Two rooms
Illustrative comparison, Room A: a founder shows friendly tasting quotes and an attractive product slide. The presentation does not reveal the test protocol, comparator or recruitment method.
Illustrative comparison, Room B: a founder shows the protocol, category-user recruitment, named comparators and the limitations of the results, alongside evidence about repeat purchase. These are fictional scenarios, not observed investor outcomes.
The two rooms hold the same product. They hold different evidence.
The first founder has a deck. The second founder has a sensory data room. The decks lose. The data rooms close.
The assumption to test
The myth is that sensory work is a marketing function. Something product teams collect after the launch, the way customer service collects complaints. Something to inform the next reformulation. Something the agency runs.
This is wrong.
Sensory is not feedback you collect after launch. It is the evidence base your launch depends on. The product claim, the deck, the round, the launch order with the retailer, the partnership with the contract manufacturer - all of it sits on top of the sensory evidence. Or fails to sit on it. The category does not forgive a product that tastes wrong. The retailer does not relist a product that does not repeat. The investor does not fund a consumer thesis the founder cannot defend on the second read.
What founders treat as a marketing function is actually the load-bearing wall of the company. Five friendly quotes is not a panel. It is a moodboard. A moodboard is decoration on a wall the building is using for support.
One-minute brief
If your product claim depends on taste, texture, preference, satiety, or repeat behaviour, the company does not have a fundraising story until it has a sensory data room.
This guide proposes four evidence areas for a consumer-led product: taste against a relevant comparator, texture and structural performance, repeat purchase and sales velocity, and documented consumer language. The evidence required depends on the product and the claim.
Each row sits in one of three tiers: Anecdotal, Informal, or Blind-benchmarked.
The framework helps a reader locate the underlying protocol and recognise what remains unknown. It does not predict the outcome of diligence or fundraising.
Founders - The moodboard tells you what buyers say in public. The data room tells you what they do in private. Funds buy the second.
Investors - Ask one question pre-investment: show me the lowest-tier evidence. Funds that do not ask end up funding consumer theses that collapse on retest.
Innovators - Sensory is the evidence base your R&D decisions depend on. Texture-analyser data without a paired descriptive panel is half a measurement.
What the evidence needs to show
If your product claim depends on taste, you do not have a fundraising story. You have a deck.
These four areas answer different questions. A tasting does not measure repeat purchase, and instrumental texture does not establish consumer preference. Prioritise the evidence needed for your next decision.
Blind-tested taste vs incumbent
Commission an appropriate consumer study with a named comparator, relevant category users and a disclosed protocol. Define sample size through the study objective, expected effect and variability. ISO 11136 provides general guidance for controlled-area hedonic testing; it does not establish a universal minimum of 150 participants.
Founder tasting and trade-fair feedback can guide exploration. They do not establish blinded consumer preference or replace a suitably designed study.
Texture / structural performance
Instrumental measurement (texture analyser, force-deformation curves, shear stress) paired with sensory descriptive analysis (a trained panel rating firmness, fibrousness, juiciness, springiness on a defined scale). Either one alone is half a measurement. Texture analyser data without a paired descriptive panel is force in newtons, divorced from human perception.
A descriptive panel without instrumental data is human perception, divorced from the physical thing being measured. The pairing is what makes the data useful. A reformulation that improves instrumental firmness but loses fibrous integrity in consumer language is a regression. The founder will not see this without the pairing.
Repeat purchase and sales velocity
Repeat-purchase and velocity data help test behaviour beyond a tasting. Choose the observation window to match the product’s purchase cycle and disclose channel, availability, promotion and price. Ninety days is an illustrative window, not a universal sensory standard. Repeat purchase is influenced by more than taste.
A tracking panel can provide category context if the sample, channel, period and definitions match the question. D2C reorder data can also be useful within its limits. Neither a named provider nor a data subscription guarantees comparability.
Qualitative texture vocabulary
Document consumer language alongside any instrumental and descriptive measurements. An illustrative reformulation record might compare versions 4.1 and 4.2 using the same scale and protocol, then examine whether consumer descriptions change. A sensory score does not measure a percentage change in Maillard chemistry.
Choose enough qualitative material to support the question and analysis. Document recruitment, coding, product versions and uncertainty. A fixed 80-100-verbatim threshold is not established here.
You have just read the four evidence types. The benchmarking, the category-specific evidence, and the deck-use rules are below.
What to benchmark against
A benchmark without an incumbent is a self-portrait.
Sensory data without a named comparator is decorative. It tells you the founder ran a test. It does not tell you whether the product is competitive. The two questions that change the conversation are: which incumbent did you compare against, and how did it score?
NECTAR’s Taste of the Industry research is one publicly described benchmark to investigate. Read the product categories, participant recruitment, protocol and comparators before applying its results to your own product.
Compare years only where products, recruitment, comparators and methods support that comparison. A change in category coverage alone does not measure a closing taste gap.
Choose comparators that match the category and eating occasion. A weak or unnamed comparator cannot substantiate a category-leading claim. The required design depends on the claim; a fixed number of comparators is not a universal rule.
What founders should stop calling a benchmark:
Founder palate against the founder’s own previous version.
Trade-fair tastings against the founder’s own previous version.
A single influencer or chef saying the product is “better than expected”.
A blind test against an unnamed competitor (the test is not falsifiable).
A blind test against a deliberately weak competitor (the test is bad-faith and reads that way).
What investors increasingly expect:
A named incumbent, ideally the category dollar leader.
A justified sample size and recruitment method suited to the target consumer and question.
The eating occasion held constant (a salmon fillet portion is benchmarked against farmed Atlantic salmon, not against canned tuna).
Where useful, replication with a further cohort and documented product version; timing should fit the research question.
How to use sensory data in a deck without overclaiming
Texture is where the category breaks. Vocabulary is where you prove you understand why. The same line applies to deck claims. The vocabulary the founder uses about their own product is itself an evidence signal.
Five claim levels appear in consumer-thesis decks. Each requires a different evidence tier to defend.
Claim level 1: some participants liked the product. Report who participated, what they tried and how feedback was collected. Anecdotes can suggest questions; they do not substantiate a broad claim about consumer preference.
Claim level 2: target consumers prefer our product to a named comparator. Use an appropriately designed comparison, with a justified sample, protocol and analysis. Informal positive feedback alone does not establish this claim.
A parity claim needs a design capable of testing the stated comparison. ASTM E1958 concerns sensory claim substantiation; mentioning a standard does not establish compliance with advertising or consumer law in any jurisdiction.
Claim level 4: our product leads the category on a specific dimension. Define that dimension and the relevant comparator set. A result against one product does not establish category leadership. Check the study design and scope before making the claim.
Claim level 5: our texture system is a defensible technical advantage. Consumer results alone cannot establish a competitive barrier. Examine reproducibility, relevant know-how or IP, competing products and the ability to maintain performance. This is a discussion framework, not a validated proof of defensibility.
A conservative editorial rule is to keep claims within the design and strength of the evidence. The five claim levels below are a discussion framework, not a legal safe harbour or a validated conversion scale.
The trap is that founders make claims at the level the strongest row of the data room would support. Investors evaluate claims at the level the weakest row of the data room would support. The gap is the diligence risk.
For R&D teams
Sensory work is no longer a marketing function. It is the evidence base your R&D decisions depend on, and it is the evidence base your CEO is going to need in front of investors whether or not it is built into the R&D budget.
Texture is where the category breaks. Vocabulary is where you prove you understand why. Texture analyser data without a paired sensory descriptive panel is half a measurement. A reformulation that improves instrumental firmness but loses fibrous integrity in consumer language is a regression dressed up as progress. Build the descriptive lexicon early. Track it across versions. Make the trade-offs visible to the CEO before they become visible to a category leader doing diligence in a meeting your CEO is sitting in.
The R&D-side prescription. Bake the descriptive panel into the formulation gate. No version ships to a customer-facing demo, a trade fair, a retail buyer, or a deck without a descriptive panel sign-off. The cost per version is small. The cost of shipping a version that scores well on the texture analyser and badly in the consumer mouth is large. The asymmetry is on the side of running the panel.
A second prescription. Document the lexicon as a living document, not a one-off deliverable. A category lexicon developed once and never updated is a snapshot. A lexicon updated every reformulation version is a system. Investors who look at R&D processes (and they increasingly do, because Issue 25 was right that Plant-Based 2.0 is an R&D-first cycle) read the difference instantly.
For founders
Two failure modes to name explicitly because both happen quietly and both lose rounds. The moodboard substitute. Founders pulling consumer quotes from Instagram and treating them as evidence in the deck. The moodboard tells you what your buyers say in public. The data room tells you what they do in private. Funds buy the second. The first is a marketing exercise dressed up as a diligence exercise; the difference is visible to any investor who has seen a credible data room before.
The single-source limitation: a retailer or D2C dataset can describe behaviour in that channel. Disclose recruitment, availability, price, promotion and the observation window. Compare it with category data only where definitions and populations are sufficiently compatible.
Scope a sensory study against the decision, relevant comparators and research design. Obtain a written quotation and timeline from the provider. No fixed budget or duration guarantees a complete data room or a funding outcome.
A second prescription. The Stress Test version of this is to reverse the audit. Pretend you are the lead investor. Read your own deck. Walk every consumer claim back to the row of the matrix that supports it. Where the support is Anecdotal, ask yourself: would I fund this from someone else? The honest answer determines whether the deck is ready or whether the data room is.
For investors
Investor confidence is bounded by the lowest tier of evidence in your sensory data room. The diligence question that separates serious investors from pattern-matchers in 2026 is one sentence: “Show me the lowest-tier evidence in your sensory data room.” Funds that ask the question filter out moodboard-grade decks before management presentations. Funds that do not ask end up funding companies whose consumer thesis collapses on the first independent retest.
Two patterns to flag in your own portfolio. The first: founders who upgrade the deck without upgrading the data room. New slides; same underlying anecdotal evidence. The deck is downstream of the data room. Fix the upstream first. A deck refresh that does not include a data room refresh is a tell.
The second: funds that accept “we’ll run a panel post-close” too easily. Closing conditions are a blunt instrument. By the time the panel runs, the cap table is locked. The downside protection the term sheet implies is weaker than the term sheet suggests. The scenario the closing condition is meant to protect against is a scenario you are no longer protected against once you have signed.
A constructive note. The Plant-Based 2.0 cycle (Issue 25) will be won by R&D-first companies. Sensory data rooms are the documentary evidence that an R&D-first company has done the work. Use the audit matrix as a triage tool: companies whose data room maps to four rows of Blind-benchmarked are doing the work and have priced in the cost. Companies whose data room maps to two rows of Anecdotal have not, and their announced raise size is implicitly subsidising the gap.
Three diagnostic questions
What is the highest-confidence claim in your fundraising deck, and which row of the audit matrix supports it?
If an investor asked your sensory firm to share the raw protocol tomorrow, would you be willing to send the email?
What is the lowest-tier evidence in your data room, and would you accept it from a company you were funding?
Sources and scope
This is an edited adaptation of the author’s full Substack essay, first published 2026-04-28. Examples and editorial frameworks are distinguished from measured evidence. Source checks and editing: 16 October 2026.
Read the original Substack version. Read the FoodEdge editorial standards or suggest a correction.
Cover photograph: Photo by Daria Nepriakhina 🇺🇦 / Unsplash. Context image, not a photograph of a company or experiment discussed here. Photo licence.
