Neodrop
Create
FeedDiscoverMy ChannelsAuto-publishJoin Our Discord
FeedDiscoverCreateMy Channels
2026-10-25

Deep Reads: The Room, the Shortlist, the Shelf

Deep Reads Weekly
Deep Reads Weekly@Ana Askari
Deep Reads: The Room, the Shortlist, the Shelf

A snack brand has ten claims for the front of a new pack, and it puts all ten into a survey. A few hundred category buyers read each one as a line of text and pick the ones they find most appealing. The protein claim wins by a comfortable margin and goes to the designer.

0:00 / 10:07
This week's three reads were published between 19 and 23 October 2026, and none of them is about a better way to analyse data. Each one is about a decision taken before the analysis starts: which room the interview happens in, which suppliers the brief gets sent to, and which surface the claim is tested on. Underneath all three sits one question: what does the setting decide, before the study has even begun?

The room

Katrina Noelle runs KNow Research and still operates an in-person research facility in San Francisco. Crispin Beale interviewed her for Research World's Insight250 series on 20 October. 1
She had just come back from POSSIBLE in Miami, where the AI stage was full of people who had built genuinely impressive technology. What struck her was where the conversation kept landing: on what people did, rarely on why. Her phrasing is worth keeping. "All of it telling you the score. None of it telling you the game." 1
The example that comes up is the Super Bowl. Hootology's research found that when people remember an ad, only 16% say it was because of a celebrity, against 61% of that year's commercials featuring celebrity talent. Noelle reads that as a listening problem rather than a celebrity problem: somewhere between the brief and the buy, the customer left the conversation. 1
Her answer to when a room changes the answer is wherever the cost of being wrong is high — new product launches, brand repositioning, anything touching health, finance or a major life decision, and any audience that is complex, emotionally engaged or historically underrepresented in research. In a room, she argues, people contradict themselves, hesitate, laugh at the wrong moment, and say one thing while their face says another. A skilled moderator catches all of it. On a screen, most of it goes missing.
She is careful with the other side, and it is the part most advocates drop. Digital research has genuinely opened doors for people who cannot travel, who have mobility problems, or who live far from a research facility. But the default has a sample of its own: online panels lean towards people with time, a reliable connection and no discomfort with technology, and treating that as sufficient is a decision about whose voices count.
The line to underline is what she says to a client who has been told they can get the same output for a fraction of the cost. She asks one question: how will you know if it is wrong? With real research you have a method you can defend; with synthetic output you have fluency. Her advice for the budget meeting is about language — researchers argue validity and depth, which lands with other researchers, while the person holding the money answers a different question: what does being wrong cost us? 1
The boundary to keep. This is an interview with the president of a firm that sells in-person research, so her position and her business point the same way. The Super Bowl figures are Hootology's, quoted inside the interview, and the interview gives no design behind them. What survives that is the question about what being wrong costs, not the invoice.

The shortlist

Maria Butylina runs SampleMine, which supplies sample to research teams. Her piece went up on Greenbook on 23 October. 2
She starts in a Turkish restaurant her team loved, with a menu so long it came as a small book. Everyone said the choice was enormous and everyone ordered the same few dishes. The parallel is deliberate: a brief lands on top of five others, the clock is already running, and the supplier whose name arrives first gets the work.
That places the choice inside the data-quality conversation, at the earliest possible point. Fraud checks, cleaning and weighting all happen after a respondent is in a survey. The supplier was chosen before any of it, and that choice decides what kind of sample is even possible — which is how Greenbook's editor frames the piece. 2
Butylina's explanation for why the shortlist narrows is not laziness. What a team knows about suppliers is scattered across heads, inboxes and spreadsheets: somebody remembers who delivered an impossible quota three years ago, someone else knows which provider was unusually strong in a particular market. Unless that memory is easy to find and compare, the practical choice is the name that comes to mind.
Her evidence is two projects she says her agency did not win. One came back with the feedback that the price was too high, and she points out that a price is the result of a dozen decisions — this one had been built on her trusted suppliers' cautious assumptions about a difficult audience. The other was lost to a competitor who sounded more confident about several difficult quotas. She says plainly that she cannot know whether a different supplier would have changed either result.
The questions she leaves behind are small enough to use: what is genuinely difficult about this brief, which suppliers have proved strong at exactly that, and are we choosing on what we know or on which names were easiest to remember.
The boundary to keep. This is an argument from the founder of a sample supplier, made without data, and her commercial interest runs towards longer shortlists. The two cases are her own recollections, which she says cannot be settled. The part worth carrying off is the framing rather than the evidence: the shortlist sets the ceiling on what a study can find.

The shelf

Thibault Cousot is chief product and technology officer at Behaviorally. His piece is Greenbook partner content from 19 October, and it is openly about a tool his company sells. 3
He opens with a snack brand that puts ten candidate claims into a survey, where a few hundred category buyers read each one as a line of text and pick the ones they find most and least appealing. The protein claim wins by a comfortable margin and goes to the designer. On the pack, that claim will compete with the logo, the flavour name and the photography for a glance that lab research puts somewhere between half a second and two seconds — and adding a claim changes how everything else on the pack gets read. A line that wins as a sentence can make the pack more persuasive and, at the same time, make the flavour name harder to find. A ranked list cannot show you that trade. 3
The workflow his team built runs the test the other way round: start from the pack and let the claim be the only thing that moves. Generative AI writes versions in the brand's voice, some factual and some aspirational, and places each onto the real design, matched to the typeface and layout, while everything else on the pack stays fixed. Ten rendered packs differ by a single line, each of which looks like something that could go to print.
Predictive models score them in four steps, in the order a shopper meets the pack: Seen, Shoppable, Seductive, Selected. Seen comes first because a claim nobody notices is not a claim. One eye-tracking study he cites, run on granola bar concepts, found that the more attention people paid to a claim, the more likely they were to choose the product carrying it. The score comes back with the trade broken out, so a gain on Seductive against a loss on Shoppable is visible to the team deciding whether the trade is worth making. 3
He also makes a claim about the data behind the models: a record of how shoppers behave in front of packs, tied to what those packs went on to sell, built from 75,000 packs tested to date.
The boundary to keep. This is partner content for Claims.AI, the product the author is describing, so the argument and the product point the same way. The design argument stands on its own. The accuracy of the predictions is not demonstrated: nothing in the piece reports how closely a predicted score matched what a pack actually sold. His last line is the one worth keeping, and it applies to anyone testing anything: every claim gets judged on the pack in the end, by a shopper with a second or two to spare, and the only choice a brand has is whether it sees that judgment before or after the print run.

What to take into the week

Three reads, three settings, and the same thing happens in each. The setting gets picked for a good reason — it is faster, cheaper or easier to sell — and then it quietly sets what the study is able to see.
The room decides what a person will admit and what a moderator can notice. The shortlist decides which supplier could ever have solved the problem. The shelf decides whether a winning sentence survives contact with a shopper.
None of the three argues that the convenient choice is wrong. Noelle uses digital research constantly. Butylina still orders the baked aubergine. Cousot lets people decide what fits the brand and what legal can support. What they argue is that the default decides for you unless you look at it.
So the check for the week is one question, for the next piece of research that lands on you or for the brief you are about to send out: where will this be judged, and was it judged there?
If the claim will be read on a shelf, the test has to involve a shelf. If the finding is going to change a launch, somebody had to hear it in a room where people could hesitate. If the sample sets the ceiling, somebody had to look past the first name that came to mind. And if the decision will be made somewhere the research never went, you already know what you are holding: a guess with a method attached.

References

  1. 1
    Why In-Person Research Still Matters — Research World

    researchworld.com

  2. 2
    Why We Keep Choosing the Same Suppliers — Greenbook

    greenbook.org

  3. 3
    Your Best-Tested Claim Might Be the One Nobody Reads — Greenbook

    greenbook.org

Deep Reads Weekly
Deep Reads Weekly@Ana Askari

This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.

Related content