Prof. Mark Edelman invited me to present a lecture and workshop on listening for advertising campaign development to his undergraduate students. In the 60 minutes allotted, I gave a quick rundown of listening – how it came about, what it is, how it’s done, and then presented case examples of listening research and listening used to achieve brand objectives. Then we went into the workshop portion.
The students are organized into agency teams, with each one’s final project being a pitch to win the Kingsford Charcoal account. After supplying them with URLs to a few listening tools – Google Insights for Search, Twitter Search, Social Mention and Ice Rocket, they had at it. In just a few minutes they were searching on key words and getting experience. I love these sessions because of the excitement that comes from discovering something new. They explored for about 10 minutes. I was called over to one of them, the others were on their own. Then I had the agencies report.
What’s always fascinating is how groups approach the problem. The un-coached ones used brand or category terms, Kingsford Charcoal, Royal Oak, charcoal, gas grill, etc. Two things were interesting: 1) as “advertising pros” their keywords were driven directly by advertising concerns, and 2) they experienced the data quality problems that arise from not disambiguating their terms. These are two of the most common problems I’ve seen in companies or groups that do listening.
Let’s take a look at the disambiguation issue first. One of Kingsford’s name brand competitors is “Royal Oak.” The students assumed that the brand name was used only for charcoal and searched on it, but were wholly unaware that Royal Oak was the name of a town in Michigan and used in a variety of ways. When Google Insights returned results, the search volumes for Royal Oak and Kingsford were opposite to their knowledge: Royal Oak was way on top. After showing the “agencies” how dirty the data was, they quickly made some changes that cleaned the data and order was restored.
Failure to disambiguate is one factor that contributes to poor listening results. If the students overlayed sentiment data on the faulty search, any conclusions drawn would be riddled with errors. Sadly, this occurs in real-life listening projects and, I believe, is one of the reasons for the failure of listening initiatives. Disambiguating can be very difficult, especially when the terms of interest, such as a brand name, or also in common use. MotiveQuest says that it took them over 250 search arguments to separate out Visa the brand name from visa the travel document for research they did during the Beijing Olympics. I have had experienced listening pros tell me that they do not always go as far as they should to disambiguate terms because of the complexity and time involved.
Now on to the search terms themselves. It’s all too common that people doing listening work use the keywords that matter to them, not necessarily searching on keywords that people speak and write. This is insufficient because people often don’t mention brand names very often. in many categories. Among foods it’s 5% or less. Searching on brand names doesn’t necessarily surface enough conversations for analysis. It’s important to broaden the search and capture more conversations by using terms that capture the category, often these are features or benefits. This is helpful. But the real gold is exploring ways consumers incorporate the brand into their lives and using keywords that reflect their language. Once students did that, “charcoal grilling for picnics,” “charcoal grill recipes,” etc. they started seeing different types of results that then spurred them to look further. In the debrief, the group that searched on the “brand in people’s lives” had richer observations and a trove of ideas to pursue compared to the others, who merely compared trends lines for “charcoal grill vs gas grill” and similar. Everyone learned this important lesson.
Takeaways: 1) put the right level of effort into disambiguation to increase the chance of getting quality results, and 2) use consumer language that captures what people do with the brand, not necessarily what the brand wants consumers to do.