News & Ideas

3 Ways Marketing Analytics Can Make the Cut

The consequences of “over promise” and “under deliver” in marketing analytics were laid bare recently in a Marketing Dive article examining the Gartner report, “Predicts 2019: Marketing Seeks a New Equilibrium.” Kevin Matray, TROVE's VP of Customer Solutions, explores how to get marketing analytics back on track.

The article opened with a pretty bold prediction:

“Sixty percent of CMOs will cut their marketing analytics departments by 50% by 2023, due to a ‘failure to realize promised improvements.’”

These are very striking numbers. Not necessarily surprising numbers, considering the many ways I’ve seen marketing analytics misapplied, but still striking in that the promise of analytics in the hands of so many companies clearly is not living up to its potential.

We at TROVE have some suggestions to help you greatly increase the success of predictive analytics in your marketing endeavors.  

But before I get to that, rest assured there is power in data. The move to data-driven decision making is the right path forward, and marketing analytics can play a productive role in this change. The problem, however, is that too many companies are making wrong turns on that path – and the mistakes are avoidable. Here are three observations to help you stay on course.

1. Data Science is not a “project,” it’s a way of doing business.

Too many companies wade into the world of predictive data science with a “project” mentality. They start with a specific question they’d like to answer then build a project around it, with all the overhead and approvals that go with it. 

Months of effort (and dollars) go by following a cumbersome internal project methodology before modeling even begins. Sometimes I even see work and spend going into “productionalizing” the model while the model is still being built. Then the team hits a wall with the model, because it just isn’t delivering the results they were expecting. 

Sound familiar? Now they are months into the project and have to make a difficult choice:  keep working on the model or shut the project down. This is not an efficient approach, especially considering a typical analytics team can only address a few projects a year.

To be efficient, you need an agile approach. Trove believes that Agile Data Science is the key, where upfront thinking (not spending) identifies a prioritized list of business questions to address via a quick Proof of Concept (POC). This is a succeed or fail fast methodology that lets organizations quickly learn which analytics to embrace BEFORE investing in an implementation. Only successful POCs move into the operational stage.

Meanwhile, the analytics team gets to address multiple business questions in parallel, quickly moving from one to the next. This agile approach greatly increases the number of business questions an analytics team can evaluate and focuses resources on only those that have been proven successful. Additionally, the output of a successful POC, while not yet operationalized, provides valuable insights in a much shorter time horizon than a traditional “project” approach.

2.  Don’t jump in the (data) lake. Learn by doing.

There is a big misconception hampering analytics teams: the idea that they need to get all their disparate data cleaned up and in one easily accessible place before they can start doing meaningful analytics. Of course CMOs are going to cut their teams in half if they are being told, “we need years to gather, clean and normalize all our data – visa vis a data lake – before we can start working on analytics.” The data “lake” quickly becomes “an ocean.”

Years! That’s the amount of time some of my prospects have cited. Well that’s years your team could have been getting its hands dirty with data and predictive analytics – years of invaluable learning! There is a better way.

It’s funny:  these same companies that want to become “data driven” don’t even apply the concept to their approach.  Instead of focusing on “all” of the data you will eventually need to run a robust analytics program, start with the data you need to answer your most imperative business questions (see Agile Data Science methodology above). Instead of boiling the ocean, work with that data first to quickly see exactly where you need to focus your cleaning/normalizing/augmenting efforts to enhance the result.

To be clear, to be successful you have to have and implement a data management strategy.  Data governance, the architecture to support it, data lakes, etc. are a good thing and required for success. It’s all about how you get there:  wade into data science sooner rather than later and let your findings drive the priority of what data you bring into your data management systems.

This approach yields additional benefits. Everyone who has worked with analytics knows that “data wrangling,” the act of gathering data to make it prediction ready, is a major component of data science. By going through this process for one analytic, much of the data can be ready for the next one, allowing for quick product- or service-specific modeling in an accelerated time frame.

TROVE’s CEO Ted Schultz is fond of saying, “Nothing improves data quality faster than putting it to work.” Put another way, data science is a “learning by doing” field. You aren’t going to learn much waiting for the lake to fill.

3. Get personal with micro-personas and dynamic segmentation across the customer lifecycle.  I spoke recently with a marketing executive who remarked that her company utilizes a customer segmentation service that places all her company’s customers and prospects into five unique segments. All their marketing efforts are tied to these segments.

Unfortunately, I hear this a lot. It’s the industry standard, but it’s also dead wrong.

Segments need to be created specifically for the business question at hand. That’s because the segment defining a good prospect for a select product changes from product to product.  Put another way, if Sally and Jim fall into the same high propensity segment to purchase product or service X, they will often fall into different segments for product Y.

This also holds true across the customer lifecycle. Marketers are mistaken if they think they can rely on a single analytic to optimize results across the different stages of the lifecycle. Success lies in applying the right analytic to the right stage, from prospecting to engagement to retention.

Specificity is key. If you want to prove your marketing analytics’ worth to your CMO, you are going to have to leave your old “shotgun” approaches at the door and get personal. Say goodbye to sending out 200,000 mailers and hoping for 2,000 replies.

Customers and prospects demand a personal touch – across the lifecycle.  A new generation of data – think of the segments offered by the traditional data services, but on steroids – and advances in modeling – driven by AI and machine learning – are issuing in an era of “micro-personas” and “dynamic segmentation.”

When your marketing team can “know” the person on the other side of your offer – the fact that they just bought a house or are a world traveler – you can tailor your offer and time it with precision.

For example, a credit-card provider offering a discount to a customer’s local home-improvement store within a month of their purchasing a new house is going to drive more business than a competitor who never knew of the move or sends the offer too late. Same goes for the world traveler. When you know your customer, you can make offers that generate more business – a free hotel room in Venice, perhaps, for buying your airline ticket with our card.

Knowing is key. A new generation of data and analytics can get you there. It’s all part of making data science your way of doing business.


Fill out my online form.

Thank you.
We'll be in touch soon.