Survey results

A survey's Results tab charts every question for you, breaks any question down by another, draws how people's answers flow from one question to the next, and shows where people stopped. It has six views in a row of tabs: Questions, Cross-tab, Flows, Drop-off, Report and Respondents. This article covers the first four; the other two have their own articles (the survey report and respondents and quality checks).

The view you are on, and the cross-tab or flow you picked, are kept in the page address, so a reload or a shared link opens the same thing. Anyone in the workspace can read the results, including viewers.

The filter bar

One filter bar sits above the Questions, Cross-tab, Flows and Drop-off views and shapes all of them. Most choices show how many people they cover.

Responses: Live, Test links or Both. While a survey has no live responses yet, the page opens on its test responses so you can see the charts working.
Source: Anywhere, Survey link (your own link) or Panel (people recruited for you). It is hidden when you are looking at test links only.
Who: Completes only (the default) or Everyone who started, which adds screen-outs and people who stopped part way, with the answers they gave before leaving.
Quality: tick boxes for Passed, Held for review, Rejected and Not judged. Rejected is unticked to start with. Only completes are judged, so everyone else counts as "Not judged".

The line on the right says how many responses are being analyzed out of the total. Above the charts, four tiles show Completed (against your target), Started, Screened out and Median length of the completes in view.

The data downloads just above the filter bar hold exactly the people the filter keeps (see survey exports).

Questions

Every question gets its own card, in the order respondents meet them, with its number, wording, question type and base. The base is the number of people who answered it, and every percentage on the card is a percent of the people who answered that question. Skips are left out of the base and shown beside it ("Base: 85 answered, 4 skipped"), so a percentage never stands on its own.

How each type is charted:

Single choice, dropdown, image choice and consent: a bar per answer with its percentage and count.
Multiple choice (pick any): the same bars, plus the average number picked. The shares add up to more than 100%, because one person can pick several.
Rating and Likert scales: the Mean (1 = the first point), the Top two and Bottom two shares, and a bar per point.
NPS: the NPS score (promoters minus detractors), the mean, the split into Detractors (0-6), Passives (7-8) and Promoters (9-10), and a column for each point from 0 to 10.
Slider and number (and a short text asked as a number): Mean, Median, Lowest, Highest and Std dev, with a chart of the spread, one bar per value when there are only a few, otherwise in tidy ranges.
Grid: a table with one row per statement, each row adding to 100% of the people who answered that row, shaded darker for bigger shares, with the row's mean and base.
Rank order: each item's Average position (lower is better), the share who Ranked first, and how many ranked it, best first.
Date: the earliest, median and latest date, with a bar per month.
Open text: every answer listed with the respondent's number (R1, R2 and so on, in the order people started), with a search box. Up to 2,000 answers show per question; the data download has every one.

A question inside a loop gets one card per loop item, labeled with the item, so you see each item's answers separately. A question only an older version of the survey asked still shows, marked "Earlier version", so nobody's answers go missing.

Cross-tab

Choose a question under Break every question down by and every other question is shown as a table: a Total column, then one column per answer to the question you chose, each with its base. You can break down by any choice question (single choice, dropdown, image choice or consent), multiple choice, rating, Likert or NPS question that is not inside a loop (an NPS question splits into detractors, passives and promoters).

A share marked ▲ or ▼ is higher or lower than everyone else's at 95% confidence. Marks only appear where both groups have 30 or more people, because smaller groups mostly mark noise. An answer nobody in view gave is left out as an empty column.

Where a question can be drawn as a flow, a Table / Flow switch above its table shows the same numbers as a Sankey chart.

Flows

The Flows view draws a Sankey chart of how people's answers to one question carry into their answers to another, for example each age group's share of each style. Each ribbon is the people who gave that pair of answers, and each person's ribbon keeps the color of their first answer. Hover a group on the left to pick out where it went.

Under Where groups differ most, up to five suggested pairs sit as buttons, the pairs whose groups differ most. The view opens on the first one.
Or pick your own: From takes any question you could cross-tab by, and To takes a choice, scale, number or date question (NPS can be a starting point but not an end).
+ Add a step carries the flow on through up to four questions in all; to take an extra step off, choose Remove this step in its menu. Extra steps need one-answer questions with at most ten answers (not pick any); numbers are put into about five bands. A longer flow counts the people who answered every step.

The two-question flows read the same cross-tab as the Cross-tab view, so the numbers always match.

Drop-off

Drop-off shows where everyone went, counting everyone who started whatever happened next, so the Who and Quality filters are not shown here.

Tiles for Started, Completed, Screened out, Stopped part way (no activity for an hour) and Answering now, with each as a share of starts.
Why people left early: screen-outs by reason (Screened out, Quota full, Removed for quality, Ended early) and by the ending card they reached.
Question by question: how many of the people who started reached each question, how many Stopped here, and how many were Screened out after it. Routing sends people past questions that aren't for them, so a shorter bar is not always a drop-off; the two right-hand columns are the people who actually left.