Skip to main content
This guide covers how your team can analyze data in projects — bringing data in, working with highlights and tags, analyzing data with Dovetail Chat, and turning findings into docs you can share.

​
1. Creating a project

Think of a project as a container of data — a set of customer calls, survey responses, support conversations, or whatever you’re bringing in — grouped around a single topic or initiative. To create one, navigate to the folder you want it to live in, and click New or the + sign in the upper right-hand corner. If your workspace has project templates, use one. Your admin team configured them so the right workspace fields are already mapped, tags are linked with descriptions, the views are set, the highlight-automation level is chosen, and setup instructions are written in — i.e. all the configuration is done for you. This creates standardization across your workspace and allows you to get straight to your data instead of rebuilding the same scaffolding each time. (Templates appear at the bottom of the list when you go to create a new object in Dovetail.) If your workspace doesn’t have any templates yet, it’s worth working with your team to create one so everyone has a standard jumping-off point. Once the project is created, give it a name and add your data. There’s no single right way to name projects, but it helps to agree on a standard across your workspace so things stay easy to find — many teams include the month and year, since that’s an easy way to tell at a glance how old something is.

​
2. The project overview

Before analyzing, fill in the project’s Overview page. It does double duty: it gives colleagues a clear understanding of the source and purpose of the data, and it feeds Dovetail’s AI context for every chat in that project. You don’t need a special format. If you already have a project brief, simply paste it into the overview. Once it’s there, you can ask the project things like “summarize the answers to the key questions in the overview so far.” There’s also a workspace-level custom context (under Settings) that shapes every chat and agent for everyone — your glossary and terminology (for example, “we call them members, not customers”), domain guidance, and tone or formatting preferences. This is controlled by your workspace’s admin.

​
3. Getting data into a project

A data object is the smallest unit of information in Dovetail — a single call, an uploaded document, a survey response set. You can bring data in a few ways:
  • Manually — upload a file from your computer, or pull from a connected integration (Zoom, Teams, Google Meet, Google Drive). Connecting an integration means Dovetail reaches across and grabs the recording directly, so you skip the download-and-re-upload step.
  • Automatically, via the calendar integration — connect your calendar (and Zoom or Google Meet) under Settings → Integrations, then set a rule so that, for example, every meeting whose title contains a certain label routes straight into a chosen project. Once configured it runs hands-free, pulling new calls in as they happen.
  • As a blank “take notes” data object — not everything comes from a file or an integration. You can create a blank data object directly in the project to type up field notes, paste in a PDF, or add any other material that didn’t originate from an upload or connected tool — so it lives alongside the rest of your data and can be analyzed with it.
A few things about automatic import worth knowing: rules are per-user, based on your own calendar and recordings, and you can only auto-import calls you own. Several teammates can each point their own automation at the same shared project. There’s also a public API and MCP write access if you want to build a custom upload pipeline.

​
What happens on import, automatically

The moment a recording lands, and before you touch anything, Dovetail:
  • Generates a transcript, split by speaker.
  • Attaches the video or audio.
  • Writes a summary. There are over a dozen summary frameworks available for you to utilize. You can always change a data object’s summary type at any time, and you can set a default in your project template.
A couple of practical tips. Upload finished presentations and reports as PDF rather than native files, since native files can render incorrectly if your custom fonts aren’t in Dovetail’s library. And when deciding whether something belongs as data or as a doc, think less about where it came from and more about its reason to exist: if it’s meant to be analyzed alongside other data, it’s data; if it’s a culmination of findings, it’s a doc.

​
4. The Data tab: exploring data objects

The Data tab is where all of a project’s data objects live. Open any object and it expands into a focused view with the source on the left and a column of icons down the right side — each icon is a different lens on that piece of data:
  • Summary — the auto-generated summary (choose from 12+ options).
  • Highlights — every highlighted moment in this object.
  • Tags — the tags that have been utilized in this object.
  • Fields — the metadata for this object.
  • Redactions — where you can find the redacted moments in the transcript.
  • Comments — lightweight notes for collaborators that don’t affect search or AI.

​
Marking participants as contacts and handling names

If you are using Contacts, this section is for you. (Unsure if Contacts is right for your team? Read our short guide here.) When a call imports, speakers come in as generic placeholders — Participant 1, Participant 2, and so on. You can relink each speaker to a real contact so the data ties back to a person (and, if you’ve synced a CRM, inherits that contact’s fields like industry or region). If you have contact automation (currently in beta) turned on, Dovetail can do this matching for you — recognizing known participants by email and attaching the right contact automatically, so you only step in for the ones it can’t place. If you need to anonymize, do it at the contact level. The speaker’s name shows in the transcript, so the way to anonymize is to give the contact a placeholder name and store their real name in a field that’s locked down — then most people see the placeholder, while the underlying profile and fields stay hidden. Locking down the contacts database lets people see who’s on a call without being able to click into the profile.

​
Detecting silence

By default Dovetail skips silence in the transcript. However, you can run detect silence to split silent stretches into taggable segments, which is useful for things like usability tests where you want to annotate what’s happening on screen even if no one is speaking.

​
Redactions

You can redact at the moment level — blur the video, redact the audio and transcript, or both. Do it manually at any time, or set workspace and project automation rules to handle it for you (the same three modes as suggested highlights apply: suggest, on, or off).
One important rule to know: you can’t download redacted content. A redacted segment is cut from any reel or download, and the original media is never exported. So to share a redacted file externally, download the original, redact it in another tool, and re-upload.

​
5. Fields in analysis

Fields are the metadata of your data. Think about fields as the unspoken context that’s true for the whole object (role, segment, industry, region, methodology, team, product line). Fields are what give your data a life beyond a single project: they let you filter and combine data across projects and folders, in both chat prompts and views. There are data-level fields (identifying information about the participant or the conversation) and doc-level fields (wayfinding for your findings — publish date, owning team, report type), and the two are separate systems that don’t transfer between each other. When updating and reviewing fields, we recommend using the following views:
  • Table — every field as a column; best for spotting empty values and for bulk editing. To bulk-edit, select rows, then use the bottom toolbar’s three-dot menu → Edit field to set a field value across all data objects.
  • Board — group by a single field group, see counts per column, and drag cards between columns to change that field’s value.
You can save any combination of filters and view type as a named view, and Save for everyone to make it a shared default. You don’t have to settle on one way of looking at your data.

​
6. Highlights and tags

A theme runs through everything from here on, so it’s worth saying up front: in the age of AI, you no longer have to tag and highlight everything before you can find an answer. Dovetail Chat reads all of your data equally, whether or not it’s been tagged. That changes the role of manual analysis — tagging and highlighting become tools for storytelling (building reels and reports), while fields become the real requirement for keeping data findable at scale. How much you tag is genuinely up to you and your team; there’s no single right way. With that framing: a highlight marks a moment as important; a tag qualifies why it matters (a pain point, a feature request, a competitor mention). The relationship between highlights and tags is also worth mentioning: a highlight can exist without a tag, but a tag cannot exist without a highlight. Tagging a passage automatically creates a highlight underneath it; highlighting alone just marks the passage. A single highlight can also carry multiple tags. To create one, select text in the transcript and choose highlight or a tag. You can adjust a highlight’s length any time by dragging the bars at its edges. A useful rule of thumb: lead with whether the highlight is the right length. A single stray line often lacks the context to stand on its own, so prefer one slightly longer, self-contained highlight with multiple tags over many fragments.

​
AI suggested highlights

Dovetail can propose highlights for you, with three modes you set per project or in the template:
  • Off — no AI help; you do all highlighting.
  • Suggest (the default) — the AI reads your tags, their descriptions, and what’s already been tagged, then proposes highlights. You accept (check) or reject (X) each one. This keeps a human in the loop, and accepted suggestions list both you and Dovetail AI as contributors, so AI-generated work is always traceable.
  • On — the AI applies highlights automatically for you to review later.
Suggested highlights run as a one-shot pass, not a continuous process. If you add new tags afterward, re-run it. To re-run across many calls at once, select them with the checkboxes, choose Select all, and use the three-dot menu’s Suggest highlights. A few things make suggestions far more accurate. First, write a short tag description. Second, phrase descriptions as positive instructions (“use this when…”) rather than lists of what not to do. Third, you can control which tags the AI draws from using the “use for highlights” toggle on a tag group — turn it on for the groups you want suggestions to pull from, and off for the rest.
Remember the overarching point: chat doesn’t weight tagged content more heavily than untagged content. Tag for the stories you’ll need to tell, not to make data findable.

​
7. The highlight canvas: visual synthesis

The highlight canvas is an infinite whiteboard for clustering highlights — affinity mapping, but with AI doing the heavy lifting. Pull highlights onto the canvas (you can filter the picker by field and/or tag, and you can pull highlights from multiple projects onto one canvas for cross-project analysis). For large sets, scroll to load everything before selecting all, since it loads in batches. Then auto-arrange in one of three ways:
  • By data — groups highlights together by their source data object.
  • By tag — clusters by tag and by unique tag combinations.
  • By theme — clusters on the meaning of the highlight text itself, ignoring tags entirely. This works even if you never tagged anything, and may surface themes you didn’t think to ask about.
You keep full manual control: rename and create groups, nest groups, drag highlights between them, copy a highlight into two places, and draw connections. You can also cluster a cluster — re-run auto-arrange on a single group to keep workshopping it into sharper sub-themes. From any selection you can build a reel, download it, or apply a tag to save the set.

​
8. Highlight reels

A reel splices multiple highlights together to be used as evidence or a storytelling device. To build a custom reel, there are three routes:
  • From the Highlights tab — filter by field or tag (a common trick is to make a tag like “final report” and tag the clips you want), select the highlights, then Add to doc to save it in Dovetail or Download to use elsewhere.
  • Inside a doc — use the / command to insert a reel, then add highlights (including from other projects), dragging to reorder.
  • From the highlight canvas — once you’ve clustered, select a group and turn it straight into a reel (add it to a doc or download it), so the synthesis you just did on the canvas becomes a shareable story.
Reel length is governed by the underlying highlights, so to trim a clip you edit the original highlight’s start and end. To cut out a middle section, make two adjacent highlights. You can also reorder clips, adjust playback speed, and add subtitles.

​
9. Dovetail Chat: asking your data

Chat is how you ask questions of your data, and it works at any scope — a single data object, a whole project, a folder, or the entire workspace. The default context follows where you are. Open chat inside a call and you’re talking to that call; from a project, the whole project; from a folder, everything in it. You can also add context to mix and match — for example, two specific projects and nothing else — using @-mentions in the prompt. Mentioned items turn blue, which is Dovetail’s signal for “this is what I’m acting on.” You can also scope a chat to a specific tag, or bring in web search and uploaded files to compare against outside sources. Citations are embedded and clickable — they bring you straight to the supporting quote, and now show the video clip of the moment, so you can spot-check any answer. The model shows its thinking as it works, and there’s a stop button. To learn more about prompting Dovetail’s Chat, check out our Prompt Guidance doc.
One thing to note: your chats are private to you. No one else can see them. To share an answer, turn it into a doc — which is the bridge to the next section.

​
10. Docs: synthesizing and sharing

A doc (formerly known as insights) is where findings come together — a report, a saved reel, product feedback, or a searchable copy of an external presentation. The fastest way to start one is from a chat response: choose Create doc and it carries the text over verbatim with citations now embedded inline, and it’s fully editable and shareable. You can also start from scratch, use a built-in report template (which prompts the AI to analyze your data and structure a full report you then refine), or generate one from an agent. Inside a doc, use the / command or the + sign on the left side to insert new sections, add structure to the doc, or format your text.

​
Sharing a doc

Sharing a doc works through access and permissions. If you create a new doc inside of a project, it automatically adopts the permissions of the project. If you create a new doc inside of a folder, it becomes a private doc that only you can see, and you can adjust the permissions at any time. Something to note: a doc is indexed by Dovetail’s AI feature based on its access settings. This means your private doc is only indexable by you and no one else until you share it more broadly.

​
11. Surveys

Surveys come in as a CSV, and a little prep makes the analysis work properly.

​
Preparing and importing the file

The golden rule is one question per column, with a single header row. Some survey exports split a question’s answers across several columns or stack two header rows — reshape these so each question is one column whose cells hold the chosen answers, and delete the second header row. To import, create a project, use the import option, and choose your CSV file. Then map the columns down the side:
  • Turn off columns you don’t need or don’t want for privacy reasons — names and email addresses, for instance.
  • Keep the answer columns you want to analyze, and make sure each has a clear question as its header.
  • You can also change responses into Data Fields.
  • The right side shows a live preview of a single response so you can confirm it looks right before importing. When it does, confirm, and Dovetail imports the data.

​
Analyzing the results

Once the data has been processed, Dovetail will automatically create a results doc for you in the Docs tab.
  • If the question was single select or multiselect, the report will create a visual for the question (pie chart or bar chart), so you can easily see the response counts.
  • If the question was open-text, the report will utilize a summary widget that summarizes all the free responses for you.
From there, analysis is the same as any other data. You can create highlights, chat with the data, and create new docs.

​
Where to next

Analysis in channels

See how analysis works for continuous, high-volume feedback.

Highlights

Surface and share the moments that matter in your data.

Data summaries

Choose the right summary framework for each source.

Prompt guidance

Get more out of Dovetail Chat with better prompts.