Marketers are bombarded with messaging about how important it is to leverage big data. But big data efforts largely fail. And one has to ask, why? Why is this effort to use the data in one's own organization so difficult?
The reason is because marketers and their service providers often fail to understand that there are three distinct problems to solve, and in their efforts to engage big data, these marketing professionals don't completely solve for all three.
1. Connect the Data Sources
The first problem that must be solved is to actually retrieve the data. In the case of a retailer, the critical data typically resides in three systems: the eCommerce or order management systems (ie, Magento), the web analytics platform (ie, Google Analytics), and the email marketing platform (ie, MailChimp or Salesforce Marketing Cloud). Unto itself, this is a monumental task. What data do you need? Has it been normalized? What primary keys will you match the data on? Is the data, once moved, actually in a format that can be used for marketing purposes like personalization (things like making sure first names are title cased) or automation (ensuring that the date fields are recognized as such in the marketing platform so that the triggering actually works)? What happens if integrations break? Who's going to fix it? How soon can they get to it? etc, etc.
Many big data efforts founder here, primarily because the marketer isn't entirely sure what to ask for, and the IT guy tends to produce exactly what is requested. The result is that inactionable data gets passed and the project stalls.
But even in the cases where the data is passed and it is right, we've only started the journey. The next step is to actually make sense of the data.
2. Analyze the Data
The next challenge is to reduce all of that data to logical analysis. This is the stuff of data scientists and statisticians. As marketers, some of us are exposed to elementary statistical concepts in school, but rarely have we explored these concepts further, and very rarely have we been in a position to put this to work. The outcome is that we really don't know what to do with the data once we've gathered it.
Raise your hand if you can run a chi-squared test for goodness of fit. Ok, raise your hand if you even know what that means. Right. These are not the standard capabilities that in the past have been taught to marketers unless you have specialized in database marketing (and most haven't). I predict, BTW, that we'll see a great surge in course offerings around database marketing and statistics to meet this unfulfilled need.
The analyses that form the backbone of making data actionable is two fold. The first is using the data to logically group different segments in your customer set. Big spenders, sleepers, best customers, average joes.... all of these can be deduced by logically breaking customers into segments based on their buying patterns. You can use established methodologies like RFM Analysis, or come up with your own groupings, but the point is that you have to come up with groups that can be marketed to in ways that more closely pertain to who they are. Guys will generally buy guy stuff. Left-handed golders tend to prefer left handed products (surprise, surprise).
The second, and trickier form of analysis has to do with anomaly detection. How and when will a marketer know that there's been a sudden acceleration of sales of a particular product? What caused it? Can it be exploited to sell more product. One of our customers, for example, sells a wide variety of LED flashlights. He made the observation that the slow hurricane season has impacted sales (he correctly suggested that it was great that people hadn't had to suffer through a storm... he just noted that there was a correlation between storms and sales of flashlights). Why? Because people stock up on emergency portable lighting in advance of a storm. This marketer could be relevant and helpful to his constituents by staying on top of when storms are coming to what geographies, and providing useful information about emergency resources, thus building his brand, and being in front of the very people who will have a need that he can fulfill.
But this sort of analysis is hard without the tools and skillsets required.
3. Act on the Observations
The final component that is required is the ability to actually act on the observations.
In our work, we've found that there are three key things that have to happen for marketers to take action.
- The first is that they must have a concept to work on. We surface these to our clients in our "Actions" feature. "Hey, you should target Oregon and Washington for sweaters because a big snow storm is coming, and assuming a 1% conversion rate, you're gonna make $150,000."
- The second is they must have the tools and the knowledge to execute. Their marketing platform, for example, needs to be able to segment on state. (Field "state" is OR or WA, for example).
- The third is that they must have the data to actually to the work with. This goes back to the original premise that you must connect the data to the platforms so that you can actually execute, and it needs to be in a user-friendly format.
This Stuff Works
Both in our direct work with clients, and in innumerable industry examples, engagement and conversion rates shoot through the roof when you use data to engage your client base. We were speaking with a marketer who increased weekly sales of a specific SKU by 200X by logically targeting people who'd purchased similar items in the past. Spanger Candy (a client of ours) reported in their single highest day of eCommerce sales, ever, when they used Windsor Circle to target specific segments of customers with relevant messaging. This stuff works.
If you've been struggling with how to use your data, and are looking for a partner to help, give us a ring...