From Assumptions to Evidence.
From Reports to Real Conversations.
8,000 hours of customer calls every month - unlocked for operations, strategy, and product.
CallMiner has underpinned contact centre quality and compliance for two years. What's changed is how we use it. With AI-powered discovery, natural language search, and a pipeline that feeds directly into our data lake and AI analytics, it's now a tool for anyone who needs to understand what customers are actually saying — product, strategy, and operations alike.
How customer conversations become intelligence — from speech analytics to AI-powered discovery.
CallMiner is an AI-powered speech analytics platform that transcribes, analyses, and categorises every customer interaction handled in the Contact Centre. Over 70% of all customer interactions at Checkatrade happen over the phone - CallMiner makes those conversations searchable, measurable, and actionable.
All calls recorded in Five9 with over 30 seconds of connected talk time and detected speech are automatically mined and transcribed.
From call recordings to conversational AI — an end-to-end customer intelligence pipeline, built in-house.
We've made it as simple as possible.
Three steps to get started. Two search methods that work together.
How it works: You describe intent in plain language. AI understands meaning, not just words.
"Show me calls where customers asked about getting more jobs or leads"
Best for: Exploration, hypothesis generation, finding what you didn't know existed.
How it works: You specify exact words, phrases, and Boolean logic. CallMiner finds calls containing those precise terms.
"sponsored listing" OR "directory" OR "boost my profile"
Best for: Precision, repeatability, compliance auditing.
Think of them as a safety net for each other — neither is perfect alone, but together they catch what the other misses.
A keyword search for "leads" won't find this. Semantic search understands the intent — it knows the customer is talking about lead volume even without using the word.
Semantic search might lump this in with account cancellations. Keyword filters confirm whether the specific terms you care about are actually present.
After searching, use the left-hand panel in CallMiner to narrow results using metadata and tags:
Agent or customer
Contact reason, outcome
Sentiment, emotion
Product categories
Agent details, trade details, call attributes
Look for the AI Assist icon in the upper-right corner of CallMiner Analyse. You can ask it to explore themes, create categories, and generate charts — all in natural language.
Once a pattern is validated, it becomes a category — a reusable filter that tags every call automatically. No more re-running searches. One click, not a query.
CallMiner's AI has already classified your calls into themes — sentiment, emotion, silence, overtalk. Pre-built. Just filter.
The AI handles the universal stuff. But it doesn't know what "Checkatrade Guarantee" or "Sponsored Listings" means. That's what we're building — the Checkatrade layer on top.
The most useful operators for everyday searches.
| Operator | Syntax | What it does | Example |
|---|---|---|---|
| Exact phrase | "words" |
Finds exact words in exact order | "home mover letters" |
| OR | OR |
Matches any of the terms | "banner" OR "display" |
| AND | AND |
Matches all listed terms | "directory" AND "price" |
| NOT | -word |
Excludes calls containing this term | "listing" -billing |
| Proximity | NEAR/n |
Words within n words of each other | "price" NEAR/10 "too expensive" |
| Call location | {seconds} |
Restrict to first/last N seconds | "hello"{90} (first 90s) |
| Speaker | =Agent / =Customer |
Target specific speaker | "sorry"=Agent |
| Wildcard | * |
Matches word variations | cancel* (cancel, cancelled, cancelling) |
("too expensive" OR "rip off" OR "not worth" OR "overpriced")=Customer
Targets only customer speech. Combines synonyms with OR. Speaker filter ensures we're hearing customer complaints, not agents discussing pricing.
("sponsored listing" OR "directory")=Agent{90}
Combines speaker targeting with call location. Useful for checking if agents pitch products early in the conversation.
("cancel" OR "leave" OR "stop my")=Customer NEAR/30 ("understand" OR "sorry to hear" OR "appreciate")=Agent
Uses proximity to find empathy language near cancellation intent. The NEAR/30 gives a 30-word window for the agent's response.
For the complete syntax reference, see the full Search Syntax Guide.
AI won't get every call right. It doesn't need to. The question isn't "is the AI accurate?" — it's "would we rather hear from all our customers imperfectly, or almost none of them perfectly?"
The exact number might be off by 10-15%, but a 3x spike is real. Investigate the trend — don't debate the count.
AI scores reflect language patterns, not intent. Listen to 3-5 of the flagged calls before drawing conclusions about an individual.
Quoting AI-generated percentages as precise facts is misleading. Say "roughly half" or "around 1 in 2" — the direction matters, not the decimal.
Automate quality monitoring for your team.
CallMiner's Score Builder converts spoken words and context into automated quality scores. Before requesting a new score, check if one already exists.
Everything you need in one place.
#ask-callminer
Questions, issues, and requests
Pitcha Rerkaram
Senior Customer Experience Operations Analyst
Pitcha Rerkaram
Senior Customer Experience Operations Analyst