How to turn survey results into a short deck people act on

The GOV.UK Service Manual gives a plan that fits most small surveys. It suggests "1 or 2 slides that outline the research you did", "5-10 slides that describe your findings" and "1 or 2 slides that show what you're doing next" (GOV.UK).

It also explains why that shape works: decks like this "make sense on their own, so can be shared and understood easily by people who miss your presentation." That second audience is the one to write for. Someone will forward slide 4 to a person who wasn't in the room.

A plan for a short survey:

1. Who you asked. The question, how many people answered, when, and how you reached them.
2. The short version. Your top three findings as three sentences. Some people will stop here.
3. One slide per finding. Usually three to six, biggest count first.
4. What you're doing next. One line per finding: the change, who owns it, when you'll check.

That's 6 to 9 slides for most surveys. Add a slide only when a finding needs a second one to show where it splits.

Because the title is often all anyone reads. A title like "Onboarding feedback" names a topic. A title like "Most people who nearly quit got stuck connecting their bank" tells the reader what you heard.

At Penn State, Garner, Alley and colleagues gave two audiences the same spoken words with different slides. One set used "a sentence headline that states the main message of the slide," backed by visual evidence. The other saw common-practice slides. The sentence-headline audience "understood and remembered the content better," and the difference was statistically significant (Penn State).

The UK's Office for National Statistics changed its chart titles for the same reason. Its rule is a title that sums up "the main message shown in the chart in 10 words or less," with dates, places and units moved to a subtitle. It found from web analytics and testing that "skim-reading is a very common practice" (ONS).

A quick test: cover everything on the slide except the title. If someone could act on the title alone, it's done.

GOV.UK lists "a headline that communicates something you've learned," a sentence or two of facts, a sentence or two on why it matters, and "a quote from a participant" (GOV.UK). For a survey with written answers, that becomes:

• The title. One sentence, under about 10 words if you can.
• The count. "23 of 80 people." Say both numbers, so nobody has to guess the base.
• The question, in its own words. Pew warns that "small differences in question wording can bring out big differences in the way respondents answer" (Pew). "What nearly stopped you?" and "What did you dislike?" get different answers.
• Two or three replies. Pick the clearest ones, not the funniest. If the group splits, show one reply from each side.
• What it means. One sentence. "Bank connection is the step to fix first" is enough.

Then run Pew's own check on each slide. Its reviewers ask, "Can this chart stand alone?" and whether, shared out of context, it has "all the critical information needed to understand it" (Pew). A slide with a count but no question, or a quote with no count, fails that test.

Pick findings that pass both of these:

1. Enough people said it. Start with the biggest counts. Read every reply under each one before you write the title, because one label often holds two different problems.
2. You can do something about it. A finding with no possible next step goes in an appendix, or nowhere.

Be careful with close numbers. Pew's advice, when two groups are within the margin of error, is to say their views are "similar or about the same" (Pew). With written answers from a small group, 11 replies against 9 is a tie. Put them on one slide, or say they're close.

Also avoid claiming a cause. Pew calls it "dangerous to ascribe causality." People who mention slow exports may also cancel more, but the survey alone doesn't show that one led to the other.

The numbers and replies below are invented to show the method. They are not from a real survey.

Say a bookkeeping app asks 80 new signups "What nearly stopped you from finishing setup?" After labelling and counting the replies, the team gets this:

Worked out from the table: the counts add up to 80, so each person is in one row. The 11 and 9 are close, so they share a slide titled "20 of 80 worried about old invoices or wanted to start on a phone." Calling either one bigger would overstate a two-reply gap.

Here's slide 3 before and after.

Before. Title: "Onboarding feedback." A bar chart of all six rows. No replies.

After. Title: "23 of 80 people nearly quit at the bank connection step." Under it: the question as asked, then three replies, such as "It asked for my bank login and I didn't trust it yet" and "My bank wasn't on the list, so I gave up for a day." Bottom line: "Fix first: explain what we read and add a CSV upload."

The second version takes ten seconds to read and says what to do. The first needs the person who made it in the room.

• A chart for every question. Show the ones that support a finding. The rest can live in a linked sheet.
• Percentages on small groups. "38%" of 8 people sounds bigger than it is. Write "3 of 8."
• Word clouds. They show single words and lose the sentence around them.
• Every quote you liked. Two or three per finding is plenty. More and people stop reading them.

If you're writing up the replies by hand, how to analyze open-ended survey responses covers the labelling and counting step, and how many survey responses you need helps with how much a small count can carry.

Most of the work in a deck is counting and picking replies. Enform reads every reply, counts what people said, and every count opens to the replies behind it, so you can check a finding before it gets a slide. It can also draft a deck from the results, which you then edit. Our roundup of survey tools that summarize open-ended answers compares how other tools handle this, and pricing is here.