Consumer panels and retail audits

Consumer panels and retail audits are the two methods that research companies use to measure what a whole country buys and watches. A retail audit records what a sample of shops sells. A consumer panel is a fixed group of households that record what they buy, or what they watch, week after week. The research company combines the two, scales the result up to the whole country, and sells it to brands, retailers and broadcasters, usually by subscription. A company that sells data in this way is called a data vendor.

Here is how a retail audit works. The research company signs up a sample of shops: a few thousand, chosen to stand for every kind of shop in every region, from supermarkets in cities to kirana stores in villages. Shops with billing software send their sales records from the till electronically. For small shops without billing software, an auditor visits regularly and counts the stock on the shelves. Sales are then worked out from the counts: the stock at the last visit, plus what the shop bought since, minus the stock now. Each shop in the sample is given a weight, which is the number of similar shops it stands for. Adding up the weighted sales gives an estimate for the whole country.

Here is how a consumer panel works. The research company recruits thousands of households that, together, match the country's mix of towns, villages, incomes and family sizes. Each household agrees to record its purchases, for example by scanning the barcode on each pack with a phone app. For television, a small meter fixed to the TV records which channel is on, and family members press a button to say who is watching. The households receive a small reward. Their addresses are kept secret, so that nobody can try to influence what they record.

Each method answers a different question. The retail audit says how much of each brand was sold, and where. The panel says who bought it, how often, and which brand they bought before. Matching the two sources every day, and correcting for gaps in the samples, is most of the engineering work.

Take, for example, a biscuit maker that wants to know its share of all biscuit sales in Bihar. Retail measurement gives that share each month, worked out from the shop sample.

Consider a television network deciding how much to charge for an ad slot during its evening serial. The ratings from meters in panel homes, sold as audience measurement, show how many people watched that slot last week.

Say a shampoo brand launches a ₹2 sachet. The panel shows whether the sachet's buyers are first-time shampoo users, or people who switched from a rival brand.

How is a consumer panel different from a survey? A survey asks people once what they remember buying or watching, and memory is often wrong. A panel records what the same households actually buy or watch, again and again, so changes over time can be seen. A brand's own first-party data is different again. It is exact, but it covers only that brand's own customers, not the whole market.