QA in Outsourced and BPO Environments
When the agents work for someone else, QA becomes a contractual instrument as well as a quality one, and the incentives point in unhelpful directions.
Outsourced operations introduce a structural conflict: the party being measured frequently controls the measurement.
The conflict
The BPO scores its own agents in most arrangements.
The score feeds a service level obligation and sometimes payment.
The BPO therefore has a direct financial interest in the score being high.
This is not an accusation. It is the structure, and it produces predictable results without anyone acting improperly.
What it produces
Compressed score distributions, near the target.
Calibration drift between client and vendor evaluators.
Selective sampling where the vendor chooses which calls are reviewed.
Disputes resolved internally, invisibly to the client.
Findings framed as agent issues rather than as process issues, since process issues may reflect on the vendor's management.
Structural responses
Client-side evaluation of a sample. The client scores calls independently, and the gap between client and vendor scores is itself the measure worth watching.
Joint calibration, monthly, on the same calls, with the disagreement measured and minuted. This is the single most effective control and it is frequently resisted.
Client access to raw recordings, not just to scores. Without it, nothing can be verified.
Random selection controlled by the client, or automated selection neither party controls.
Analytics run on the client side, over full coverage, which removes the sampling question entirely for anything automatable.
Contractual right to audit the evaluation process, exercised occasionally.
Designing the contractual measure
Do not put the QA percentage in the contract as the quality measure. It is gameable, its uncertainty is large, and it will be optimised.
Prefer outcome measures: resolution rate, repeat contact, customer feedback, compliance failure rate from client-side analytics.
Where a QA score is contractual, specify the scorecard, the sampling method, the calibration requirement and the client's right to re-score, in the contract rather than in an appendix nobody reads.
Specify what happens when client and vendor scores diverge. Without a defined process, divergence becomes a monthly argument.
Multi-vendor operations
Different vendors, different evaluators, different calibration. Cross-vendor comparison measures the evaluation as much as the operation.
Cross-vendor calibration is essential and rare. The same calls, scored by all parties, with the spread published to all of them.
Client-side analytics is the only genuinely comparable measure across vendors, because it applies one method to everyone.
The findings problem
Process findings in an outsourced environment are frequently about the client.
A confusing product, an unworkable script, a broken system are the client's to fix, and the vendor raising them can look like excuse-making.
Create a route that is not adversarial: a joint forum where process findings are raised with evidence, and where the client's own analytics has already identified them independently.
A vendor whose process findings are dismissed stops raising them, and the client loses the most valuable intelligence available from an operation they do not staff.
What good looks like
Both parties score independently and compare.
The gap is measured, minuted and trending downward.
Client-side analytics provides the comparable numbers.
Contractual measures are outcomes, not QA percentages.
Process findings flow both ways and get acted on.
Disputes are visible to the client in aggregate, so an unusual pattern is detectable.
The joint calibration session
The control that does most to resolve client-vendor score divergence, and the one most often resisted.
Monthly, same calls, both parties score independently in advance.
Reveal together. Discuss the items with the widest spread.
Minute the agreed convention, which becomes binding on both.
Measure the gap and track it. A closing gap is the evidence the arrangement is working.
Rotate who selects the calls, so neither party can choose favourable ones.
Include difficult calls deliberately.
Where the gap will not close on an item, the item is ambiguous and should be rewritten rather than argued about monthly.
A contract specifying joint calibration with a measured and reported gap is a materially stronger quality clause than one specifying a score target.