. Some floors divide by connects instead; pick one and stick to it. If 120 of the 300 answers are people, contact rate is 12 percent of dials. Watch for the gap between connect and contact rate suddenly widening, which usually means more machines or a change in AMD.
- Transfer rate = transfers divided by live contacts. If 12 of those 120 contacts are transferred, transfer rate is 10 percent. Watch for drops by hour late in the shift, and for big differences between agents on the same list.
Billable transfers and billable rate
Read the funnel numbers together. A falling transfer rate with a steady contact rate is a floor or script problem. A falling contact rate with a steady transfer rate is a data or dialer problem. Then look past the funnel, because transfers are what your floor sends, and billable transfers are what your buyer pays for after duration, criteria, duplicates and returns.
- Billable transfers = transfers that met the contract terms and were not returned.
- Billable rate = billable transfers divided by transfers. If 12 transfers produce 9 billable, the billable rate is 75 percent.
- Billable transfers per fronting hour = billable transfers divided by logged-in fronting hours. This lets you compare agents, shifts and bots on one scale.
What a falling billable rate is telling you
Watch for a billable rate that drops while the transfer rate rises. That pattern usually means weak callers are being pushed through, often near the end of the shift or near the end of a month with an incentive target. Pull the buyer's return reasons by campaign and by agent. "Short call" and "did not know about transfer" point at the handoff, and how a warm transfer should run is the fix. "Wrong age" and "wrong state" point at qualification, which is a script and coaching problem.
Two supporting numbers help here. Closer answer time is the average seconds from the start of a transfer to a closer picking up; when it climbs, holds get longer and short calls follow. Transfer talk time is the average length of a transferred call on the buyer's side, from their report. A falling talk time with a steady transfer count means the closers are getting calls they do not want, and it usually shows up a few nights before the returns do.
Look at the billable rate by agent as well as by campaign. Two fronters on the same list with the same transfer rate can have very different billable rates, and the gap is usually in how they introduce the caller, not in how many callers they qualify.
Cost per billable transfer
This is the KPI that tells you whether a campaign makes money at all.
- Cost per billable transfer = (seat costs + VoIP + data + tools and bots + share of overhead) divided by billable transfers, for the same period.
- Worked example in neutral units: if a night costs 900 units in total and the floor produces 60 billable transfers, cost per billable transfer is 15 units. Compare it with what the buyer pays per billable transfer, in the same currency.
- Break-even billable rate = cost per transfer sent divided by payout per billable transfer. If you need more of your transfers accepted than you are getting, the campaign loses money however busy the floor looks.
- Margin per billable transfer = payout per billable transfer minus cost per billable transfer.
AMD accuracy
Watch cost allocation, too. VoIP and data are often lumped together across campaigns, which hides a losing campaign behind a good one; split them even roughly. The next biggest cost lever is answering-machine detection, which decides which answered calls reach a fronter. It fails in two directions, and both cost money.
- False positive rate = live humans dispositioned as machines divided by all calls dispositioned as machines. These are real contacts hung up on. Measure it by listening to a sample of machine-dispositioned calls each night. Even a small sample, listened to every night, catches a bad setting within a shift or two.
- False negative rate = machines passed to a fronter divided by all calls passed to fronters. These waste fronter time and inflate the connect rate.
- Detection time = average seconds from answer to the AMD decision. Slow detection means a real person hears silence and hangs up. The trade-off is covered in AMD and cost per transfer.
Abandon rate and QA score
These two are as much about risk as about performance. Any failure on a compliance point, such as a missing disclosure or an ignored DNC request, should fail the call outright regardless of its total score.
- Abandon rate = abandoned live-answered calls divided by live-answered calls, per campaign, over the period your rules specify. The US limit is commonly summarized as 3 percent over 30 days per campaign; confirm the current rules and how they apply to you with counsel. Watch for spikes when agents log off and a predictive mode keeps dialing at the old pace.
- QA score = points passed divided by points possible on your scorecard, averaged per agent and per campaign. Watch for scores that differ by team lead on the same calls; calibrate by having two leads score the same five calls each week. Also watch for scores that are high across the board while returns rise, which means the scorecard is checking the wrong things.
- QA coverage = calls reviewed divided by calls made. The lower the coverage, the more one bad call is hidden or one unlucky call is overweighted.
A nightly KPI sheet for a transfer floor
Put these on one sheet, one row per campaign, and review it at the end of every shift (around 4am PKT or 4:30am IST), then again when the buyer's numbers arrive. Bot seats belong on the same sheet. B3 Voice bots write dispositions back to your dialer in your existing codes, so the rates calculate the same way for bot seats as for human ones, and every bot call is transcribed, so QA coverage on those seats is complete rather than sampled. The bot's own report is explained in reading a bot's disposition report.
- Dials, connects, contacts and transfers, with the three funnel rates
- Billable transfers, billable rate and the buyer's return reasons
- Billable transfers per fronting hour, by agent
- Cost per billable transfer and margin
- AMD false positives from a sample of 20 machine-dispositioned calls
- Abandon rate, month to date
- Average QA score and every compliance fail, by name
