Last week I visited with fellow technologist and Big Data evangelist Flavio Villanustre at LexisNexis. During the visit we discussed advances in data-driven marketing, a topic that I’ll be covering with a panel of experts at the upcoming OMMA Atlanta conference.
Here are some of the big takeways from our conversation:
- The biggest challenge that marketers face today is having access to data. The more data one has, the better data models one can build. And better data models drive better predications. Marketers must take every opportunity that they are given to collect and share data.
- The role of unstructured data (eg; photos, videos, audio) in data analysis will increase over time. For example, Google announced in 2012 that researchers used 1,000 computers to find cats in pictures. The impressive thing about this finding is the ability of computers to identify a particular object with accuracy without human intervention. This level of machine learning demonstrates that computer-enabled data analysis is something that we can take advantage of in the not-so-distant future!
- We are recognizing the value of predictive analytics. While descriptive analytics, or the collection of basic metrics (eg: visitors, page views, leads, likes, +1’s, pins, etc.), is important to understand what’s happened in the past, companies want to leverage data to predict the future and drive more revenue/increase profit.
- A very small number of companies worldwide (only 3%; according to Gartner Research) are beginning to use prescriptive analytics. Prescriptive analytics is a complex type of predictive analytics that allows one to test out multiple marketing models. It provides an optimal solution given a set of objectives, requirements and constraints. This is where say a company can test the impact of various promotions/discounts and shipping rates on a customer’s purchasing behavior.
- Marketers have to accept the successes (and failures) of data-driven decisioning. We hate to turn decisions over to a computer because we believe that we’re smarter — we’re human after all! Unfortunately, computers typically beat our “gut feel” — our intuition is just inherently faulty (according to HBR). Marketers need to accept that we’re biased and that we must adopt a “test and optimize” processes where the answer from one set of marketing experiments inform the next set of experiments.
It is easy to see how these advances in marketing — whether it is through the introduction of marketing technology platforms, data collection practices or data analysis processes. While it is tough to predict the future (thanks Yogi Berra), I believe that marketers that ignore data-driven decisioning are poised to lose (in the long run).
Late last week, Amazon made headlines (again) for receiving a patent on a method that predicts the eventual shipping destination of a product (see coverage on TechCrunch). The proposed method enables Amazon to leverage various data points to begin shipping an item before it is actually ordered — something that is both creepy and cool! In a way, it isn’t any different than the big data solutions that the NYPD is actively using to fight off crime or what top national universities use to recruit students.
While it may seem somewhat futuristic, it actually isn’t much different than a grocer watching the weather and ordering extra beer in anticipation of a run before a snowstorm. For Amazon, it is also the natural extension of their Subscribe & Save Program which gives customers a 15% discount on regularly purchased products, such as toilet paper, cereal, and dog treats. As this program enters its third year, Amazon is poised to not only see an incremental improvement in operating costs through real-time ordering/inventory management but also to prevent others from taking this same path.
As some retailers may be concerned that their competitive advantage is slipping while data-driven organizations like Amazon are gaining marketshare, the reality of the situation is that retailers aren’t that far behind. They simply need to initiate a “test and learn” methodology where there’s a focus on measuring an outcome through data collection and analysis. As part of a holistic data strategy, retailers can collect and stitch together information from several good data sources. For example, retails can understand:
- Shopping intent by collecting web traffic data on product reviews, weekly circulars, online shopping lists/wish lists.
- Buying patterns by collecting in-store transactional data (along with loyalty program data).
The key is for retailers to define a strategy!
This also reminds me of a recent slide that I came across from a Big Data presentation by McKinsey that showcases the profitability that UK-based Tesco experienced over the past 15 years:
It seems that the inflection in profitability didn’t occur until a significant investment was made in data collection/analysis.
For retailers looking at this graph, it is telling a story that there’s no better time than now to make such an investment.
What additional steps do you think retailers should take to secure a foothold in the marketplace?
I recently saw a funny video on big data that I just had to share. The concept of big data is that meaningful information can be extract from voluminous data — the kind that is too big for a standard database.* Nonetheless, big data is valuable in that it can predict what customers want (before they know that they actually want it!):
Big data is so hot right now primarily because (A) there’s lots of data sources (eg: Google Analytics, eCommerce transactions, in-store/catalogue orders, social media, etc.) and (B) data can be cheaply processed. But it seems that marketers believe that big insights only come with big data! To that point, it also seems that small data is all but useless! But is that really true?
Well, while bigger may seem better I recently
was reminded learned in my Data Analysis MOOC (aka online class) that big insights can also be found in smaller data sets! It just isn’t always feasible to have a super-sized data set — client may not have a large amount of data. And fortunately for them and for data heads, small data along with inferential data analysis (along with random sampling) can deliver big insights.
To get started, you just need a business problem, a theory and data to prove (or disprove) your idea! So don’t let the lack of big data stop you from doing the analysis.
*Big data is more than just voluminous (see O’Reilly).