Last week, I went to 2 separate events on big data analytics; 1 sponsored by the AMA and another by a tech meetup group. I came away disappointed and convinced that businesses are running their big data analytics like a bad CRM program.
Here are the key takeaways from today’s discussion on the dangers of big data:
Having a lot of data doesn’t necessarily improve marketing
Why building customer relationships outweighs remarketing
How to go from big data to insights
Data visualizations improve understanding
Art and science of marketing: Mixing the right potion
Big data analytics and marketing
Big data refers to the ever-increasing volume, velocity, variety, variability and complexity of information. For marketing organizations, big data is the fundamental consequence of the new marketing landscape, born from the digital world we now live in. [source]
Just having a lot of data doesn’t inherently improve your marketing efforts. Instead, consider this:
All information has a cost, even though we’ve seen a sharp drop in the cost of storing data. So, consider carefully what data to store, which data can be stored offline (where storage costs are significantly lower), and which data should never be stored (and maybe never be collected). Just because you can, doesn’t mean you should collect all data.
Gathering person information invades the privacy of users and many people feel uncomfortable giving up this information. Instead, they’ll avoid using your site, accessing your assets, maybe even develop a negative attitude toward your brand. Seriously consider whether the value of collecting personal data outweighs these costs.
Building customer relationships
Big data creates a host of opportunities for marketers. Unfortunately, based on presentations last week by some of the industry leaders, IBM and Oracle, big data only means “Big Brother”.
These industry leaders, as well as many others, advocate using big data to stalk users online so you can sell them something. This approach views big data of an extension of CRM programs that invade users’ privacy without providing any benefit to the user.
Sure, tagging user identity to behavior provides great marketing benefits. Knowing that a particular user visited a page on your site, viewed one of your ads, or made certain comments on Facebook, is really valuable information. It helps the organization qualify leads, turn off problems for complaint handling, or serve up appropriate products. Amazon and Netflix are the leaders in turning these big data analytics into customer value.
And, that’s really the key — they create customer value by tracking online activity to a specific user. Handing over personally identifiable information (PII) is a tit-for-tat arrangement where the user gets some concrete value. When a firm uses that PII for lead generation, the firm sees all the benefit and the user sees a cost in terms of getting sales materials they didn’t request.
Building customer relationships should be your goal. Using big data to create opportunities for customers to find and buy your product outweighs efforts to force feed your product to prospects.
Going from big data to big insights
Here are great ways to use big data analytics to improve your market performance:
Identify customer segments
In the past, brands developed relatively anemic target markets based on vague notions of what types of consumers might find their products useful. Later, they built customer personas that were richer, including psychographic and behavioral elements. Now, firms can use big data to tell them exactly what their customers look like.
Even without PII, you can collect data from cookies, customer surveys, and feedback forms to help you gather information about who’s buying your brand or who’s visiting your digital properties. Then, it’s a simple matter to feed the information into a statistical package like SAS or SPSS to generate clusters that conform to segments within your user base.
Now, you can create highly targeted messages designed to hit the hot buttons of each segment, which increases conversion dramatically.
A new use of big data analytics is identifying customer lifetime value (CLV) which, put simply, is the value (sales-costs) of a customer over their lifetime (as your customer). Customers who purchase more, have a higher CLV, for example, but if they make a lot of returns (cost) or require a lot of special attention (cost) their CLV might be lower than someone who doesn’t purchase as much. A young customer might have a higher CLV despite small purchases because they might reasonably continue purchasing your brand for a long time. Some customers represent such a small (or negative) CLV, it’s not worth marketing to them.
Optimize the customer experience
Tools like Google Analytics, as well as dashboards provided by other social networks, give you information that helps manage and optimize the customer experience.
As an example, look at this flow diagram from Google Analytics. You determine how visitors move through your site, identifying what content is most productive in moving visitors toward purchase, especially by segment. In this case, this isn’t an e-commerce site, but you could easily see a situation where you track things like shopping cart abandonment to identify whether certain segments (you can change the default channel grouping to accomplish this) or certain pathways contribute or more (or less) to abandonment.
A new feature in Google Analytics, called cohort analysis, further helps you optimize results using big data.
Optimize marketing efforts
Offers: Brands often offer incentives to drive sales, but not all offers are created equal. For instance, I once did a mail program where we offered a relatively cheap ($5) product as an incentive. Through A/B testing we determined that this particular offer was more appealing to our target audience than other possible offers.
Messages: With traditional advertising, we have little opportunity to fine tune our messages. It’s kinda like throwing spaghetti at the wall – some sticks and some doesn’t. We have no clue what makes some stick over others.
Big data helps overcome this problem because we now KNOW what’s sticking. For example, if we create a post of Facebook or an ad there, we KNOW if it performed better (or worse) than others. That way, we can create content similar to what works best (BTW, sometimes the market performance of a piece of content is function of timing (time of day; day of week; how many times you shared the message) than the absolute message so you should monitor this factor as well in optimizing messaging.
In-store marketing: beacons are the newest tool for generating big data from in-store shoppers. Shoppers transmit information about where they are in the store, how long they stay there, and whether they made a purchase. Harvesting this big data helps with store layout, identify which products attract the most attention, or identify problems with speedy checkouts, as well as other store atmospherics issues.
Better forecasting and budgeting
By understanding the ROI of specific marketing activities, you can optimize the way you spend your marketing $$$$. You can also generate better, more accurate forecasts through big data analytics.
Sales people constantly complain about the quality of the leads they get from marketing. Of course, they would like leads that simply require a phone call to write up the order. But, the fact remains that often sales leads are little more than a signup for an email newsletter or viewing a particular page of a website. And, that’s bad.
Lead scoring involves tracking behavioral information or purchasing information from 3rd party vendors that helps identify which prospects are leads (have need, money, authority, and desire for your products) versus tire kickers.
I once developed a lead scoring algorithm for a firm based on reader behavior from their list of nearly a million email subscribers. The algorithm, derived from an econometric model, used frequency, recency, and readership of particular articles to score leads. Subscribers who scored above a certain threshold became leads for the sales force.
Image courtesy of HBR
As humans, we are numerically challenged. It’s hard for us to see patterns within columns and tables of numbers. Thus, visualizations help build insights we might miss if we saw them in an Excel spreadsheet. Using the correct visualization is necessary to derive insights, so don’t just convert tables to graphs without considering how the data is best visualized.
HBR provides a guideline for determining how to create visualizations using a visualization — a 2X2 matrix describing the type of data available against the type of insight desired. You can read the entire article using the caption link.
Art and science of data analysis
I’ve discussed earlier that data analysis is part art; part science. That’s especially true of big data analytics. So, I’m a little concerned when I hear data wranglers confine their discussion to issues of Python, R, SQL, and other analysis tools. Sure, it’s great to know how these tools work, but you first need to understand how to ask the right questions — which means you need to be a subject domain expert, first. So, if you’re doing marketing intelligence, you need to be a marketer first, an analyst second
. Otherwise, you’re not really doing marketing intelligence, you’re back in the stone ages where marketers waited for weeks or months to get insights from analysts able to pull the right data and run the right analysis.
Doing marketing intelligence also requires someone who understands other social analytics tools such as SPSS or SAS because these tools are critical for building algorithms, creating clusters (segments), and model building, which attempts to assign causal relationships among factors that impact conversion.
Finally, your data wrangler needs to understand text analytics as IBM estimates 80% of all that big data is unstructured (words, images, etc). If you can’t get a handle on what that unstructured data means, you’re losing most of your insights. And the balance between art and science in evaluating unstructured data definitely swings closer to art, than science.