Reports & exports

Graham insight report

Graham writes a plain-English insight report for your study. It covers:

An executive summary (setup, sample size, and the headline story).
Recommended groups: how many, why, and card-by-card commentary with agreement levels and the labels participants used.
Where participants agreed and where they disagreed (and what that risks).
Segment notes (differences across screener segments), data quality, and prioritized next steps.

The report's groups are the card groups the tool builds from how often participants placed cards together, cut to your chosen group count, over the responses currently in view. They are not the group names participants typed; each group is named by your saved name (or Graham's), otherwise the most common label participants used for it. A set-aside or filtered-out response does not feed the report. A short line on the Report tab says the same, so this is clear on screen.

On a card sort you can also choose which clustering the groups come from, with a toggle on the Report tab: Average agreement (the default) or Best pairs (strongest-pair / single-linkage, the same choice the Dendrogram offers, which often holds up better with fewer clean responses). The report, its charts and its exports all follow the method you pick, and a saved report remembers which one made it.

How it works:

Click Generate report. It builds from your current group count, method and filters.
It's saved per study, so it persists across sessions and devices. Changing the group-count slider or the grouping method hides the old report and offers a Regenerate. The report always matches the group count and method it was made at.
If Graham can't write the report, you get a basic version instead: the facts and figures, without the written interpretation. You'll see an amber banner at the top saying Graham didn't write this report, with the reason and a Try Graham again button. The report is also marked "basic report" next to the participant count, which is what shows on a print or export.
The most common reason is that Graham ran out of time on a large study. Trying again usually works. If the banner says Graham couldn't be reached, that one's on us and we've already been alerted, so there's no retry button.
A basic version never replaces a proper report you already had. If you press Regenerate and that run fails, your existing Graham report stays saved.
Re-opening a report never rewrites it. All four study types keep the report you already have and serve that, so going back to the tab is instant and costs you nothing against your daily Graham allowance. It is only rewritten when the underlying data has actually moved: a new response landing, a quality verdict changing, or a late reconciliation of a bought audience. Pressing Regenerate always writes a fresh one.
On a read-only (Viewer) seat the Generate and Regenerate buttons are not shown. A viewer can read and download a report somebody has already written, but not start one.

By plan: the Free plan shows one Graham insight report on screen (view only, no downloads). Paid plans (Individual and up) get unlimited Graham insight reports and can download them (see below).

Exports (paid plans)

Downloads are available on Individual, Team, and Enterprise, and they are the same five on every study type (card sort, pairwise, Best-Worst and findability):

The written report, from the Report tab once Graham has written it:

PDF: a real PDF file, one click, built from the same outline as the Word file with the charts in it, set in the SortedResearch faces (embedded, so it looks the same on every reader). It is not the browser's print dialog; Print is a separate button if you want the on-screen version on paper.
Word (.docx): the written report with the charts embedded (and, on a card sort, the card key). Branded.
PowerPoint (.pptx): a branded deck: title slide, a slide per section, and chart slides with talking points.

The Word file and the PDF are printed from one shared outline, so they always carry the same sections in the same order.

The data, from the Download row just under the tab strip, on every view of the Results tab and needing no report:

Excel (.xlsx): one workbook with every table the results page can show, one per sheet, plus a Raw data sheet. The sheets are built from the same numbers as the page's CSV buttons, so a sheet and a CSV cannot disagree.
Raw data (CSV): the whole study in long format, one row per judgement (a card placed, a pair chosen, a best and worst pick, a task attempt), with the participant's id, timings, quality flag and every screener answer repeated on each row. This is the shape R, SPSS and a pivot table expect.

Each chart on the Analysis views also keeps its own Download CSV of the numbers behind it, for example on a card sort:

Similarity matrix: card-by-card agreement percentages.
Popular placements: card × category counts.
Suggested groups: group name, card count, agreement %, and card list.
Raw assignments: one row per participant-card placement (long format).

On the Free plan, reports and data are view-only, with no downloads. Upgrade to export.

If a single chart fails to capture during a Word/PowerPoint export, the export still completes without that chart rather than failing entirely.

Pictures in your results and downloads

If your pairwise items carry pictures, they follow the data through:

On screen, a thumbnail sits beside each item in the ranking table and in the opportunity matrix key, so you can tell at a glance which item a row is about.
In PowerPoint, the pictures get their own slides, each labelled and numbered to match the ranking, with animations marked as animated. A deck carries up to 24 pictures; past that it says how many it left out rather than quietly showing a partial set.
In Word, the same pictures are embedded in the document. An animation appears as a still, because a Word document cannot play one.
In CSVs, nothing changes: the columns are the same as they have always been.

The pictures inside a downloaded file are part of the file, so a deck or document you save keeps working after you send it on.