TURF Analysis in India: How to Choose the Portfolio That Reaches the Most People

Sawtooth Software's webinar page says TURF stands for Total Unduplicated Reach and Frequency and is used to identify a subset of messages or items to offer that will reach the greatest number of consumers possible.

Two measures do the work. Reach is the share of respondents who would pick at least one item in the set. Frequency is the average number of items in the set that each respondent would pick, counting everyone reached. A set can have high reach and low frequency, or the reverse. Reach is the usual decision measure.

The word 'unduplicated' is the key. A person who likes three items in your set is counted once for reach.

TURF fits when you can offer only some of the items, and each person needs only one of them to be reached. It does not fit when people need everything in the set, such as the parts of a single product. Use concept testing when you are comparing whole ideas and not building a portfolio.

TURF needs one yes or no per person per item: would this person buy, use or choose it? There are two ways to collect that.

The first is a multi-select question: 'Which of these would you buy?', with every item shown. Our survey builder documents multi-select questions with minimum and maximum selection rules. Set a cap, such as 'pick up to three', if you worry about yes-saying. Our design notes say Indian respondents show elevated yes-saying, so a cap forces them to choose. The trade-off is that a cap understates reach for people who like many items, so state the cap when you report.

The second is a MaxDiff exercise, which gives a ranked list of preference. Use MaxDiff when you have many items, roughly 8 to 30, and need a preference order first. The formula list for our analytics engine includes TURF for multi-select questions and greedy TURF reach curves for MaxDiff results.

Keep the item list to things you could actually make. Show images for flavours or packs, since our builder supports image, video and audio stimuli. Randomise the order of items and use the same wording and length for each, so no item gains from its description.

Pick a sample large enough for the answer you need. Reach is a proportion. If a set reaches 90% of 400 respondents, the 95% margin is about 2.9 points. At 200 respondents it is about 4.2 points. Work out the margin before you compare two close sets (see significance testing).

The numbers below are made up to show the arithmetic. Ten respondents each said which of five chip flavours they would buy. Real studies need far more people.

The two most popular flavours are Mango and Masala. Between them they reach respondents 1 to 6, so six people. Reach is 6 out of 10, or 60%. Frequency among those reached is 11 picks across 6 people, about 1.83. All five Masala fans already like Mango.

Now compare other pairs. Mango and Lemon reach respondents 1 to 6 and 7 to 9, so nine people: 90% reach, with every person picking one flavour. Mango and Guava reach eight people. Masala and Lemon reach eight.

The best pair beats the most popular pair by 30 points. A third flavour, Guava, picks up respondent 10 and takes reach to 100%. That is the logic of a greedy build: start with the item that reaches most, then add the item that brings the most new people. Here greedy reaches the best answer at each step. It can miss the best set in other data, so check close calls by comparing all combinations when the item list is short.

TURF counts stated interest. People say they would buy more than they do, so treat reach as a ceiling, not a forecast.

It counts a person as reached even if they like one item only slightly. It also ignores volume, margin and the cost of making each item. A flavour that adds 5 points of reach but doubles complexity may not be worth it.

It is silent on price. Run a Gabor-Granger or Van Westendorp study for that. And the answer depends on the items you listed: leave out a strong option and TURF cannot find it.

Report sets with similar reach together. If two sets are within the margin of error, choose on cost, fit with the brand and ease of supply.

If your job is a portfolio survey of smartphone-using Indian consumers, TURF is on the platform's list of research designs, so you can build the question, set the quotas and read reach in the results. This is what a 400-response study costs at list price.

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. SuperJ has a 25M+ consumer footprint in India only. The survey-eligible panel is a subset, so availability depends on targeting. Members complete Aadhaar-based KYC once, and each study gets a zero-knowledge proof, not identity documents. Speeders, straight-lining and inconsistent answers are removed before analysis. See panel methodology, Poseidon and pricing. The limit: smartphone users only.