Sampling and coverage
Nobody reviews every call, so which calls get reviewed decides whether your figures mean anything. This is how Qualitify chooses, why it chooses that way, and how to read what it did.
Where it is
Quality
The same page as the scorecard, below it. Owners only — a reviewer works the queue and cannot change what goes into it, which is the point.
How much to review
A number of calls per agent per period. That is the whole setting, and it is per agent on purpose: a flat company-wide number would quietly give your busiest agent the most scrutiny and your quietest almost none.
Set it lower than feels right
Reviewing a call properly means listening to most of it, scoring it and writing something the agent can act on. That is fifteen to twenty minutes. Four calls per agent per week, genuinely reviewed, beat twenty skimmed — and a queue nobody can keep up with is the single most common reason a quality programme stops after six weeks. Raise it once you know how long reviewing really takes on your team.
Work out your reviewer’s capacity before setting it: agents × calls each × twenty minutes. If that exceeds the hours your reviewer actually has, the queue will grow forever and the Pending evaluations figure on the overview will tell you so.
How calls are chosen
From the answered, recorded calls in the period, spread across your agents, up to the number you set for each.
- Unanswered attempts
- Never sampled. They are counted in your activity figures, because dialling is work, but there is nothing to listen to and a queue full of them is a queue nobody works.
- Very short calls
- Can be sampled, and are worth seeing occasionally — a pattern of ten-second calls is itself a finding.
- Both directions
- Incoming and outgoing both. If your team’s inbound and outbound calls need different standards, that is a reason for a second scorecard rather than for filtering here.
- Both sources
- Desktop and mobile recordings go into the same queue and are scored against the same scorecard. The reviewer can see which is which, so they know what the audio will sound like.
How fair sampling works
The section of the page that explains itself in the product, and worth understanding because it is what makes the figures defensible when an agent disputes them.
The naive approach — review whatever is at the top of the list — concentrates on whoever calls most, at the time of day the reviewer happens to work, which produces a measurement of your reviewer’s schedule rather than your team. Qualitify spreads the sample across agents and across the period instead.
What that buys you
- Every agent is measured on a comparable number of calls, so their averages can actually be compared.
- Nobody can be reviewed only on their worst morning, or only on their best.
- An agent cannot work out which calls get reviewed and perform only on those.
- The figures survive being questioned, which is the real test of a quality programme.
Per-employee balance
The page shows how the sample actually landed per person. Read it when somebody’s average looks surprising — the usual explanation is that they have four reviews and everyone else has twenty, which is not a performance difference.
Reading a sampling run
Each time sampling happens it is recorded as a run, and Run history keeps them. A run tells you what was chosen, for whom, and what happened to it afterwards.
| What you see | What it means |
|---|---|
| Assignment | Which calls were picked for which agent, and which reviewer they went to |
| Decisions | What the reviewer did with each one: scored it, or skipped it with a reason |
| Per-employee balance | How many calls each agent contributed, so you can see the sample was even |
Run history is also the answer to “why was this call reviewed and not that one”, which comes up the first time somebody is unhappy with a score.
Skipped calls
A reviewer can skip a call rather than score it, with a reason. That is a feature, not an escape hatch: a call that is thirty seconds of a wrong number cannot be scored against a sales scorecard, and forcing a number out of it would be worse than skipping.
Skips are kept and visible, which matters — a reviewer skipping a third of their queue is telling you something, either about your sampling or about your scorecard, and the figure is how you notice.
Coverage, and what it is for
Coverage is the share of calls that got reviewed. It is the number that tells you how much weight your quality figures can carry.
- Low coverage and a high score is the dangerous combination. It reads as good news and is mostly an absence of evidence.
- Uneven coverage between agents means their averages are not comparable, whatever the dashboard shows side by side.
- Falling coverage means the queue is outrunning your reviewers. Either lower the number per agent or add reviewer time — raising the target while coverage falls makes the problem worse.
- Stable, even coverage is the point of all of this. Then a score moving means something.
When the queue is not behaving
| What you see | Usually because | What to do |
|---|---|---|
| The queue is empty | No answered, recorded calls in the period — often a recorder problem rather than a sampling one | Check Calls for the period; if it is empty, start with the recorder troubleshooting |
| The queue keeps growing | The target is above your reviewers’ capacity | Lower the calls per agent, or add reviewer hours |
| One agent has far fewer reviews | They made far fewer recorded calls — often calling outside the recorder | Compare their call count with the team’s; check their Source values |
| Nothing new since a date | Recording stopped around then | Check the recorders rather than the sampling |