As marketers, we’re all familiar with the term “big data”–but what about “bad data”?
According to marketing experts, bad data leads to bad decisions, which could result in bad use of marketing dollars. That was the consensus of this week’s “CMO.com Wants To Know,” for which we reached out to executives for a better idea of what makes good data, what qualifies as bad data, and how to identify it.
Carol Chung, SVP of Media Technology at DigitasLBi, told CMO.com:
There absolutely is such thing as bad data–although, the way I see it, “bad” is the lower end of a very broad spectrum based on accuracy, usability, and value.
Bad data can range from the absolute “bad”–i.e. inaccurate and/or fraudulent to some more controversial qualities, like being unverified, old, reported, subjective, panel-based, probabilistic, to just simply having too much data. Some of these types of data are widely acknowledged as acceptable data points for indicative analyses, but not for definitive decisions. On the other hand, good data is fresh, actionable, accurate, and insightful.
[In terms of identifying bad data], the common approach is to get your hands on as much data as possible. I’d caution against that. Analysis paralysis is a very real thing when dealing with big data. I think of data the same way I think about technology. It’s a tool used to serve a specific purpose. The clearer you can be about the objective you’re trying to achieve or the question you’re trying to answer, the more likely you are to choose the exact tool you need. Never choose the tool before the project. Same goes for data.
Bad data is often confused with not having the right data set that you need. In terms of defending against fraudulent or inaccurate data, make sure your sources are verified, trusted, and certified.
Sidra Berman, VP of Marketing at Savi, told CMO.com:
There is such a thing as bad data, and it can cost companies billions of dollars annually. Bad data is often caused by companies falling into big data traps or biting off more than they can chew, where they focus on obtaining massive quantities of data rather than quality information that can be used to solve specific business issues or drive strategic marketing initiatives.
The data collected by companies may be fragmented, incomplete, or clouded by irrelevant information, making it difficult to glean actionable insights that can be applied within their businesses. For companies to be successful with big data, they should start with a few key data sources, ensure data quality, and implement the improvements found from analysis. Once that project is successful, additional data sources can be added to provide further operational intelligence and improvements.
Good data will be any meaningful information that provides actionable intelligence for an organization. This requires boards, CMOs, and data analysts to understand what information is important to their success and what information is irrelevant. When companies like Netflix or Amazon use historical viewing or purchasing data to make recommendations to their users on what movie to watch or what widget to buy, that data analysis is meaningful, and the result adds to their bottom line.
These organizations are successful in sorting through enormous data volumes to make company predictions and decisions because they know where value resides. This allows Amazon, for example, to sort through data on previous customer purchases, viewing habits, promotions, region, time of day, and more.
Bad data comes from multiple sources, and it is not always the data’s fault. Bad data could be due to human error, siloed data collection and management, data being in various formats (especially semistructured and unstructured), or a general lack of expertise to interpret and understand it. Businesses and marketers must first understand what business problems they are trying to address and, only then, ensure that data is normalized and handled properly to catch errors before it is used to make important decisions and waste marketing dollars.
Neda Stoll, Senior Marketing and Business Strategy Manager at Adobe (CMO.com’s parent company), told us:
Bad data does exist. With digital, there is so much data out there. As a result, marketers are challenged with figuring out which data to look at and which to use. But just because you can track something doesn't mean it is what you need to measure.
The key to identifying bad data is understanding your primary goal, first and foremost, and then building a data plan based on that. Marketers need to start with identifying their objectives because if you want to drive sales, for example, the way you look at data and measure it is different than how you would measure building awareness. A measurement plan will tell you what is important to measure, and good data will be the data that is actionable and gives you insights.
Selina Petosa, Chief Creative Strategist at Rational Interaction, told CMO.com:
Absolutely, i. In our digital age there is a vast amount of data available, but not all data is good data. IoT data, [for example], presents a huge opportunity for brands, but the data guiding decisions needs to be accurate–and the data is only as accurate as the system and sensor within the devices. Building a strategy on inaccurate data might yield questionable branding and advertising strategies.
It’s crucial to not only determine which data is accurate, but also which is relevant, as the data is only useful insofar as it leads to a value add from a branding perspective. Also, bad data can misrepresent targets for a campaign, and these skewed results are not only a costly mistake, but can also damage the brand’s reputation with the consumer.
Good data is the data that enables marketers to deliver real-time information that adds value to consumers’ lives. It’s not enough that it’s accurate–it also needs to drive business decisions and really move the needle. Wearables, for example, present an opportunity to capitalize on personalized wellness data and deliver real-time updates based on location and other timely and relevant elements of the user’s day-to-day activities. For example, a wearable might be able to determine that a user is dehydrated—which marketers know from quality/reliable location and health data—presenting the perfect moment to push a targeted ad for a nearby store that sells bottled water.
Bad data is just the opposite–it’s inaccurate or it’s irrelevant (or, worse, both). For example, if my Fitbit or Apple Watch inaccurately records my sleep patterns and then suggests that I try a new product purported to help sleep, the data and the recommendation are without merit or value.
As with any data collection, there are some key considerations to keep in mind. First, how accurate are the sensors and systems collecting the data? Have they been rigorously tested and are the results repeatable? What’s the scope of the data, and is it representative of the targeted population? Also, how secure is the data? Insecure data might be compromised and invalid.
In the age of IoT, it’s also going to be crucial to analyze the device capturing the data as well (i.e., how sophisticated/reliable is the device?). Evaluating these criteria will help marketers weed out the bad data, avoiding costly marketing mistakes, and/or abusing relations with the end user by pushing irrelevant or untimely ads.
Here's what the Twitterverse is saying about big data: