Stop scoring random conversations. Score the ones that went wrong.
In one minute: why random reviews miss the conversations that matter, and how to find four times more.
See the methodology Not ready to talk? Run the 2-minute simulator
Read the transcript1 min read +
Every month, your team handles thousands of conversations. Somewhere in there, a few hundred went wrong.
Your team leaders review a hundred of them, picked at random. Like most companies: sixty-three percent do.
Nine land on a problem. The rest, the wrong answers, the broken promises, nobody ever sees.
But conversations that go wrong leave traces. A low score. A reopened ticket. A transfer. A customer who comes back.
Bring them into one place, and let them choose. Sixty percent on signals, twenty on what's new, twenty still at random, so everyone stays in view.
Same hundred reviews. Forty problems found, instead of nine.
I'd be glad to look at your reviews with you. Let's set up thirty minutes to talk it through — pick the time that suits you.
AI voice · Captions onCredits
Music: “Marty Gots a Plan” — Kevin MacLeod (incompetech.com), licensed under Creative Commons: By Attribution 4.0. Voice: Chanon Asarisi (AI-cloned with consent).
- 63 %of companies pick the interactions they evaluate at randomCOPC Global Benchmarking · 2022
- 35 %of US contact centres say their QA is very effective at improving the customer experienceContactBabel · 2024
- 81 %of support agents say most of their conversations are never reviewedSolidroad State of CX 2026 · vendor survey, 500 agents
Random QA sampling finds problems at the rate they happen. That’s the problem.
problem conversations found in 100 random reviews.
When just under 10 % of conversations go wrong, a random pick lands on one about once in ten. The rest of your review hours go to conversations that were fine.
problem conversations found in the same 100 reviews.
Most reviews go first to what your tools already flagged — low CSAT, reopened tickets, escalations, long handle times, negative sentiment. A small random share stays, so you still see the baseline.
Illustrative month, as in the video and the simulator: thousands of conversations, a few hundred that went wrong, 100 reviews split 60 % on signals, 20 % on what’s new, 20 % at random. Run the simulator to use your own numbers.
Three things you can run next month — with or without us.
- 01
Your current picking method, mapped
How your team chooses which conversations to review today, and which kinds of problem that method can never reach.
- 02
Your alert signals, ranked
The signals your helpdesk, CRM and phone system already hold — low CSAT, reopened tickets, escalations, long handle times — and the one or two worth connecting first.
- 03
A three-basket plan and scoring grid
How to split next month’s reviews between flagged conversations, what’s new and a random control share, plus the grid to score them consistently.
Built for the people who own service quality.
- Customer service
Head of Customer Service
Show leadership that quality is improving, with a method you can defend.
- Quality
QA lead
Spend your review hours on conversations worth coaching, not on ones that went fine.
- Orders & after-sales
Order management & after-sales
Catch delivery, invoicing and returns issues before they turn into lost accounts.
- Support operations
Head of Support Ops
Use the data your tools already collect before you buy another one.
Let’s look at your reviews together.
30 minutes with Chanon Asarisi, no commitment. You leave with a sampling plan you can run next month.
- 0–10 minHow your team picks conversations to review today
- 10–20 minThe alert signals your tools already hold — and the ones to connect
- 20–30 minYour three baskets, and the review grid to score them
- No commitment
- No slide deck
- You keep the plan
Run the 2-minute simulator
Your volumes, your review rhythm, the way you pick conversations today — and how many problem conversations three sampling baskets would catch with the same effort.
Fair questions.
What is a QA sampling plan?
A rule for choosing which customer conversations your quality team reviews each month. Instead of a purely random pick, it mixes conversations flagged by alert signals, what’s new (agents, products, processes) and a small random control share.
Which alert signals should decide what QA reviews?
Start with the ones your tools already record: low CSAT or NPS answers, reopened tickets, escalations, unusually long handle times and negative sentiment. Pick the two you trust most, then add more.
Do we need a new QA tool?
No. We start from the data your helpdesk, CRM or phone system already holds. If a useful signal is missing, we tell you which one and why.
We only review a few dozen conversations a month. Is it worth it?
Yes. The smaller your sample, the more each review has to count — that’s exactly where aiming beats random.
Is this a sales call?
It’s a working session. You leave with a plan you can run on your own. If you want help putting it in place, we can talk about that separately.
Who is Quantum Growth Advisory?
A consultancy for customer service and sales teams, working on CRM, data and AI across HubSpot, Salesforce and Monday.
Meet your peers on November 25th in Paris.
One evening for customer service, CRM and commercial excellence leaders. Four workshops, including one on the KPIs that prove service performance (NPS, CSAT, resolution rate) — with or without AI.
- When
- Wednesday, November 25th · 6:30 pm – 9:30 pm
- Where
- Salons Hoche · 8 avenue Hoche, 75008 Paris
- Format
- 30 seats · 4 workshops · networking
- COPC Inc.Global Benchmarking Series 2022 — Contact Center Quality Assurance (900+ executives worldwide)
- ContactBabelUS Contact Center Decision-Makers’ Guide 2024 — 189 contact centres
- SolidroadState of CX 2026 — survey of 500 support agents (vendor research)
- Illustrative exampleThe month shown in the video and the simulator: 100 reviews, 9 problems found at random, 40 with three baskets. Not a client result.
