The Enemy of Data Science: Noisy Signals in the Enterprise

To understand customer signals we must create a body of knowledge that can power data science and predictive algorithms.
Customer Experience

The past five years have seen a great evolution in enterprise applications from workflow automation to intelligent applications. The giants of enterprise computing agree that the future is centered on intelligent applications, data science, and analytics.

I agree that all companies need a better and deeper understanding of customers and target markets, and they need their enterprise applications to help them execute. The challenge now comes not in the technology, but the data itself.

The Importance of Clean and Complete Signals

Intelligent enterprise applications would do well to learn from the issues in consumer marketing. Targeted messaging and recommendations from our favorite brands are commonplace, but those systems are very dependent on the quality of the underlying information.

Everyone's had this experience: I buy something on Amazon for one of my teenage sons and that confuses the Amazon engine, which starts recommending online games that I would never buy for myself. There is nothing wrong with the data science algorithms at Amazon, they are just working on non-relevant signals. Netflix gets just as confused when my sons use my Netflix account to watch teen shows.

The world of customer signals in the enterprise is much more challenging than it is for Amazon and Netflix. Those services rely mostly on the first-party data that I agreed to share. Enterprises don't have the continual engagement with prospects that Amazon has with its audience, and must rely on both first-party data (collected by the enterprise) and third-party data (sourced by external providers).

The problem is that B2B first-party data changes very rapidly. On average, 25 percent of employees change roles every year. Company firmographic data, such as revenues or employee count, changes frequently as companies grow, shrink, merge or go out of business.

Third-party data is used by enterprises to augment first-party data. Traditionally companies looked for quantity (e.g. give me more names, more leads, more companies) but today the focus has shifted away from quantity to quality (e.g. I need accurate data, and as much insight on this person/company as I can get). The struggle to obtain and maintain information quality and accuracy is universal.

The Informed Enterprise

Data science, predictive, and recommendation technologies have the potential to help us understand where the opportunities lie, and how to most effectively pursue them. This will take CRM from a past focused on workflow and automation to a future focused on intelligence and prescription.

We see customers apply data science to improve lead scoring, customer engagement (frequency and quality), and messaging. Currently most of these efforts are focused on silos of data (inbound lead flow, customer interactions, targeting databases), but they will quickly evolve to integrate as much information as is available about prospects, customers, and their interests.

The Right Signals are the Foundation

How can we all prepare for this new world, so that we are at the forefront of the informed enterprise? We must create a body of knowledge that can power data science and predictive algorithms:

  1. Does our CRM have clean, normalized data? If we have the same company or person represented in multiple, disconnected records no algorithm in the world will be able to make sense of that information.
  2. Does our marketing database have the up-to-date and normalized information that we need to segment, score, and route leads?
  3. Do we have a way to maintain the marketing and CRM information up-to-date (hopefully in near-real-time)?
  4. Do we have the additional data that represent buying signals, mapped to the right companies and people? Signals include both behavioral information (is this company hitting our website) and demographic/firmographic information (does this company fit our qualifying criteria? Is this the right decision-maker, and if not who is?)

If we start with a clean dataset rich with the right signals, any of the great predictive and analytic technologies available in the market will do a great job for our businesses. Otherwise, garbage in will inevitably result in garbage out-and it will continue to be a noisy world.