What would voice biometrics save your contact centre?

Three numbers give you an estimate. Every other assumption is on the page, set to a starting value, and yours to change. The estimate, the key insight and the exhibits update as you type.

Your inputs

Your contact centre

All calls in the period, before any self-service.

Agent time per call, including after-call work.

Fully loaded cost of a call an agent handles.

Voice biometrics

Calls from customers with a voiceprint.

Enrolled callers matched in the IVR.

Enrolled callers matched with an agent.

Today’s process

IVR identifies 50%, authenticates 40%, contains 50%. Agents identify 95% in 20 s and authenticate 95% in 40 s.

How identification and authentication work before the change. Starting values, not benchmarks: replace them with yours, and use 0 for a step you do not automate.

Estimated saving with voice biometrics

£847,273

Agent cost saved before technology costs, on 1,000,000 calls in the period you choose, at £5.00 an agent-handled call

  • £478,125 from 95,625 more calls finished in self-service
  • £369,148 from 28 seconds off the average agent-handled call
Strong authentication
79%
Percentage of calls authenticated using voice biometrics.
Zero-friction calls
790,128
Number of calls with seamless biometric authentication.
Callers no longer turned away
33,131
Callers who fail security today and would be authenticated.

The key insight

Most of the saving, £478,125 of £847,273, comes from 95,625 more calls finishing in self-service; the rest comes from agents spending 28 fewer seconds on security per call.

Exhibit 1Agents ask security questions on 11% of calls, against 74% today
  • Twice as many calls finish in self-service
  • 629,497 fewer calls need an agent to ask security questions
  • AFinished in self-service10% → 20%
  • BAuthenticated, then to an agent10% → 20%
  • CIdentified automatically, authenticated by an agent, 9% matched by voice29% → 11%
  • EIdentified and authenticated by an agent, 38% matched by voice45% → 47%
  • DFGNot identified or not authenticated6.4% → 3.1%
  • Voice biometrics at an agent: the agent asks no security questions
See the numbers
How each call ends: share of all calls today and with voice biometrics, and calls with it
OutcomeTodayWith voice biometricsCallsChange
A Authenticated automatically and finished in self-service10%20%195,625+95,625
B Authenticated automatically, then to an agent10%20%195,625+95,625
C Identified automatically, authenticated by an agent29%11%107,703−177,297
D Identified automatically, not authenticated2%under 1%1,047−13,953
E Identified and authenticated by an agent45%47%470,428+19,178
F Identified by an agent, not authenticated2%under 1%4,572−19,178
G Not identified3%3%25,000No change

Source: SymNex security path model, your inputs. Shares are rounded, so a column may not add to 100%.

Exhibit 2Voice biometrics authenticates 79% of calls without a security question

471,378 calls are matched by voice at an agent, so the agent asks no security questions.

  • Voice biometrics in the IVR
  • Voice biometrics at an agent
  • The IVR, as today
  • An agent asks security questions
  • Not identified or not authenticated
  • AAuthenticated automatically, finished in self-service20%
  • BAuthenticated automatically, then to an agent20%
  • CAuthenticated by an agent11%
  • EIdentified and authenticated by an agent47%
  • ✕D, F, G: Not identified or not authenticated3.1%

Source: SymNex security path model, your inputs. Lane width is the share of all calls; the thinnest are drawn wider so they can be seen. Shares are rounded.

Exhibit 3More self-service brings 56% of the saving
  1. £478,125More self-service95,625 more calls finished in self-service
  2. £369,148Shorter agent calls28 seconds off each of 900,000 agent-handled calls
  3. £847,273Estimated savingIn the period

Source: SymNex security path model, your inputs. Self-service is priced at your cost per call.

Exhibit 4Agent calls are 28 seconds shorter, and 95,625 more calls never reach an agent
  • Today
  • Voice biometrics in the IVR
  • Voice biometrics at an agent
  • The IVR
  • An agent

In the IVR

  1. Identified in the IVRVoice biometrics does not change this step
    No change500,000 → 500,000 calls
  2. Authenticated in the IVRMatched by voice biometrics or by the IVR’s own checks, with no agent
    +191,250200,000 → 391,250 calls
  3. Finished in self-serviceOutcome A: no agent at all
    +95,625100,000 → 195,625 calls

With an agent

  1. Matched by voice at an agentPart of outcomes C and E: the agent asks no security questions
    +471,3780 → 471,378 calls
  2. Asked security questions by an agentThe rest of outcomes C and E
    −629,497736,250 → 106,753 calls
  3. Not identified or not authenticatedOutcomes D, F and G: the caller fails security
    −33,13163,750 → 30,619 calls
Agent workload
MeasureTodayWith voice biometricsChange
Calls handled by an agent900,000804,375−95,625
Average handle time300 s272 s−28 s
Agent time on security, per agent-handled call46 s18 s−28 s
Calls needing security work by an agent80%61%−19 points
Calls authenticated94%97%+3 points
of which by voice biometrics0%79%+79 points

Source: SymNex security path model, your inputs. Bars are shares of all calls; the stages are nested, so they do not add to 100%.

Exhibit 5Cost per call moves the answer most

Your saving is £847,273. Each row moves one assumption by the range shown and holds every other input where you set it.

  1. Cost per call £4.00 to £6.00£677,818 to £1,016,727
  2. Automated authentication rate 30% to 50%£944,136 to £749,986
  3. Manual authentication time 30 to 50 seconds£750,559 to £943,987
  4. Enrolled calls 75% to 95%£758,281 to £933,415
  5. Automated containment rate 40% to 60%£767,431 to £926,397
  6. Average handle time 240 to 360 seconds£939,560 to £785,748

Manual authentication rate does not change the saving in this model. Call volume is left out: the saving grows in step with it.

Source: SymNex security path model, your inputs.

Your estimate

£847,273

Estimated saving in agent cost with voice biometrics, before technology costs, on 1,000,000 calls in the period you choose.

Not sure what share of your calls will be enrolled? The value chain calculator works it out from your offer, consent and match rates, and sends it back here.

How the model works

The calculator follows the security path from Chapter 7 of Unlock Your Call Centre. A call meets up to five kinds of decision: automated identification, automated authentication, self-service, manual identification and manual authentication. At each one, a share of callers succeeds and the rest take the other branch.

Multiply the rates along a branch and you have the share of calls that ends in each of the seven outcomes, A to G. The exhibit uses the book’s example rates; the calculator above uses yours.

The saving is what agent-handled calls stop costing: calls that now finish in self-service, each at your cost per call, and the security seconds taken off the calls that still reach an agent. It is a gross figure, before the cost of the technology, of enrolling callers and of the IVR. It covers the calls in the period you choose, so a year’s calls give a yearly saving.

Exhibit 6Rates multiply down the path: at these rates only 12% of calls finish in self-service
The security path: the share of callers taking each branchEvery call starts at automated identification. At automated identification: Success (60% of these calls) leads to automated authentication; Failure (40% of these calls) leads to manual identification. At automated authentication: Success (40% of these calls) leads to self-service; Failure (60% of these calls) leads to manual authentication. At self-service: Success (50% of these calls) leads to outcome A; Failure (50% of these calls) leads to outcome B. At manual authentication: Success (86% of these calls) leads to outcome C; Failure (14% of these calls) leads to outcome D. At manual identification: Success (90% of these calls) leads to manual authentication; Failure (10% of these calls) leads to outcome G. At manual authentication after manual identification: Success (86% of these calls) leads to outcome E; Failure (14% of these calls) leads to outcome F. Outcomes: A, authenticated automatically and finished in self-service: success; B, authenticated automatically, then to an agent: success; C, identified automatically, authenticated by an agent: success; D, identified automatically, not authenticated: failure; E, identified and authenticated by an agent: success; F, identified by an agent, not authenticated: failure; G, not identified: failure.AutomatedID60%40%AutomatedAuth40%60%Self-service50%50%ManualAuth86%14%ManualID90%10%ManualAuth86%14%100%ABCDEFGSuccessFailureThe security path: the share of callers taking each branchEvery call starts at automated identification. At automated identification: Success (60% of these calls) leads to automated authentication; Failure (40% of these calls) leads to manual identification. At automated authentication: Success (40% of these calls) leads to self-service; Failure (60% of these calls) leads to manual authentication. At self-service: Success (50% of these calls) leads to outcome A; Failure (50% of these calls) leads to outcome B. At manual authentication: Success (86% of these calls) leads to outcome C; Failure (14% of these calls) leads to outcome D. At manual identification: Success (90% of these calls) leads to manual authentication; Failure (10% of these calls) leads to outcome G. At manual authentication after manual identification: Success (86% of these calls) leads to outcome E; Failure (14% of these calls) leads to outcome F. Outcomes: A, authenticated automatically and finished in self-service: success; B, authenticated automatically, then to an agent: success; C, identified automatically, authenticated by an agent: success; D, identified automatically, not authenticated: failure; E, identified and authenticated by an agent: success; F, identified by an agent, not authenticated: failure; G, not identified: failure.AutomatedID60%40%AutomatedAuth40%60%Self-service50%50%ManualAuth86%14%ManualID90%10%ManualAuth86%14%100%ABCDEFGSuccessFailure
  1. ASuccessAuthenticated automatically and finished in self-service
  2. BSuccessAuthenticated automatically, then to an agent
  3. CSuccessIdentified automatically, authenticated by an agent
  4. DFailureIdentified automatically, not authenticated
  5. ESuccessIdentified and authenticated by an agent
  6. FFailureIdentified by an agent, not authenticated
  7. GFailureNot identified

Source: Unlock Your Call Centre, Matt Smallman