TURF Analysis (Total Unduplicated Reach and Frequency)

What TURF analysis answers

TURF stands for Total Unduplicated Reach and Frequency. It was first used in media planning to choose the combination of magazines or TV programmes that reaches the most different people, and it is now a standard tool in product and portfolio research. The typical question is: if we can only offer three flavours, which three should they be? or which five SKUs should a small store carry? or which four messages should a campaign rotate?

The answer is usually not the three most popular items. Popular items often appeal to the same people: if most chocolate lovers also like vanilla, carrying both adds little new reach. TURF counts how many different respondents like at least one item in a set, so it rewards items that appeal to otherwise unserved customers. That is why it is the right tool for line optimisation, assortment planning and message selection, and why it complements preference methods such as MaxDiff and conjoint.

Key measures

• Reach: the percentage of respondents (or of the weighted sample) who are reached by at least one item in the portfolio. A respondent is reached by an item when their answer meets the threshold, for example “definitely or probably would buy” (top-2 box) on a five-point purchase-intent scale, or simply 1 on a yes/no question.
• Frequency: the average number of items in the portfolio that each respondent likes. Two portfolios with the same reach can differ in frequency; higher frequency means more choice for each customer and more chances of a sale, so the tool uses it to break ties.
• Incremental reach: how many new respondents each additional item brings in. The ladder shows diminishing returns and helps decide how many items are worth offering.

How the calculator searches

The tool converts every answer into reached / not reached with your threshold and stores each item as a bit pattern over respondents, which makes overlap calculations very fast. It then evaluates every possible combination of the chosen portfolio size (an exhaustive search) whenever this is feasible in the browser; for 10 items and portfolios of 3 that is 120 combinations, for 20 items and portfolios of 5 it is 15,504. It lists the best combinations ranked by reach and then frequency, so you can see the runner-ups and choose between sets that are practically tied on reach but differ in cost, margin or production complexity.

The tool also builds a greedy reach ladder: it starts with the single item of highest reach, then repeatedly adds the item that brings in the most new respondents. The ladder is easy to present and shows the shape of diminishing returns. For very large problems where the exhaustive search would take too long, the best combination is taken from the greedy solution and the tool tells you so. Greedy solutions are usually optimal or within a fraction of a percentage point, but not always, which is why exhaustive search is used whenever possible.

Worked example with the built-in data

The example simulates 300 consumers rating ten ice-cream flavours on a five-point purchase-intent scale, with three underlying segments (families, health-conscious and adventurous). Using the top-2 box (rating ≥ 4) as the definition of reach, the most popular single flavour is Chocolate (57% reach), followed by Vanilla (51%) and Cookie Dough (47%). But the best three-flavour line-up is Chocolate + Mango Sorbet + Coffee, which reaches 88% of respondents. Picking the three most popular flavours would reach only about 81%, because Chocolate, Vanilla and Cookie Dough all appeal to the same family segment. Mango Sorbet is the least popular single flavour, yet it is in the best portfolio because it reaches the health-conscious segment that the others miss. Notice also that the greedy ladder’s first three steps give 88.0%, slightly below the exhaustive optimum, illustrating why the full search matters.

Setting up the data

Use one row per respondent and one numeric column per item. Answers can be 0/1 (would buy / would not buy, have used / have not used) or ratings such as purchase intent (1–5), liking (1–7) or the MaxDiff scores of each respondent. Choose the threshold that defines “reached”: 1 for 0/1 data, 4 for a top-2 box on a five-point scale, 6 for the top-2 box on a seven-point scale. If your sample is not representative, add a weight column; reach and frequency then use the weights. Leave out respondents who did not answer the item battery.

Choosing the threshold and portfolio size

The threshold has a large effect on TURF results. A strict definition (top box only, “definitely would buy”) gives lower reach but focuses on strong demand; a looser one (top-2 box) gives higher reach and more ties. A common choice is the top-2 box, with a sensitivity check using the top box. For the portfolio size, run the analysis for several sizes and look at the ladder: if the fourth item adds only one or two points of reach, the line-up of three is probably the efficient choice.

Reporting TURF results

Report the number of respondents, how reach was defined, the portfolio size, the best combination with its reach and frequency, and the next best alternatives. A ladder chart is the clearest visual. A typical sentence is: “TURF analysis of ten flavours (N = 300; reached = purchase intent ≥ 4) identified Chocolate, Mango Sorbet and Coffee as the optimal three-flavour range, reaching 88.3% of respondents with an average frequency of 1.40.”

Strengths and limitations

• Strengths: intuitive, easy to explain to managers, and directly aimed at a portfolio decision. It handles overlap between items that simple popularity rankings ignore.
• Stated intent: purchase-intent questions overstate real buying; treat reach as relative, not as a sales forecast.
• Binary view of appeal: TURF ignores how much more someone likes one item than another once both pass the threshold. Frequency partly compensates, and probabilistic TURF (with MaxDiff or conjoint utilities) is an extension for advanced studies.
• Costs and constraints: TURF optimises reach only. If some items are more expensive to make or stock, use the list of near-optimal combinations and choose with the costs in mind, or force must-have items into the line-up by checking only combinations that contain them.
• Ties: with small samples many combinations have nearly the same reach; differences of one or two percentage points are usually within sampling error.

TURF together with other tools

TURF is often the last step of a product study. Use MaxDiff first to measure the appeal of a long list of items, then run TURF on each respondent’s top items to choose the line-up. Use latent class or cluster analysis to understand the segments behind the overlaps, and Gabor-Granger or Van Westendorp to price the chosen products. All of these tools are on this site and run with your data in the browser.

Key references: Miaoulis, Free and Parsons (1990), Marketing Research; Krieger and Green (2000), Journal of Marketing Research, on TURF and its extensions.