Turning horrible data into meaningful data [INSIGHT]

by Nigel Hollis

In a recent conversation, Graham Page, Head of Global Research Solutions, described chatter on social media as “horrible data”. In doing so, he was alluding to the fact that the data is inherently messy but also subject to many different influences. If you want to make sense of trends in social data, you have to be able to turn that horrible data into meaningful data.

Data handler

That is precisely what Millward Brown’s Digital Behavioral Analytics is designed to do. The system is designed to parse social and search data into key components and remove the subjectivity that is so often applied to this type of data. Critical to this process, is the application of leading-edge dynamic linear models to automatically split the raw digital data into base and short-term trends, and then decompose the latter into seasonal, campaign and event related changes.

So what sort of findings come out of this type of analysis?

•     First, the degree to which people will talk about a brand, or category, varies dramatically. A new, meaningfully different brand may well generate more than its fair share or searches and social commentary, but as the brand ages and grows the ’new’ factor will decline. Brands need to work hard to stay newsworthy.
•     Second, underlying search levels reflect changes in the interest and salience of the brand over time and so, in general, have a stronger relationship with sales performance. However, this differs by brand and category. Raw search volumes will relate well to weekly sales for an e-commerce brand, but base trend and seasonal effects will be more relevant for an automotive brand.
•     Third, brand campaigns are the number one generator of conversations about a brand (37 percent), followed by celebrities (25 percent) and news stories (17 percent). The uplift we see in social is heavily influenced by the overall quality of the creative. By contrast, search volumes show a bigger increase for ads with news in them, and we see a stronger relationship between the persuasive power of the ad and people seeking more information about the brand.

To make horrible data not just meaningful but actionable, it is important that marketers get findings quickly. An important enhancement to the analytic process has been the recent automation of the analytics, to allow weekly updates to be provided in a timely and cost effective manner.