The voice of your customers

Your customers tell you constantly what works and what does not — in emails, meeting notes, tickets, notes. The problem is never a lack of information: it is that the information is scattered across thousands of exchanges nobody has time to re-read.

Insights read those exchanges for you and extract two things.

# Mentions and topics

A mention is one specific point raised in an interaction: a feature request, an adoption blocker, a churn signal, a compliment. Every mention keeps a verbatim — the exact wording, in its original language — and a link back to the interaction it came from. You can always go back to the source.

A topic groups the mentions that talk about the same thing. "The export is too slow", "the report takes two minutes to load" and "I can't get my data out on Fridays" form a single topic, even though no two customers used the same words. This grouping is recomputed every night, so a topic can grow, merge into another, or split when it becomes too broad.

Worth remembering

A topic cuts across your accounts. That is what makes it possible to answer "how many customers are telling me this?" — a question no single interaction can answer.

# What the screen shows

The topics table is sorted by number of accounts affected by default — not by number of mentions. One customer repeating the same thing ten times weighs less than ten customers saying it once.

Each row carries:

the mention kind (feature request, adoption blocker, churn signal…) and its family;
the number of accounts and mentions over the period;
the new accounts — those raising this topic for the first time;
the movement, which says what changed (see the next page);
the ARR or MRR exposed, meaning the revenue carried by the accounts involved;
the average sentiment and the maximum severity.

Clicking a row opens the topic: its verbatims, its trend, and the tasks attached to it.

# Sorting out what matters

Two levels of decision, independent of each other.

Triage applies to every topic and does not expire:

Pinned — the topic comes back first, whatever sort you picked;
Muted — the topic disappears from the lists. Its mentions keep counting in the totals: muting hides a grouping, it does not delete data. This is the right move for recurring noise or a bad grouping.

The product lifecycle only applies to topics that express a request. It records where the decision stands — under review, planned, shipped, declined — with a note and, if you want, a link to the matching ticket in your product tool. Every change is traced: who, when, what.

Why it helps

A topic marked shipped or declined lets you answer the customer the next time they raise it, instead of rediscovering the request from scratch.

# Filter, save, export

Filters apply to mentions (kind, sentiment, severity, period) and to accounts (segment, owner, tags, stage, agreement). So you can ask questions like "what are my accounts above €50k ARR coming up for renewal complaining about?".

A useful set of filters can be saved as a view and shared with your team. And everything on screen exports to CSV, topics or mentions — the export replays exactly the filters you applied, so the file says the same thing as the page.

Feature adoption

Movements