Weekly & Monthly PDF reports

Two reports, two audiences

Both reports are generated from the same per-call analysis, but they answer different questions and go to different people.

The monthly review never re-reads calls or re-scores anything. It consumes the weekly analyses as they were signed off, so a number in the monthly deck can always be traced back to the week it came from.

What the reports are built on

The analysis behind both reports is deliberately separate from what the dashboard shows.

• Facts come from the platform — transcripts, agent extensions, call dates and durations, the real agent roster, call counts per agent, and your workspace targets and Call Score weights.
• Judgement comes from the AI analyst reading the transcript — whether the customer was satisfied, whether the script was followed, whether an upset customer was escalated, whether the problem was solved, whether the agent stayed professional, the severity, the flags with their quotes, and the impact tags.
• The platform’s own sentiment, summary and risk tags are not used in the reports at all. They power the app; the report’s quality view is a second, independent read.

Weekly Quality Report

One report per workspace per window. The default window is the last week — six working days — and the period covered is printed on the cover.

Section 1 — where the workspace stands

• The counts table Department, period covered, calls handled versus calls audited, and the number of agents. Handled is every call received in the window after byte-identical duplicates are removed; audited is the subset that carried a usable transcript and was scored.
• The headline One sentence naming how many agents need attention this week, and whether any of them is serious enough to be taken off customer calls today.
• Department call score Opening, Handling and Closing percentages plus the blended average, across all audited calls.
• Calls vs goal, and average call duration Calls handled against the workspace goal (calls-per-day target × working days), and the average call length against the duration benchmark. Both are read against your own targets, so they mean nothing until those are set.
• The five things that matter on every call Five stacked bars, one per quality perspective, splitting the audited calls into fine, small and serious. This is the fastest read in the report: a long red segment tells you which of the five is the problem before you look at any agent.

Section 2 — customer frustration, and flagged conduct

The section opens with a count of calls where the customer was clearly frustrated — a network outage, a billing dispute, a repeat unresolved issue — split into two numbers:

• Handled calmly — the agent kept a professional tone under pressure. This is a credit to the agent.
• Frustrative behaviour — the agent mocked, dismissed or raised their voice at the customer. This is a conduct finding.

Below that, every call graded Action Needed or Critical gets a card, worst first, carrying:

• the agent’s name and extension;
• a 🎧 Listen link that opens the call in the app;
• a short note on what happened;
• the sentence worth flagging — an exact quote from the transcript;
• the compliance flags raised.

Every flag on a card carries a verbatim quote as evidence. A finding without a quote does not reach the report, which is what makes it defensible in a compliance review.

Section 3 — quality, every agent, worst first

The AI review, kept deliberately separate from the platform’s numbers. Each agent gets a ✓ / ⚠ / ✗ on five perspectives:

The What to do rating is each agent’s worst perspective across their calls, not an average:

• Doing great — every perspective clean.
• Doing well — one small miss.
• Needs coaching — two or more small misses.
• Below standard — one serious miss.
• Serious — act now — two or more serious misses, or any Critical call.

Two labels are worth knowing. An agent with fewer than three audited calls is marked score not reliable yet — treat it as a signal to look, not a verdict. An agent with no scored calls shows a dash and the reason, and is excluded from every workspace average.

Section 4 — agent-wise performance, from the platform

The same agents, scored the way the application scores them, so you can see whether a problem is quality or productivity.

The table is sorted worst-first by status, so the row you need to act on is always at the top. An agent can be On track on Call Score and still be Off target because they took two calls against a goal of thirty — that is the status doing its job.

Section 5 — the punch list and the method

• What to do this week, by person — every action has a named owner and a date. Critical findings are dated today and say to take the agent off customer calls and report the call to compliance; coaching actions are dated two days out. Where three or more agents skipped the same opening step, a team-wide refresher is added. The top performer on Call Score gets a praise item — the list is not only bad news.
• How we score — the method in plain language, plus the impact-tag legend.
• Coverage — how many calls were fetched, scored, and skipped, how many byte-verified duplicates were removed, and anything excluded as out of scope. It’s the audit trail for the counts in the first section.

The severity scale

Every audited call lands in one band, and the band decides the language used about it.

What is checked before a report goes out

Three audits run on every report, and none of them is optional:

1. Skip-list audit — every call the pipeline skipped is re-fetched and re-tested, because a skip is a call that silently never got reviewed.
2. Duplicate audit — a call is only removed as a duplicate when its length and file size match byte for byte. Anything short of that is promoted back and scored.
3. Content-attribution audit — each written analysis is matched back against the transcript it claims to describe, so a summary can never drift onto the wrong call.

Monthly Executive Review

One deck per client per month, pooling every workspace across the month’s four weekly cycles. Week 1 is the baseline; Week 4 is current.

Seven sections, in this order. Departments, issue boxes and watchlist rows are all repeating blocks, so a bigger company produces a longer deck with the same design — the sections stay the same whatever the size.

Section 1 — the dashboard

Four KPI cards, each with the movement behind it:

• Call coverage — 100% of calls audited, against the under-0.5% a manual QA team reaches by sampling.
• Departments improving — how many of the workspaces moved in the right direction, and how many need attention.
• Customer dissatisfaction — the fall from Week 1 to Week 4, with both figures shown.
• Average call duration — the same, in minutes.

Then four trend charts, each a four-point series across the month:

Section 2 — what changed behind the numbers

A short list explaining the movement in words — which behaviours changed, not just which numbers did: upset customers handled calmly, repeat offenders coached early, script gaps closed, long calls trimmed toward the benchmark. It reads as the causes behind the four trends above it.

Section 3 — department breakdown

Every workspace as a department, with its monthly call score, its calls per month and a status, sorted best-first. The section names the one department to step into — the note is explicit about which is trending the wrong way, because that’s where the next coaching cycle earns the most.

Section 4 — company-wide issues

Six tiles per department, each a recurring QA category with a status:

Statuses read On track · Improving · At risk · Needs attention. Improving is the important one: it means still below target but moving, which is a different management decision from At risk, which means below target and flat.

Section 5 — defaulter watchlist

The few agents flagged repeatedly across the month: those rated Below standard or Serious in two or more of the four weeks, or carrying repeated compliance flags. Each row shows their Week 1 rating, their status now, and a recommendation.

The point of this section is trajectory, not a snapshot. The same agent is followed on the same issue, week to week — improvement means the coaching landed; a flat line after four weeks means it didn’t, and the recommendation says so.

Section 6 — the value created, in money

Two savings, both computed from call volume and call length alone. Neither needs an estimate of what a customer is worth, which is what keeps the figure arguable-with rather than hand-waved.

• QA auditing replaced Auditing the month's calls by hand would need a number of reviewers — the call volume divided by how many calls a reviewer can audit in a day, divided by the working days in the month. That headcount is priced at a fully loaded weekly rate.
• Dipping average call duration The minutes taken off the average call between Week 1 and Week 4, multiplied by the month's call volume, priced per agent-minute.

The arithmetic is printed alongside, with the rate assumptions used — a standard researched rate set, stated openly so you can substitute your own and watch both totals move. The headline is the monthly figure, with the annual figure beside it.

Section 7 — where it’s helping, and what to decide

A short list of where the platform is earning its place, then the decisions that need a leadership answer this month — which department gets the next coaching cycle, which company-wide fix to roll out, which individual has run out of coaching runway, and confirmation of the rates behind the money section.

How to read them

• Start with the five bars, not the agent table The first section of the weekly report tells you which of the five things is failing across the workspace. Fixing a workspace-wide pattern beats coaching five people individually on the same miss.
• Read the two agent sections together An agent weak on quality but strong on the platform numbers is following the script while handling customers badly. Strong on quality and weak on the numbers usually means a script step is being skipped on otherwise good calls. The two are kept separate so that neither can hide the other.
• Treat the worst-perspective rating as a prompt to listen Both the quality rating and the backend status take the worst signal rather than an average, deliberately — an average hides the one call that matters. Open the call before you act on it; every flagged card has a Listen link for exactly that.
• Check the small-sample warning before judging an agent Fewer than three audited calls is marked as not reliable yet. It is a reason to review more of that agent's calls, not a performance conclusion.
• Use the monthly deck for direction, the weekly for action Nobody is coached from the monthly review — it has no individual detail by design. It tells you which department and which recurring issue to spend the next month on; the weekly report tells you who to sit with on Monday.
• Trust the counts, and check them if you like The coverage line closing the weekly report reconciles: fetched, scored, skipped and duplicates removed. If the audited number looks low against what you sent, that line says why.

Getting a report

Reports are produced and delivered by our team as part of your plan — there is no button in the application that generates one. If you want the weekly report for a workspace that isn’t receiving it, a different window covered, or the monthly review added for your leadership team, talk to us through the contact form.