Cavefish Founder Presents 'Making Emotions Measurable' at Media Cymru's PLAYBACK Festival

Cavefish founder and CEO Jonathan Prescott presented "Making Emotions Measurable" at PLAYBACK, Media Cymru's Festival of Innovation, held at Tramshed and Tramshed Tech in Cardiff on 16-17 September 2026. The talk argued that every piece of content has an emotional fingerprint, and that creative teams should be able to see it while they are still making the work, not after it has been transmitted.

PLAYBACK brought together the media sector projects funded by Media Cymru over four years, with keynotes from teamLab, panels on inclusion, sustainable production and financing, and a programme of lightning talks on AI and emerging technology. Cavefish took part in Media Cymru's R&D programme in 2025, in collaboration with Transport for Wales.

From 326 dashboards to one question

Prescott opened with his background running data and technology teams, including one organisation with 326 dashboards that could measure customers, operations, performance and conversion, but not how people felt. That gap, and the harder follow-up question of how you want people to feel, led to Cavefish and its EchoDepth platform.

"Reach tells me somebody was there. Completion tells me they stayed. Recall tells me what they say they remember afterwards. But emotion didn't happen afterwards. It happened moment by moment. Every asset has an emotional fingerprint, and if we can see that fingerprint early enough, we can do something with it."

- Jonathan Prescott, Founder & CEO, Cavefish

Attention, emotion, memory, value

The talk framed communication as a chain: attention gets you seen, emotion gives the experience meaning, memory is what survives, and value is what commissioners and funders increasingly ask producers to demonstrate before a single shot is taken. Drawing on Fredrickson and Kahneman's Peak-End research, and the later work showing that the overall emotional journey also shapes memory, Prescott argued that what matters is the shape of the experience: where emotion rises, falls, peaks and ends.

EchoDepth analyses observable signals continuously through an asset, including facial action units and vocal pitch and energy, and maps them across valence, arousal and dominance. The output is a curve that an editor can read without knowing the terminology, showing whether an audience felt what was intended, when it was intended.

The price comparison test

To test the idea, Cavefish analysed four adverts from four competing price comparison companies, made by different agencies, against a benchmark corpus of 138 high-performing ads. The analysis used five scored dimensions with the trace resolved at 25 frames per second. Comparisons resembling System1-style measures are model-derived proxies, not official System1 Star Ratings.

• Convergence, not ranking. All four ads landed in a narrow band (3.96, 3.82, 3.65 and 3.65), with remarkably similar emotional fingerprints despite different briefs and agencies.
• Different structures. Two ads peaked roughly a third of the way through, one at about 51% of the runtime and one at around 90%.
• A shared blind spot. All four scored maximum on warmth and comedy, but averaged 2.3 on right-brain features such as characters in relation, eye contact and a sense of place, against a benchmark of 7.
• A neutral baseline. Face coverage ranged from 35% to 54% against a 99% benchmark, measured joy intensity from 0.06 to 0.09 against 0.25, and the dominant emotion in all four was neutral.

The finding, Prescott said, was not which advert was best, but that an entire category appears to have made essentially the same emotional choice: loud, warm and funny, with a largely neutral emotional baseline that completion rates and reach cannot reveal.

Moving measurement before the edit locks

Panels, focus groups and post-campaign analysis remain valuable for context and explanation, but their findings often arrive after the money has been spent. Cavefish's aim is to move measurement earlier, to the storyboard, animatic and rough cut, so that versions can be compared while the decision is still reversible. In Cavefish's model that means no panel to recruit for each iteration, analysis in minutes per asset, and the same trace applied consistently across versions.

"Why test one final film when you could compare three rough cuts? Why have the argument, 'I prefer version A', 'No, I prefer version B', when you can put evidence into the room? Not to replace creative judgement. To sharpen it."

- Jonathan Prescott, Founder & CEO, Cavefish

An offer to Welsh productions

Prescott closed with an invitation to the Welsh screen sector. Cavefish will analyse the emotional fingerprint of the first three Welsh productions that come forward at no cost: a rough cut, two competing versions, or any edit where everyone has an opinion but nobody has evidence. Cavefish wants to hear where the technology is right and where it is wrong, and to build Welsh faces, voices, delivery and genres into its benchmark corpus.

Productions interested in taking part can contact jon@cavefish.co.uk.