On Monday morning, the customer-support dashboard announced good news: average handling time had fallen. The queue was moving, and the chart sent to leadership was green. Across the room, an agent closed a query about a duplicate bill with the prescribed reply. She had noticed the same issue in three contracts, but following it meant opening account histories, speaking to billing, and leaving a case open longer than the target allowed. With the productivity goal in view, she chose what the dashboard rewarded: settle the visible fragment and set aside the slower investigation.
That is not a story about a careless employee or a malicious machine. It is about an organisation that uses a metric as its language of value. When average time, closed tickets, or a satisfaction score occupy the whole conversation, they also decide which experiences are easy to tell and which are silenced. A short case fits into a number; an anomaly connecting several errors, an incomplete repair, or an unlabelled doubt does not. The cultural problem starts before an algorithm makes a recommendation. It starts when only work that a dashboard can add up is recognised as work.
A number is not a neutral summary#
Measurement matters. A service that does not know its waits, abandonment rates, or costs will struggle to improve. The difficulty begins when an indicator is treated as a complete photograph of reality rather than a choice about what to notice. Average handling time can expose an unfair delay. It can also conceal that the reduction came from sending difficult questions elsewhere or closing conversations that will reopen tomorrow.
Measurement matters. A service that does not know its waits, abandonment rates, or costs will struggle to improve.
Every metric carries a quiet grammar. It defines who appears efficient, which task is called delay, and which account of a problem looks like an inconvenient exception. If staff are praised for closing quickly, a record containing five erroneous charges becomes a threat to individual performance. If finding a shared cause is recognised, the same record can become evidence of a faulty billing rule. Data do not compel either interpretation; the policy governing their use does.
That is why it is not precise enough to say that metrics simply “dehumanise.” They organise attention. An organisation can use them as a provisional compass alongside listening and professional judgement, or as a tribunal that dismisses evidence which will not fit in a field. In the latter arrangement, people learn what stories are safe to bring forward. They report compressible successes, avoid cases that disrupt the rhythm, and rename a systemic problem as an individual exception.
The price of pushing long cases away#
Long cases are uncomfortable because they break the promise of easy comparison. Resetting a password is not equivalent to discovering why a household has been billed twice for months. Yet an average mixes them and presents the difference as a deviation. The immediate result is fragmented support: every contact fixes a small piece, nobody holds the whole account, and the customer has to repeat the story at each turn.
There is a less visible cost. When an investigation is abandoned to protect an indicator, the service loses knowledge about its own design. In the billing example, three lengthy contacts might point to a migration rule applied twice, confusion between products, or a notice that encouraged people to accept an incorrect charge. Closing each ticket separately saves minutes today but expands tomorrow's volume. The chart improves while the system produces the work it will later penalise.
The burden is not shared evenly either. People with little time, less confidence in digital channels, or more precarious finances are often least able to make five contacts. When speed is rewarded but the persistence needed for repair is ignored, a complete answer becomes a privilege for someone who knows how to keep pressing. No one needs to make an explicit decision to exclude them. It is enough for the effort demanded of the person to remain invisible in the service's assessment.
Making narrative evidence usable#
The alternative is not to replace numbers with vague impressions. It is to give narrative evidence a clear place in the work. A support story can look anecdotal while containing variables the dashboard never anticipated: contradictory messages, a date that makes an appeal impossible, a language barrier, or a rule that affects joint-account holders differently. Its value is not that one story automatically represents everyone. Its value is that it identifies where a closer look is warranted.
A monthly narrative review can begin with a simple discipline: choose three discarded cases, not three exemplary ones. The team retrieves records that were closed because they ran long, transferred without resolution, or marked too complex. For each, it reconstructs four things: the hidden cost borne by the person; the decision induced by the metric; the possible pattern; and the concrete change to test. This is neither a meeting to vent nor an attempt to review every conversation. It is a stable opening for the material that incentives had encouraged staff not to mention.
The selection should leave a trail. Recording why a case was chosen, which documents were consulted, and what uncertainty remains keeps the review from becoming a contest of anecdotes or the preference of the loudest person in the room. Billing, support, and operations may read one record from different positions. That difference becomes useful when it turns into testable questions: How many contracts share this combination? At which step does the duplicate charge arise? What notice did the customer receive before accepting it?
From account to testable adjustment#
Listening is insufficient if the outcome ends as meeting notes. Every reviewed case should yield an adjustment with an owner, a date, and a verification signal. For duplicate bills, the team might create a temporary tag that groups related reports, stop linked tickets from auto-closing, and ask billing for a sample of migrated contracts. The change does not promise to erase the defect at once. It makes clear who will check whether the hypothesis was right.
That adjustment also changes the role of the metric. Rather than treating average time as a verdict, the team can read it alongside reopenings, the number of contacts required to resolve an issue, and the number of root causes identified. None of those figures is complete. Taken together, they make it harder to gain speed by moving cost to the customer or the next shift. The indicator stops dictating one behaviour and begins to ask a question: what have we actually improved, and for whom?
The narrative review should not become another quota that staff must satisfy at all costs. If a team is required to produce three “transformative stories” every month, embellished accounts and artificial categories will follow. The obligation is to keep the channel open, document decisions, and revisit their effects. Some months, three records will expose one common cause. Other months, they will show that no pattern exists and that the right response is better individual guidance. Being candid about that result is part of the control.
Changing everyday recognition#
Metrics govern because they shape promotions, bonuses, meetings, and messages of praise. A statement of principles will therefore not correct their effects on its own. The organisation has to examine what it celebrates. A supervisor who protects time to investigate a recurring billing error needs to know that a worse average for one week will not count against her. Someone who documents a barrier affecting many people deserves recognition comparable to the person who clears a queue rapidly.
It also helps to separate uses. An indicator that is helpful for planning shifts can be harmful when used to judge every agent. Average handling time helps forecast capacity; without context, it does not explain the quality of a conversation or the value of detecting a structural fault. Stating this separation in writing prevents a dashboard designed for operations from quietly becoming a disciplinary device.
For the next cycle, the practical question is not “which number should we remove?” but “which behaviour are we making rational?” Set aside an hour, bring back three long cases that were left behind, and name the cost that fell outside the dashboard. Then choose one adjustment: a tag, an automatic pause, an escalation route, or a second measure. Publish who will verify the result and when. When a verifiable account can sit beside the indicator, the metric stops silencing important work and finally begins to help the organisation see it.
