Call quotas: hitting numbers without burning out

Calls made and conversations had are different quantities, and almost every problem with calling targets comes from measuring the first while wanting the second.

Calls made is easy to count, easy to compare, and entirely within the caller's control. Conversations had — meaning contacts where the intended person actually engaged — is the thing that produces business, and it depends on the list, the timing, and whether the calls were any good.

When a target is set on the first, behaviour adjusts to it, and the adjustments are predictable: dialling numbers unlikely to answer because an unanswered call still counts, ending marginal conversations early to keep moving, skipping the notes, and calling without preparation. Each is rational under the target and each reduces the number that matters.

This is not an argument against counting calls. Volume is real, and a caller making very few calls will not produce results. It is an argument for tracking both, side by side, because the ratio between them is where the useful information sits.

Beyond a certain rate, additional calls per day stop adding results and start subtracting them, for three reasons that compound.

The first is preparation. A call made with thirty seconds of context — who this is, what happened last time, why they are being called — goes materially better than one made blind. At a high enough rate there is no time for it, so every call is blind.

The second is the notes. Records written immediately are usable and records written later are not, so a rate that leaves no gap between calls destroys the log, which destroys every future call.

The third is delivery. Voice quality degrades measurably across an unbroken run of calls: flatter, faster, less responsive. The last calls of a long block are the worst ones, and they are made to numbers that deserved the same attention as the first.

So pushing the rate up converts good calls into mediocre ones. The total goes up and the outcome does not, which reads as the calling not working rather than as the rate being wrong.

The cost that is most often ignored is that some calls take something out of the person making them, and the effect carries into the next one.

A hostile call, a complaint, a conversation with someone in genuine difficulty, or simply a run of flat refusals leaves a residue: elevated stress, a defensive tone, and a reluctance to dial that gets read as laziness. Dialling the next number immediately means the next person receives a caller who is still in the previous call.

A few minutes between difficult calls is not slack, it is part of the work, and building it into the plan is more effective than expecting people to absorb it. So is varying the work — putting note-writing, list preparation or administrative tasks between blocks of calls rather than clustering all the calls together.

The scheduling implication is uncomfortable for anyone setting targets: a caller with fewer calls scheduled and gaps between them will often produce more conversations than the same caller scheduled back-to-back, because every call is better.

The most consequential diagnostic error in calling is attributing a low connect rate to the person dialling.

Whether a call is answered by the intended person depends mostly on things the caller does not control: whether the number is correct and current, whether it belongs to the right person, whether that person is reachable by phone at all, and whether they have any reason to expect the call. A caller working a list of stale numbers cannot fix it by trying harder, and being told to make more calls guarantees the outcome stays the same.

The signature is distinguishable. If connect rates are low across every caller working the same list, the list is the problem. If one caller's rate is much lower than colleagues working comparable lists, that points at timing or approach. And if connections happen but conversations end immediately, the list may be reaching real people who were never plausible customers — which is again a list problem.

Checking which of these applies takes an hour of looking at records and prevents months of the wrong remedy.

Targets are not the problem; single targets on the easiest metric are. A workable arrangement has a few properties.

Track both numbers, and treat the conversation count as the primary one. Calls made becomes a diagnostic — a low count with a healthy conversation count is not a problem, and the reverse is.

Set the volume figure as a floor rather than a maximum, and set it low enough to be met while preparing properly and writing notes. A floor that requires skipping preparation is a target to produce bad calls.

Count quality events explicitly, so the things that take time are visible as work: notes written, next actions agreed, contacts correctly disqualified. Disqualification deserves credit specifically, because a caller who establishes that someone is not a customer has done something valuable and will avoid doing it if only positive outcomes count.

And review the list's contribution separately from the caller's, so that a bad list is diagnosed as a bad list rather than as insufficient effort.

Calling is unusual among business tasks in that most of its individual outcomes are negative, and a practice that does not account for that will not last.

What helps is measuring inputs the caller controls alongside results they do not, so a good day is recognisable even when nobody bought anything. Blocks rather than a continuous day, so there is an end in sight. Varied work between blocks. And an explicit understanding that a refusal is a normal outcome rather than a failure, which sounds like a platitude and has a practical effect on tone: a caller who has internalised it sounds different, and people can hear it.

The structural point is that a person's capacity for this is finite and rebuilt by rest rather than willpower. A plan that assumes otherwise produces a few strong weeks followed by attrition, which is far more expensive than the calls it gained.

Automix's Wani is a planned voice follow-up agent, documented at /docs/wani. It is not available to customers and no calls are placed by it today, so the workload described here is a person's.