How to measure share of search

The measure is free, takes an afternoon to set up, and moves months before revenue does. Most of the difficulty is in the setup, not the maths.

Share of search is your brand’s search volume as a proportion of the total for your competitive set. That is the entire calculation. Everything that makes it useful or useless happens before you divide.

Pick the competitive set properly

Not the companies you dislike. The companies a buyer genuinely considers alongside you. Four to eight names is workable; beyond that the series gets noisy and the chart becomes decorative.

Two tests. Would a customer shortlist them in the same conversation as you? Do they sell to the same buyer with a similar price and model? If the answer to either is no, leave them out and note why, because somebody will ask next quarter.

Choose the terms

Use brand terms, not category terms. “Nike” rather than “running shoes”. Brand queries are the demand that already belongs to someone; category queries are demand nobody owns yet, and mixing them produces a number that means neither thing. The two are worth tracking separately, which we set out in branded versus category search.

Watch for collisions. Common-word brand names pick up unrelated searches, and a brand that shares a name with a film, a place or a person will look far healthier than it is. Where a name is ambiguous, use the brand plus a category word consistently across the whole set.

Pull the data

Google Trends is free and compares up to five terms at once, which is why the five-name limit tends to shape the set. It gives relative interest rather than absolute volume, which is fine, because share is a ratio anyway. For absolute numbers, a keyword tool or Search Console for your own brand will do.

Pull at least three years. Anything shorter and you cannot tell a trend from a season.

Normalise and smooth

Convert each brand to a percentage of the set total for each period, so the series is comparable even when overall category demand moves. Then apply a rolling average, usually four or twelve weeks, because weekly data in most categories is mostly noise and reading noise as signal is the fastest way to lose faith in the measure.

What to look at

• Your floor, not your peaks. Campaigns create spikes. A rising floor between spikes is the thing that compounds.
• Movement relative to the set. Everyone falling together is a category event, not your problem.
• The gap between you and the leader, tracked over years rather than months.

The mistakes that waste the effort

Changing the competitive set between reports, which makes the series meaningless. Reading a single month. Including a competitor whose brand name is also a common word. Reporting the index value rather than the share. And presenting it as a performance metric when it is a leading indicator: it tells you where demand is heading, not what happened last month. Where it belongs in reporting is covered in the dashboard that gets read.

What does Google Trends actually give you?

Not search volumes. Google Trends returns relative interest on a scale of 0 to 100, where 100 is the highest point in the window you asked for, and it is calculated from a sample of searches, so the same query can return slightly different numbers on different days. Terms with very little volume come back as zero, which means too few searches to report, not none at all.

The practical consequence is the most important rule of the method: put every brand in the competitive set into the same request. Brands compared inside one request share one scale, so their ratio is a true share. Brands pulled in separate requests are each scaled to their own peak, and dividing one by another produces a number that means nothing.

Which search terms break the measure?

Building our share of search index for India surfaced the failures that matter most, and they are all about names rather than maths:

• Common words. Dhan is a stockbroking brand and also the Hindi word for wealth, inside names such as Jan Dhan. Its raw share looked like a third of the category; we left it out rather than publish an inflated number.
• Names inside other names. A broad match on Mahindra also counts searches for Kotak Mahindra Bank, so maker names were the wrong unit for cars.
• Brands that share a word with other businesses. Bare searches for Jio and Airtel include streaming, banking and finance products, so telecom is measured on recharge searches instead.
• How people actually type a name. Searching for Swiggy Instamart undercounts a service most people call Instamart.
• Rebrands. Blinkit took that name in December 2021, so its series starts from zero. A rename is a break in the data, not a surge.

The test before trusting any term is to read the queries Google Trends lists as related to it. If they are about something else, the term is measuring something else.

A worked example

For quick commerce we put Blinkit, Zepto, Instamart and BigBasket into one request for India over five years, summed each brand's weekly values by month, and divided by the total for the set. Over the last three months of the series, Blinkit held 56.4 percent of the set's branded search, Zepto 28.0, Swiggy Instamart 10.7 and BigBasket 4.9, against 44.5 percent for BigBasket at the start. The full monthly series is in the downloadable data.