The Compare tab is where researchers turn audience filters into structured side-by-side reports. Pick two segments, and Tellet generates a presentation-quality comparison covering shared themes, segment-specific patterns, statistical differences, and a per-question breakdown — all backed by the same interview data you see in Overview and Questions.
Hub — every comparison you generate is saved and listed in one place. Open old comparisons, delete them, or start a new one in two clicks.
Statistical overview — completed interviews, completion rate, average and median time, started interviews (when screener data is available), and screening pass rate (when screener entry data is available for both segments), side by side with deltas.
Sample profiles — demographic distribution for each segment, as small bar charts in the segment colours so you can read who is in each group at a glance.
Executive narrative — an AI-written summary that opens with the single most important difference and ends with actionable recommendations. Its length scales to your sample size (shorter for small studies, fuller for large ones), so it never reads as padding. Renders in the project's analysis language.
Shared themes — patterns that show up in both segments, with notes on how each manifests differently.
Unique to A / Unique to B — patterns specific to each segment, side by side.
Per-question breakdown — for each interview question, a one-paragraph summary per segment plus an explicit list of key differences and shared patterns, with a category bar comparison when the question has been categorised.
Stale detection — comparisons remember which filter snapshots they were generated against; if you re-run analysis on a filter, comparisons referencing it are flagged as stale so you know to regenerate them.
Export — copy the full report to the clipboard, or download it as a Markdown file ready to paste into Notion, Google Docs, or a slide deck.
64+ languages — the report renders in the project's analysis language.
Navigate to a project, then click the Compare tab in the secondary navigation, alongside Overview, Questions, Transcripts, and Research Assistant.
The Compare tab appears once your project is marked Completed (from the project status menu). While a project is still collecting responses the analysis keeps changing, so comparisons — along with audience filters and the Research Assistant — stay hidden until you close data collection.
Make sure you have at least two audience filters for the project. Filters can be created from the filter dropdown in the Overview, Questions, or Transcripts tab. (If you have fewer than two, the Compare tab shows a guidance message.)
Open the Compare tab. If no comparisons exist yet, click New Comparison in the empty state. If comparisons already exist, click + New Comparison in the top-right of the hub.
In the modal, pick Segment A and Segment B from the dropdowns. The two filters must be different.
Click Compare. Tellet generates the comparison and shows live progress messages ("Loading segment data...", "Generating comparison report..."). Generation takes a little while and scales with the size of the project.
When the report is ready it opens automatically. The URL contains the comparison ID, so you can bookmark or share the link with teammates with project access.
Both filters must have completed analysis. If you try to compare a filter that hasn't been analysed yet, you'll get a clear error message.
The two filters must be different — comparing a filter against itself is rejected.
Each filter must match at least one respondent. A filter with zero matching conversations has nothing to analyse, so it can't be used in a comparison.
Reports follow a deliberate top-down narrative order:
Hero card — the two segment names side by side with sample sizes (e.g. "7 of 25 respondents" vs "3 of 25 respondents") so you immediately see how the comparison is weighted.
Executive Summary — the narrative opens here, not at the bottom. This is what you read to a client first.
Statistical Comparison — the metric table with delta column.
Sample Profiles — demographic bars per segment.
Shared Themes — themes that appear in both segments.
Unique to {Segment A} and Unique to {Segment B} — side-by-side cards showing what's specific to each.
Per-Question Comparison — accordion of questions with summaries for each segment, key differences, shared patterns, and a category bar chart when categorisation is available.
Each segment is consistently colour-coded throughout the report:
Segment A uses Tellet's emerald (green)
Segment B uses Tellet's pink
This makes it easy to follow a single segment across multiple visualisations — the same colour appears in the metric table headers, profile bars, segment-specific finding cards, and per-question summaries.
Each segment card shows "N of M respondents", where N is that segment's respondent count and M is the project's total respondent count, the same total used by the Executive Summary. Segments can overlap (a respondent can belong to more than one filter), so M is not the sum of the two segments' sizes, and it stays the same no matter which two segments you compare within a project. The narrative is calibrated to the segment sizes: small samples are flagged as directional, larger samples are reported with more confidence. When a segment has very few respondents, the report shows a warning noting that the results for that segment may not be representative.
The hub lists every saved comparison for the project:
Title — "Filter A name" vs "Filter B name"
Sample sizes — respondents in each segment
Date — when the comparison was generated
Stale badge — appears when one of the filters has been re-analysed since this comparison was generated
View / Delete actions
Click any row to open the report. Click the trash icon to delete a comparison (deleting removes the saved report; you can generate a fresh one at any time by comparing the same two filters again).
In the comparison report header you'll find two buttons:
Copy — copies the full report (executive summary, statistical table, themes, unique findings, per-question breakdown) as plain text, formatted for pasting into a document or email.
Download — downloads the entire report as a Markdown file. The filename is comparison-{filter-a}-vs-{filter-b}-{date}.md and the contents are formatted with proper headings, tables, and sections so it renders cleanly in any Markdown viewer (Notion, GitHub, Obsidian, GitLab, etc.).
When you re-run analysis for a filter, every saved comparison that references that filter is automatically flagged as stale. Stale comparisons still open and read fine — the data they show is just from the previous analysis snapshot. To get fresh insights, regenerate the comparison: open the Compare tab, click + New Comparison, pick the same two filters again, and Tellet creates a new report against the latest data.
If the "Refresh analysis" gate appears because a filter's saved analysis is out of date, clicking Refresh analysis re-runs it. If that filter matches zero respondents, Tellet tells you right away that there's nothing to analyse instead of leaving the refresh running.
Comparisons are also automatically deleted when one of their filters is deleted — the hub updates immediately.
Comparison data is accessible from the Research Assistant chat. When you ask a question that involves comparing two segments and a completed comparison already exists, the assistant retrieves that report rather than re-running the comparison. If the comparison does not yet exist or is still generating, the assistant will prompt you to create it from the Compare tab first.
Languages — the report content is in the project's analysis language. The interface around the report (button labels, dialog text, error messages) is currently English and Dutch only.
One pair at a time — you can only compare two filters per report. To compare three or more segments, generate multiple pairwise comparisons.
Sample size minimum — comparisons technically work with very small samples but the smaller segment drives a confidence rating: very small segments are flagged as directional, and larger segments are reported with greater confidence. Aim for a healthy number of respondents in each segment before treating differences as conclusive.
Screening pass rate conditionality — the screening pass rate metric is only included in the statistical overview when screener entry data is present for both segments. Projects that do not capture screener entry data will not show this metric.
"Both filters must have completed analysis before comparing" — One of the two filters hasn't been analysed yet. Apply the filter from the Overview tab, click Run to generate the analysis, wait for it to finish, then come back to Compare.
"Cannot compare a filter against itself" — You picked the same filter for both Segment A and Segment B. Pick two different filters.
"Refresh analysis" fails for a segment with no respondents — If a saved filter matches zero conversations, there's no data to analyse. Tellet reports this immediately rather than leaving the refresh spinning. Adjust the filter's criteria so it matches at least one respondent, then refresh again.
Stale badge appears — One of the filters has been re-analysed since this comparison was generated. Click + New Comparison with the same filter pair to refresh the report against the latest data.
Comparison times out — Very large projects (hundreds of interviews, dozens of questions) can take longer to generate. Try narrowing the filters, then retry.
"Comparison failed. Please try again." — A temporary error occurred. Try once more — if a result was already generated, Tellet will reuse it.