New research in marketing
This article is part of a series of articles from the WARC Guide to new research in marketing.
There is growing confusion around how marketers can apply ‘share of search’; Shann Biglione of Zenith USA sets the record straight
In his latest column, Mark Ritson makes the case for replacing share of voice with share of search. While share of search (SOS) plays an increasingly important role in tracking the health of communications, presenting it as an alternative to share of voice (SOV) is a mistake.
How does share of voice inform marketers, and why is it a problem metric?
As retold by Mark, one of the big developments in budget planning was what is referred to as the “Jones curve”, named after John Philip Jones, who showed the curved relationship between SOV and market share. Basically, because of the correlation between the two, a brand could look at its current share of market and share of voice to estimate whether it is over or underspending versus its competitors. SOV, and more recently Binet and Field’s popularization of Excess Share of Voice (ESOV), have been staples of the media industry. But Mark knows his onions of course and is absolutely right to point out they’re running into a big problem: SOV has become really hard to estimate.
Back in the days when a brand would be buying mostly mass media, one only needed a tracking company to check whether an ad appeared or not. The reporting allowed marketers to get a good sense of how much media weight competitors are applying to a market, and further channel by channel analysis offered an indicative look under their strategic hood. To this day, asking for a competitive report remains one of the most common requests brands make to their media agencies. As with every data run, the quality of the analysis relies partly on the abilities of the analyst, but mostly on the reliability of the data.
The big issue is that media is increasingly bought on channels and platforms where advertising is a lot harder to track. Because a lot of media formats are bought on an audience/impression basis and not on a placement one, trackers are hard-pressed to see all the ads, let alone the volume of ads that are being deployed. Entire platforms (usually the walled gardens) are not tracker/spider-friendly. It’s easy for a system to see an ad was shown during Saturday Night Live’s broadcast, but it’s a lot harder to identify how big a campaign that ran only on Facebook really was. Even big formats like a YouTube masthead can be a problem as they are increasingly sold on an audience basis, making it hard for the tracker to know how many people have really seen it.
Bluntly, the tracking of digital spend is a bit of a shitshow, often leaving share of voice reports with big disclaimers that digital is very likely underreported, if at all. This was fine when digital spending represented less than 10% of the total, but is now a big headache for analysts and marketers relying on that data. Ask a media team for an SOV analysis and watch their eyes roll in despair.
Looking for an alternative is by no means a waste of time. However, what is critical to understand at this stage is that share of voice is an input that marketers have strategic and operational control over. The beauty of measuring SOV is that yes, you can use it to estimate your ranking and it correlates pretty well with your market share but, more importantly, you can affect this ranking almost immediately by just opening or tightening your media budget. If your share of voice is inadequate, you can decide to fix it. Share of voice has this ability to serve as a marketing thermostat: it lets you measure and control the heat.
Why is share of search interesting… and why can’t it replace share of voice?
The idea here is not to downplay the worth of share of search. The concept itself is fairly self-evident: by comparing how much my brand is searched relative to other brands, I may be able to get a decent sense of my position in the market. And yes, there often is a correlation between your share of search and your share of market. (If you want to find out more about the reliability and practicalities of share of search, I highly recommend reading James Hankins’ practical demonstration of how to estimate it, and how to use it.)
SOV is a measure of input. A thermostat. But when it comes to your share of search, you’re measuring an outcome. It’s a thermometer.
That correlation isn’t the only arousing part for marketers. Share of search analysis, when set up properly, doesn’t just stop at estimating the current attention your brand commands. More advanced solutions looking at specific queries allow you to better understand how the category is searched and how you rank within those behaviors – for example, are people looking to compare products or trying to figure out how to use the products? This can really help prioritize the kind of messages and investments a brand should focus on. In trend-based categories like fashion or beauty, this can be a very powerful weapon to adjust your comms strategy, and why performance media often plays a central role in their overall planning process. Personally, I’m a fan, and I welcome seeing more brand dashboards paying attention to it.
However, there is one big problem and a few realities when it comes to Mark’s suggestion. The biggest hurdle – and the one that discredits the idea most - is that mentioned above: SOV is a measure of input. A thermostat. But when it comes to your share of search, you’re measuring an outcome. It’s a thermometer, and you can’t really use it to turn up the heat.
You can decide to change your share of voice immediately, but your share of search is not something you directly control. So of course, it correlates well with market share. It’s basically a proxy measure for it! But correlating with market share isn’t the only function a replacement to share of voice needs to serve. We would also need a measure that is directly controlled by a brand’s marketing investments.
On top of that, people who use share of search extensively know that it also has important caveats.
The first reaction is often that not every category is searched the same way – you’re more likely to search for a car than chewing gum. It is definitely true that not every category is equal in its ability to extract representative data. But it’s actually more reliable than one might assume, and it’s always worth taking a shot at studying that elasticity.
No, some of the more insidious issues are found under the hood, with three in particular:
- As a metric, SOS needs to deal with the fact it captures a lot of movements that are not controlled by the brand. A common occurrence, for example, is a celebrity making a statement about said product in the news. One could argue this gives you an indication of how visible your brand truly is, but it goes to show how share of search is far from being a proxy for your marketing choices.
- The way you categorize keywords can insert a strong bias too. All searches are not equal and fine-tuning which keywords define a brand search can be tricky. You also run into issues when the brand is either a common word (e.g. Purple for mattresses), or when people tend to associate a brand with a category (e.g. Chapstick for all lip balms or Kleenex for tissues). This can taint the data, and in some cases you’ll need to find proxies to make things work.
- Brand and product search is going through its own fragmentation. While Google remains a valuable place to look (and certainly a decent sample size), one now has to question whether ranking Amazon searches isn’t as good if not better. I’ve seen plenty of situations where the report can look quite different from one platform to the other, especially when they serve different purposes. This can be mitigated by analysing which ones correlate better with sales, but is important to consider as sometimes you may want to model around both, and how search behaviors can change.
All this is to say that share of search certainly has promise, but it also has issues. It’s understandable that it gets marketers excited – it really should. But it’s disappointing to see an academic of Mark’s influence jump into it without cautioning about what search analysis is and does.
Interestingly, beyond the industry finding a collaborative way to fix it, the solution to the SOV problem might actually be about… moving entirely away from it. In many cases, SOV analysis can have inflationary effects on categories’ media budgets. A bit like an arms race, obsessing about how big your competitor’s budget is can easily drift categories towards unhealthy levels of spend. And ultimately, taking bottom-up sufficiency approaches to estimate what your media weight should be might be the best outcome for both the effectiveness and efficiency of your brand’s investment. One share of search will even play a role in!
After all, we’re an industry that prides itself in being consumer-centric. So how about we spend more time and effort building plans based on the consumers?
Read more articles from the WARC Guide to new research in marketing.
Decoding decisions: Making sense of the messy middle
Jonny Protheroe and Sian Davies
Moving to a positive attention economy with Attention Adjusted Net Reach
Professor Karen Nelson-Field
The true cost of advertising attention
James Hurman with Peter Field
Using neuroscience to make the case for more diversity in advertising
Chasson Gracie, Joshua Richards, Daniel Pozzebon, Kevin Keane and Jasmin Amin
Outcome-based marketing v2.0: Profitable growth by targeting consumers in the movable middle
Mobile Marketing Association and Neustar