Who are the callers who mismatch?
When a caller fails to match, the instinct is suspicion. Put in your call volume, how often fraudsters call and your system’s error rates to see how many of the callers it turns away are your own customers.
Genuine customers among the callers who mismatch
99.2%
49,980 of 50,376 mismatches, on 1,000,000 calls
- 49,980 genuine customers turned away (false rejects)
- 396 fraudsters turned away, of 400 who called
- Genuine customers per fraudster caught
- 126
- Turned away for every fraudster the system stops.
- Fraudsters accepted
- 4
- False accepts: fraudsters the system let through.
The key insight
99.2% of the callers who mismatch are genuine customers: for every fraudster turned away, 126 genuine customers are turned away too.
Genuine customers turned away, beside every imposter call 49,980 against 400
The imposters, enlarged 400 calls
- Genuine customers turned away49,980
- Fraudsters turned away396
- Fraudsters accepted4
Source: SymNex mismatch model, your inputs. Areas are in proportion, except that any group is drawn at least a few pixels wide.
| Caller | Calls | Accepted | Mismatched | Share of mismatches |
|---|---|---|---|---|
| Genuine customers | 999,600 | 949,620 | 49,980 false rejects | 99.2% |
| Fraudsters | 400 | 4 false accepts | 396 | 0.79% |
| All calls | 1,000,000 | 949,624 | 50,376 | 100% |
Source: SymNex mismatch model, your inputs. Imposter calls accepted: 1% of all imposter calls.
| False reject rate | One in100 | One in250 | One in1,000 | One in2,500 | One in10,000 |
|---|---|---|---|---|---|
| 1% | 50% | 71.6% | 91% | 96.2% | 99% |
| 2% | 66.7% | 83.4% | 95.3% | 98.1% | 99.5% |
| 3% | 75% | 88.3% | 96.8% | 98.7% | 99.7% |
| 5% | 83.3% | 92.6% | 98.1% | 99.2% (your case) | 99.8% |
| 10% | 90.9% | 96.2% | 99% | 99.6% | 99.9% |
Source: SymNex mismatch model, with your other inputs held. Your case is outlined.
Your result
99.2%
of the callers who mismatch are genuine customers, on 1,000,000 calls.
How the model works
The calls you enter are every call in the period you choose where the caller says they are a registered customer. One call in every attack rate is a fraudster. If you count people calling for a customer, that percentage of calls is theirs. Everyone else is a genuine customer.
The system turns away genuine customers at the false reject rate, and accepts fraudsters at the false accept rate. People calling for a customer are accepted at the false accept rate times their success multiple, because relatives and friends often sound alike. Everyone who is not accepted mismatches.
The model assumes:
- every caller is enrolled and checked, so no call skips the system;
- each fraud attempt is a separate call, with no repeat attempts and no targeting of unenrolled customers;
- the error rates hold for every call, though in practice they vary with the line, the device and how long the caller speaks.
Every figure is in proportion, so it is an expected value, not a forecast for any one period. Use the error rates your partner measured at your operating threshold, or your own calibration results.
The calculator runs in your browser and nothing you type is sent anywhere. Chapter 12 of Unlock Your Call Centre covers how to design the mismatch process for the genuine majority.