Research Highlight Reel: Every Clip Needs Its Count

A research highlight reel is what the board actually watches. Why each clip should carry its theme and how many participants said the same, and who picks it.

Twelve in-home interviews are coded, the themes are on the slides, and the debrief with the marketing director is on Thursday. One deliverable is still missing: the research highlight reel, three minutes of participants saying the findings in their own words. So on Wednesday night the analyst opens the recordings and starts scrubbing for the quote she remembers from the fourth interview. She finds a better one in the ninth. Nobody writes down which theme it came from or how many other people said the same thing.

Key Takeaways

• A highlight reel is made for decision makers who will not watch the sessions, and it is usually assembled last, from the analyst's memory of the footage.
• Vendor guides recommend reels from 60 seconds to 15 minutes, and a Qualtrics guide suggests the strongest 3 to 5 clips, chosen for clarity.
• Only one of the three guides reviewed here asks for clips to be representative, and none says how to show it to the audience.
• A 2020 meta-analysis of 61 papers found anecdotes outweigh statistics when emotional engagement is high, which is why the most vivid clip tends to win the room.
• Each clip should carry the theme it came from and how many participants said the same thing, with the final choice left to the analyst.

What a research highlight reel is for

The reel exists because the people who make the decision will not watch the sessions.

A research highlight reel is a short video assembled from selected clips of research recordings, such as interviews, focus groups, usability sessions or video survey answers, made to show stakeholders the findings in participants' own words. The Userlytics glossary describes it as a focused summary rather than raw footage, built to highlight themes, critical incidents, quotes and typical behavior.

A theme is a pattern of meaning the analyst identifies across several participants, built by coding transcripts passage by passage. A clip is evidence for a theme. A theme is the claim.

Keep that distinction in view. The reel shows clips. The decision rests on themes. Somewhere between the two, the link usually breaks.

How the reel gets made today

PERSONAL EXPERIENCE On most qualitative projects I have run or reviewed, the reel is the last thing built and the first thing shown.

1. Finish the coding and write the themes into the debrief deck.
2. Remember which participant said something memorable about each theme.
3. Reopen the recordings and scrub through them to find the moment.
4. Cut the clip a few seconds wide, so the sentence lands cleanly.
5. Order the clips into a story, add title cards and music.
6. Blur faces or check consent, if there is time before the meeting.
7. Send the link to the client ahead of the debrief.

Step two is where the evidence chain breaks. The clip is chosen from the analyst's memory of the recordings, not from the coded passages behind the theme. Userlytics notes that clip selection is often the most time-consuming part of the job, and that is exactly the part done under the most pressure.

What the vendor guides recommend

UNIQUE INSIGHT Read three guides side by side and the advice converges on short reels and strong clips. It does not converge on how to prove a clip is typical.

The comparison is ours, built from the three pages linked in the table.

Qualtrics makes a point worth reading closely. Three respondents saying the same thing clearly, it says, persuade more than eight saying it with varying degrees of articulation. That is good editing advice. It also means articulation becomes the filter, and the people who speak most fluently on camera are not a random sample of the people in the study.

Userlytics is the only one of the three that asks for clips to be representative, meaning they reflect a recurring theme and not an isolated outlier. None of them says how the audience is supposed to know which one it is watching.

Three guides, one deliverable, three lengths

A range chart of recommended highlight reel length, on an axis from 0 to 15 minutes, for three vendor guides. Qualtrics, in an article from April 2026, recommends 60 to 90 seconds built from the strongest 3 to 5 clips, chosen for clarity and consistency, and does not name representativeness as a criterion. Suzy gives a range of 1 to 7 minutes with an ideal of 1 to 5 minutes, chooses the most important videos that support the key findings, and does not name representativeness. Userlytics, in a glossary entry from June 2025, recommends 5 to 15 minutes and is the only one of the three that names representativeness, meaning a clip should reflect a recurring theme and not an isolated outlier. Every number comes from the vendor page named and none is a Cassi.ai measurement.

The guides disagree on length by a factor of ten. Only the orange row asks whether the clip is typical, and none says how to show it on screen.

Why the most vivid clip wins the room

A reel works because a face is more convincing than a percentage. That is also the risk.

A meta-analysis by Freling and colleagues, published in Organizational Behavior and Human Decision Processes in 2020, pooled 61 papers on the persuasive impact of evidence type. When emotional engagement was high, statistical evidence was less influential than anecdotes. When engagement was low, statistics persuaded more.

Anecdotal bias is the tendency to give a single vivid story more weight than aggregate evidence pointing the other way. A launch decision watched as a highlight reel is a high-engagement setting by design. So the clip of the one participant who loved the pack, filmed in good light, can outweigh the slide saying most of the group found it hard to open.

Who answers for that? The insight lead who presented it, usually months later, when the launch underperforms and someone asks how many people actually said what the board remembers.

What has to travel with every clip

UNIQUE INSIGHT A clip that leaves the analysis should carry its claim with it, on screen.

• The theme. Which finding the clip illustrates, in the same words as the debrief slide.
• The count. How many participants expressed that theme, out of how many, for example 7 of 12.
• The type of clip. Whether it is typical of the theme or the strongest expression of it.
• The counter voice. When the theme is split, one clip from the other side.
• The selector. The analyst who chose the clip and signed off the reel.

None of this needs a new tool. It needs the clip to come from the coded passage, so the theme and the count are already known when the clip is cut. The same information belongs in the methodology note of the report, so anyone reading the deck later can check how the clips were chosen. We described the same chain for text in qualitative coding with AI without losing the evidence.

Where the clip gets lost, and how to keep it

The break happens because transcripts and recordings live apart. The analysis happens in text. The reel is cut in video. The only bridge is the analyst's memory of where something was said.

Word-level timestamps are time codes attached to each word of a transcript, so any coded passage points to the exact second of the recording. WhisperX, a transcription method published in 2023, was built because standard Whisper output gives inaccurate utterance timestamps and no word-level timing out of the box. With word timing, a coded passage becomes a clip address. The analyst no longer scrubs. She picks from the passages already coded under the theme.

This is the order we recommend.

1. Transcribe with word-level timestamps, and keep them on every coded passage.
2. Count, for each theme in the debrief, how many participants expressed it.
3. Pull candidate clips only from passages coded under that theme.
4. Choose the clip, and mark whether it is typical or the strongest.
5. Put the theme and the count on screen as a lower third.
6. Add a counter voice when the theme is split.
7. Check consent and protect faces before the reel leaves the team.

From coded passage to clip on screen

A flow in five steps down the left side. Step one, a transcript with word timing, where every word carries the second it was said in the recording. Step two, a passage coded under a theme: the analyst codes the text and the time code comes along. Step three, count the theme: how many participants said it, out of how many. Step four, candidate clips are pulled only from the passages coded under that theme, not from memory. Step five, the clip goes on screen with its label: the theme, the count, and whether it is typical or the strongest expression of the theme. On the right, an orange box shows the usual shortcut: the night before the debrief, the analyst scrubs the recordings for the quote she remembers, and the clip arrives with no theme and no count; a dashed arrow shows that this shortcut skips steps 2 to 4 and lands directly at the end. A closing band says the analyst picks the clip and signs the reel, and the label shows the audience what the clip stands for. The figure contains no measured numbers.

The shortcut on the right is how most reels are made. It is faster the night before and slower every time someone asks how many people said it.

Step seven is easy to skip because tools appear to handle it. The Qualtrics support page states that its automatic redaction of personal information applies to the transcript only, and not to the audio or video.

Where Cassi.ai comes in

@Cassi.ai is a software engineering company specialized in the pains of market research, innovation and insights, founded by Cassiano Albuquerque. We work with research agencies and corporate insights teams.

Video Lab is a Cassi.ai product that analyzes ethnographic and in-home test videos and returns behavioral KPIs, with a timestamp and a frame as evidence for every conclusion.

Our Qual Lab takes interview and group material as video, audio or transcript. Video Lab, built for footage, transcribes speech with timestamps, describes what happens on screen, and ties every conclusion to the second of the recording and the frame behind it. The key moments become a library of clips with faces under mosaic, ready for the executive presentation, and uncertain items go to a person before they count.

The theme and the count on each clip are the discipline we agree with each team at the start of the project. The software keeps the clip attached to its passage. The analyst decides which clip represents the finding, and signs the reel as she signs the report. More of our work is in the cassiai.com portfolio.

FAQ

How long should a research highlight reel be?

Guides disagree. Qualtrics suggests 60 to 90 seconds with 3 to 5 clips, Suzy 1 to 5 minutes, and Userlytics 5 to 15 minutes. Shorter reels make each choice weigh more.

How do you choose clips for a highlight reel?

Start from the themes, not the recordings. Pull candidates only from passages coded under each theme, then choose for clarity, and label whether the clip is typical or the strongest.

How do you show that a clip is representative?

Put the theme and the count on screen, such as 7 of 12 participants. When the theme is split, add one clip from the other side.

Can AI make a highlight reel from interviews?

It can transcribe with word-level timestamps and suggest clips for each coded theme. Choosing the clip that represents the finding is an analytical judgment, and it stays with the analyst.

A reel is the only part of a study many decision makers will see. If the clips arrive without their theme and their count, the board is judging the most fluent participant instead of the finding.

If your team still builds the reel the night before the debrief, pick one clip from your last reel and try to say, in under a minute, how many participants said the same thing.

About the author

Cassiano Albuquerque is the founder of @Cassi.ai. He has spent twenty years in market research operations and now builds AI systems for research teams. To see how clips stay attached to the themes they prove, write to him through the contact page or on LinkedIn.

Published by Cassi.ai. Read the full article at https://www.cassiai.com/blog/research-highlight-reel.