> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dovetail.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Run a voice-of-customer program
Most companies don't have a listening problem. They have a consolidation problem. Support knows what's breaking, sales knows what's blocking deals, research knows why, and product sees none of it in one place. Everyone is looking at a different slice, so every conversation starts by arguing about whose numbers are right.
A voice-of-customer (VoC) program fixes that by building one intelligence layer instead of many scattered dashboards. Every source of customer feedback flows into the same place, gets analyzed the same way, and becomes something every team can query and act on.
This page is a map, not a manual. It walks through the four stages of a VoC program in Dovetail and points you to the feature pages where the actual steps live.
Dovetail's VoC workflow is built on [Channels](/help/channels). If you're setting one up for the first time, read [Channels 2.0](/help/channels/channels-beta) first — it's the model the rest of this page assumes.
***
## Stage 1: Centralize the feedback
The first job is getting everything into one place. Not a summary of everything — the raw material, continuously synced.
In practice that means connecting the sources where customers are already talking to you:
* **Support** — Zendesk, Intercom, Freshdesk, Front, Jira Service Management, ServiceNow CSM, Zoho Desk
* **Reviews** — App Store, Google Play, G2
* **Surveys and product signals** — Qualtrics, Sprig, Pendo, PostHog, Usersnap
* **Sales and CS conversations** — Gong, Grain
* **Anything else** — CSV, the [Dovetail API](/integrations/dovetail-api), or [Zapier](/integrations/zapier)
You can put every source into a single combined VoC channel, or split them into separate channels by product area or team. Both work — the choice depends on whether those sources serve the same decisions.
See the full list of supported sources and how to connect, configure, and manage them in a channel.
Deeper research — interviews, usability tests, sales call recordings you want to analyze closely — belongs in [Projects](/help/projects) rather than a channel. Both feed the same workspace, so [Chat](/help/chat) and [Agents](/help/agents) can reason across them together.
Decide what each channel is for before you connect anything. [Build a Channels strategy](/help/channels/build-channels-strategy) covers how to scope channels so they stay useful as volume grows.
***
## Stage 2: Make sense of it
Centralizing feedback is only useful if someone can read it. At real volume, nobody can.
Channels 2.0 does that work by organizing everything into two layers:
**Ideas** are summaries of what's worth doing next, each one grouping the feedback that points at the same underlying problem or request. They're ranked by how much they matter — your business context, how often something comes up, how many accounts are affected, and how much ARR sits behind it.
**Evidence** is the raw material underneath: the original tickets, reviews, survey responses, and conversations. It's where you go to check a claim, pull a quote, or see exactly who said what.
This is what separates a VoC program from a feedback inbox. You get a ranked list of what to fix, and every item on it traces back to named customers and real revenue.
Two things make the ranking meaningfully better:
* **Context.** Tell the channel what you care about with focus areas and linked workspace docs — a strategy doc, a tone-of-voice guide. Ideas are shaped by what you tell Dovetail matters. See [Adding context to your Channel](/help/channels/channels-beta#adding-context-to-your-channel).
* **Contacts.** Idea rollups show ARR, plan tier, and account only if your [Contacts database](/help/contacts/index) is connected to your CRM. Without it you get volume; with it you get weighted priority.
Channels needs enough material to find real patterns. Aim for at least 200 data points before you judge the quality of your ideas.
***
## Stage 3: Make it self-serve
A VoC program that depends on one person producing a monthly deck isn't a program — it's a bottleneck. The goal is for anyone to get an answer without asking you first.
Stakeholders ask questions in plain language and get cited answers across projects, channels, and docs. "What are churned accounts saying about pricing?" is a question anyone can ask.
Track the movement, not just the moment. NPS, CSAT, sentiment, and keyword trends over time, with drill-in back to the feedback behind every number.
Chat answers point-in-time questions. Dashboards answer whether things are getting better or worse. You want both.
For teams who live in Slack or Microsoft Teams, [Chat in Slack and Teams](/help/chat/chat-in-slack-and-teams) puts the same answers where they already work — the single highest-leverage thing you can do to get a VoC program adopted.
Save a view in your channel for each team — by product area, region, or segment — so squads return to the same slice of feedback instead of rebuilding filters every time.
***
## Stage 4: Act, and close the loop
The last stage is where most VoC programs quietly die. Feedback gets analyzed, a report gets written, and nothing changes.
**Push work into the tools your team already uses.** From any idea, send to Jira or Linear to create a ticket that stays linked back to the idea it came from. That link is what lets you trace a straight line from a customer complaint to the thing you shipped.
**Automate the reporting.** Instead of assembling a VoC update by hand each month, configure an [Agent](/help/agents) to do it on a schedule or in response to an event — summarize what's new in a channel, flag an emerging issue, or escalate risk to Slack, Teams, or email. Every output stays traceable to its source.
**Write it up when it needs to persist.** [Docs](/help/docs/getting-started-with-docs) includes a Voice of Customer framework for turning analysis into a shareable report, with citations that link back to the underlying data. Use it for quarterly reviews and exec narratives — not for the weekly cadence, which an agent should handle.
**Tell customers what changed.** When you move an idea to **Resolved**, use `Notify customers` to draft a personalized update to the contacts whose feedback drove it. This is the part almost nobody does, and it's the part customers remember.
Closing the loop is also what keeps feedback coming. Customers who hear back are far more likely to tell you something useful next time.
Want to pressure-test a decision before you commit to it? A [Digital twin](/help/agents/digital-twins) is an agent built only from what a specific segment has actually said, so you can ask it questions the way you'd ask a customer — and ask it to cite the interview or ticket behind any answer.
***
## Who does what
A VoC program only works when it isn't one person's side project. Roughly:
| Team | What they get from it | Where they spend their time |
| :----------------- | :----------------------------------------------------------------------------- | :-------------------------------------------- |
| **Product** | A ranked, revenue-weighted view of what to build next | Ideas feed, sending to Jira or Linear |
| **Research** | High-volume signal to target deep research at, and a place to publish findings | Projects, Docs, channel context |
| **Support and CX** | Early warning on emerging issues before they become ticket spikes | Dashboards, saved views, agent alerts |
| **Sales and CS** | What's blocking deals and putting accounts at risk | Chat, digital twins, account-level rollups |
| **Leadership** | One picture of the customer instead of five competing ones | Dashboards, VoC docs, scheduled agent reports |
One person usually owns the program — keeping channel context current, tidying ideas, and making sure the loop actually closes. That role matters more than the tooling.
***
## What good looks like
**Signs it's working**
* Stakeholders ask Chat before they ask you
* Roadmap discussions cite ideas and evidence, not anecdotes
* Tickets link back to the feedback that prompted them
* Customers get told when something they asked for ships
* The same numbers show up in product, support, and exec reviews
**Common mistakes**
* Connecting every source at once, before deciding what each channel is for
* Leaving channel context empty, then judging the ideas it produces
* Skipping CRM enrichment, so priority is volume-based rather than revenue-weighted
* Treating a monthly report as the output — the output is decisions
* Never closing the loop, so the feedback stream quietly dries up
***
## Where to go next
Ideas, Evidence, context, and closing the loop.
How to scope channels so they stay useful over time.
Automate the reporting, alerting, and escalation.
Every source you can connect to Dovetail.