Coaching at Distance: Quality Assurance That Survives Offshore
Quality assurance in a contact centre is not really about scoring calls. It is about changing what happens on the next one, and the mechanism by which that change occurs is a conversation between a person and their team leader.
Move the operation six thousand miles away and the scoring survives intact. The conversation is what degrades — quietly, without anybody deciding to let it, because the informal apparatus that supported it no longer exists.
A team leader in a British office overhears a difficult call and leans across. A team leader managing a floor in Manila for a client in Manchester receives a scorecard, a calibration disagreement and a monthly pack. Those are not the same instrument.
What actually goes wrong
Quality frameworks degrade in a predictable order, and specialists who assess offshore quality frameworks tend to find all three stages present in the same operation. The first is the commonest condition in the industry.
Scoring without coaching. Calls are reviewed, scores are recorded, a report is produced, and nothing reaches the agent in a form that changes behaviour. This is the most common state of quality assurance anywhere, and distance makes it more likely rather than less.
Calibration drift. The client’s view of a good call and the provider’s view diverge gradually. Nobody notices until a complaint arrives and the two parties discover they had been scoring different things for eight months.
Volume-weighted sampling. Reviewing in proportion to how often an interaction type occurs rather than to what an error would cost. The result is thorough review of the routine and almost none of the rare, consequential conversation that actually generates complaints.
“Everyone has a QA framework and most of them are measurement systems with no actuator attached,” said Ralf Ellspermann, founder and chief strategy officer at PITON-Global. “You can tell within ten minutes. Ask what changed last month because of a quality finding. If the answer is a training reminder, nothing changed.”
Designing for consequence
The single most useful structural change is to stop sampling by volume.
Classify interaction types by what an error costs: a mishandled vulnerability disclosure, a complaint not recognised, a regulated statement given incorrectly, a safeguarding signal missed. Those categories warrant something close to complete review. Routine servicing needs only enough sampling to spot drift. Consequence weighting remains uncommon, even where service standards across Philippine call centres are otherwise strong.
This feels inefficient to anyone managing to a review-volume target and it is the only version that finds the failures that matter. An operation reviewing two per cent of everything is reviewing almost none of the interactions where two per cent is nowhere near enough.
Making coaching work across distance
“The calibration session is the most valuable hour in the relationship and the first one cancelled when anyone is busy,” said John Maczynski, chief executive of PITON-Global. “Clients who protect it get an operation that scores the way they would. Clients who skip it get an argument eight months later.”
Four practices hold it together.
Coaching time protected in the rota
If coaching happens when the queue allows, it does not happen. Scheduled, protected time per agent per fortnight, treated as production rather than as a nicety, is the difference between a coaching culture and a coaching policy.
Side-by-side listening, synchronously
Reviewing a recording together — agent and team leader, same call, same moment — is considerably more effective than a written score delivered later. It requires scheduling across time zones, which is precisely why it gets dropped.
Joint calibration with the client, monthly
Both sides score the same recordings and then argue about the gaps. Done quarterly it is too late to stop divergence; done monthly it keeps one definition of good in play rather than two.
Team leaders with a span that permits coaching
A leader responsible for thirty agents is an administrator with a coaching objective attached. The ratio is the quiet determinant of whether the rest of this is physically possible, and it is the first thing cut when a programme comes under margin pressure.
The automated layer changes the brief
Quality assurance now has two populations to examine, and most frameworks were built for one.
Human interactions require the review described above. Automated resolutions require a separate sampling regime entirely — reading what the system actually said, verifying whether the customer’s issue was genuinely resolved or merely contained, and testing whether context survived the handoff when the system escalated.
Nothing repays sampling like the moment of transfer, and almost no one samples it — it sits in the gap between two reporting sets and therefore belongs to neither team.
Evidence a UK buyer can actually ask for
Three questions, none of which requires a provider to disclose another client’s material.
How is the sample selected, and what is the rationale for the proportions? A flat percentage across all interaction types means consequence weighting does not exist.
How much protected coaching time does an agent receive, and how is it evidenced? This is a rota question with a factual answer.
Describe three recent occasions on which a review finding altered something — the mechanism rather than the client: who raised it, who signed it off, and how quickly it reached the agents.
That third question carries the most weight, because it is the one that separates a measurement system from a management system.
See also: The Role of Machine Learning in Networking
Coaching, or scoring?
Not the average score, which climbs in every operation over time for reasons largely unconnected to whether service improved.
The useful indicators are the gap between the provider’s scores and the client’s on the same calls, which should be small and stable, and the proportion of quality findings that resulted in a documented change, which should be well above zero and frequently is not.
An operation where those two numbers are healthy is coaching. One where the average score is high and those numbers are unexamined is scoring, which is a different activity that happens to use the same forms.
