Sensory Testing in India: Blind Taste Tests, Hedonic Scales and In-Home Use Tests

Discrimination and descriptive work use small trained groups. Liking must come from ordinary consumers. ISO 8586:2023 calls a sensory panel "a true measuring instrument" whose results depend on its members, and covers food, beverages and home and personal care products (ISO 8586:2023). The Indian standard IS 6273 (Part 1), adopted in 1971 and reaffirmed in 2012, draws the same line. It asks for a trained panel of 5 to 10 for intensity work and a semi-trained panel of about 25 to 30 for a first screen of acceptability. Liking comes from a randomly selected consumer panel whose findings rest on at least 100 independent judgements (BIS IS 6273 Part 1).

The triangle test gives each assessor three coded samples, two the same and one different, and asks which is the odd one. ISO 4120:2021 describes it as a forced-choice procedure for deciding whether a perceptible difference or similarity exists. It is statistically more efficient than the duo-trio test but "has limited use with products that exhibit strong carryover and/or lingering flavours", and applies only to homogeneous products (ISO 4120:2021 scope).

The duo-trio test gives a reference first, then two coded samples, and asks which matches it. ISO 10399 says it is less efficient than the triangle test but easier for assessors. It has a constant-reference form for when assessors know one product and a balanced-reference form for when they know neither (ISO 10399:2017 scope; a 2026 revision has replaced this edition).

How many assessors? A university sensory analysis course gives 20 to 40 for a triangle test, as few as 12 when the difference is large, and 50 to 100 to prove similarity (Ankara University, FDE443 Lesson 10). The FAO's worked example from ISO 4120 uses 12 assessors: 8 correct out of 12 calls the products different at the 5% level (FAO).

Use them for a supplier switch, a cost-down recipe or a shelf-life check, then move to consumers for liking.

A descriptive panel rates the intensity of agreed attributes: sweetness, crunch, fragrance strength, lather. ISO 13299:2016 sets out the process for establishing a sensory profile, including comparing a product with competitors (ISO 13299:2016). The DLG practice guide says the number trained "should always be one and a half to twice the final number", because some drop out (DLG Expert Report 07/2017).

Consumer hedonic tests use the 9-point scale, from 1 "dislike extremely" to 9 "like extremely", with "neither like nor dislike" in the middle. It was developed to measure the food preferences of US soldiers (Peryam and Pilgrim, 1957), and a 2023 paper calls it "the most widely used scale for consumer acceptance testing globally" (Frontiers in Food Science and Technology, 2023). ASTM E2943 covers the unbranded two-sample test on this scale, with preference by forced choice (ASTM E2943).

Consumer sample size follows the difference you need to detect, not habit. ISO 11136, the standard for hedonic tests in a controlled area, gives worked examples. Detecting a 60/40 preference split at 5% risk with 90% power needs 259 consumers. A one-sided test needs 211 consumers. Proving a new product is not worse than the old by a margin of 0.1 needs 853. For one-sided difference tests "a statistician shall be consulted" (ISO 11136:2014/Amd 1:2020). The 100-judgement floor in IS 6273 is a minimum, not a target.

Brand changes taste. In the well-known Coke and Pepsi study the two drinks are "nearly identical in chemical composition". Yet knowing the brand of one drink "had a dramatic influence on expressed behavioral preferences" (McClure et al., Neuron, 2004). A 2021 strawberry yoghurt study ran blind and informed sessions: the nationally known brand's name raised acceptance, while texture and overall impression did not differ between tests (Research, Society and Development, 2021). In our own skincare blind test, five packs scored almost the same on appeal once the logos were removed.

Test blind to improve the formula, and add a branded cell with the same questions to measure what your brand adds or costs.

Central location test (CLT) or in-home use test (IHUT)? A CLT, or hall test, controls temperature, portion and order, and collects every answer the same day. An IHUT shows how the product fits real routines. The Iowa State food product development manual lists IHUT advantages as "product usage under true consumer conditions", answers from more than one household member and "sampling of high-fatigue products over time". Its disadvantages are cost, little control over use and return rates that "may be as low as 25%". It describes an ideal IHUT as 100 consumers in three markets (Iowa State University, Food Product Development Lab Manual).

Do the two agree? A frozen pizza study with 137 CLT panellists, 46% of whom also returned home-use results, found no significant difference between locations except on crispness and tenderness. It recommended the CLT for cost and control (Agriculture and Natural Resources, 2006). For detergents, shampoos or floor cleaners, performance depends on water, surfaces and habit, so test at home.

Order changes scores. The first sample is judged on its own; the second against the first. Balance the order so every sample appears in every position an equal number of times, then randomise within that balance (Ankara University, FDE443 Lesson 7). In a triangle test the six arrangements (ABB, AAB, ABA, BAA, BBA, BAB) are each served equally often; the FAO example uses each twice across 12 assessors (FAO).

Use random three-digit codes, and avoid single letters, double digits or anything that stands for a company, a test number or the serving order. The same course gives limits on samples per sitting: eight to ten biscuits, six to eight beers. For heavy carry-over products such as smoked or spicy meats, bitter substances or greasy textures, it allows only one or two (Ankara University, Lesson 7). Spicy snacks, pickles and masalas fall in that group, so plan more respondents, not more samples per person.

IS 6273 says about 30 g is enough for a difference test, about three sips or bites. It prefers room temperature for control, but for preference tests the temperature at which people normally eat the product (BIS IS 6273 Part 1). Do not test straight after meals or coffee breaks, and test at the time of day the product is normally consumed.

Our survey routing supports monadic and sequential monadic rotation and randomisation, so the questionnaire blocks follow the tasting order you brief. The monadic and sequential monadic guide covers how that choice changes sample size.

Our platform screens, recruits and runs the questionnaire. Someone else puts the product in the respondent's hand: a venue team for a CLT, a courier or fieldwork partner for an IHUT. Panel members stay anonymous to researchers, and no address is collected or shared. Delivery needs a separate consent step that you or your partner manage. Carry the respondent's code across so answers can be matched.

Brief the placement team on:

• One sheet per product: blind code, true identity held only by the panel organiser, batch, manufacture date, storage and shelf life.
• Preparation method in full: water quantity, boil time, serving temperature, identical for every sample.
• Serving order per respondent ID, from your balanced design.
• Allergen and ingredient disclosure. The Iowa State manual requires the survey to carry allergens and the ingredient statement; put the same on the pack label.
• For an IHUT: how many days, how often, what not to mix it with, and when the questionnaire link opens.
• For a CLT: water, a plain cracker, a spit cup and a score sheet.
• A return plan for unused product.

Then give your partner the respondent IDs that passed screening, and give us the codes each respondent received. The post-tasting questionnaire goes only to those IDs.

Report the mean and the top-two-box share (8 and 9) for each hedonic question, by product, with the base. In a sequential monadic design the same people rated both products, so use a paired test. In a monadic design, use a between-groups test. The significance testing guide covers both.

For JAR questions, run penalty analysis. Collapse the five points into too little, just about right and too much. Compute the drop in mean liking for each off-target group against the just-about-right group, weighted by the share of respondents in it. The ASTM manual on JAR scales (2009, 120 pages) documents this as "penalty analysis or mean drop analysis" (ASTM MNL63). A common working rule is to act only on penalties that are significant and affect roughly 20% or more of consumers (MetricGate).

Read the open-ended answers yourself before coding. Then write one line per product: liked or not, why, what to fix, and at what price.

What we would do. We would run the first round blind, in a hall, in sequential monadic order, and act only on the JAR penalties that clear the rule above. Then take the winner home for a week in an IHUT before anyone sees a pack. If the branded cell moves the score a lot, you have learnt something about the brand, not the recipe.

If your job is the screening, recruitment and post-tasting questionnaire for a sensory test among Indian consumers, this is the pilot and main study at list price, with respondents included. Product placement is costed separately with your fieldwork partner.

Pro is sold in bundles of 400 to 1,000 responses and costs 20% less per response than Plus. Enterprise plans come with a dedicated research team and are priced on request. Screened-out respondents do not count towards a response bundle.

The SuperJ network has 2 crore Indian respondents. The survey-eligible panel is a subset, and availability depends on your targeting. Members complete Aadhaar-based KYC once, and each study receives a zero-knowledge proof of verification. Surveys run in as many Indian languages as the study needs. Quality checks flag speeders, straight-lining and inconsistent answers. Read /panel-methodology and pricing.

The limit: the app reaches smartphone users only, so offline households need face-to-face fieldwork. We do not handle the product; placement stays with you or your fieldwork partner.