Predictive Analytics, or Predictive, as we call it for short, is a term that most marketers are aware of today. While there is no lack of content around the topic, we’d like to share a definitive guide that answers the most burning questions around predictive: What is it? How do you use it? Who is it for? And of course, how do you select the right vendor?
Since an actual post on everything related to predictive analytics would end up being rather long, we’ve decided to break up this topic into multiple parts. In this post, we’ll define predictive analytics and explain the importance of data quality in predictive.
Predictive analytics has been defined in many ways. The more technical definition reads, “predictive combines vast amounts of historical and contextual data to create a probabilistic model that predicts which actions or audiences have high likelihood to succeed and which have a high probability to fail.”
However, a description of the emerging technology category can be drastically simplified – a system that uses your past data to predict outcomes. Or, one of our favorite explanations which came from the Chief Analytics Officer at Swift Capital, predictive is simply the 80/20 rule.
We know 80% of our revenue will come from 20% of our prospects. And 80% of pipeline will come from 20% of our campaigns. Predictive identifies the 20%.
You don’t have to just take our word for it – we’ve also compiled industry experts answers to the question. While their responses vary, the underlying concept is the same – predictive involves predicting where you’ll have success.
With the growing hype around Artificial Intelligence (AI), Big Data, and Machine Intelligence, it can be difficult to understand how predictive analytics stands out from the crowd. Gartner helps distinguish predictive by comparing it to other forms of analytics:
Beware that most companies’ current marketing models are based on an implicit assumption: what happened in the past will happen in the future in the same way.
“This involves shifting from a historical orientation to become more forward-looking. This is a crucial difference, the aim is not just to describe and understand the past but also to explore ‘what is likely to happen’. For this level of difficulty, we are making decisions about the future which we can only express with limited certainty. Diagnostic and descriptive analytics can be precise. Predictive analytics is, by definition, based on probability, and must be accepted as delivering ‘degree of confidence’ only.” (Mobey)
To illustrate how predictive analytics is different from other forms of analytics, let’s look at the most straightforward predictive use case – direct marketing.
In direct marketing, predictive analytics identifies the probability that a certain customer will respond positively to an offer given all the relevant past information on similar customers. A follow-up of this prediction could be to contact all customers with a probability of positive response above a certain threshold. This completes the predictive analytics use case since the management team decided which customers to contact based on how much to invest across the channel mix, which was determined with a fact-based prediction.
SiriusDecisions, B2B advisory and research firm that has been longtime proponent of advanced data-driven technologies, offers an equation that answers this question holistically.
While the algorithms used to make predictions are powerful, predictions are only as good as the underlying data sources. As the above graphic briefly mentions, vast sources of external data are used for modeling, which has naturally forged use cases beyond predictive analytics. Most notable is the ability to source net-new prospects and to improve existing customer and prospect datasets (enriching, cleansing, diagnostics). Depending on the maturity of your organization – which we’ll cover in a later section – your starting point with a solution may be fixing critical bad data issues prior to leveraging sophisticated predictions.
Before diving into how predictive can help deliver value, it’s important to take a closer look at the data (signals) that make the insights, leads, and services uniquely powerful.
Think of data as an iceberg.
The tip of the iceberg is made up of your traditional buying signals – firmographics such as location, headcount, and industry. While your accuracy and fill rates for even these basic data points may be subpar, they are typically available for segmenting and reporting within your CRM and marketing automation technologies (MAT).
Predictive analytics platforms, on the other hand, track and correlate all the buying signals that are “submerged”. Deeper-levels of signals expose business motivations and challenges not available when just using firmographic data, providing added context to your leads and customers which helps drive more informed marketing decisions.
For a more detailed list of signals tracked by Radius, read our Segmentation Datasheet.
With more signals, marketers are enabled with key indicators specific to a business’ propensity to buy and respond to marketing campaigns. These insights can then be leveraged to further personalize and extend campaign messaging and channels. Below are a sample of three customer profiles that advanced signals can reveal:
For a detailed breakdown, read our predictive playbook on How to Sharpen Targeting with Micro-Segmentation
While it’s important to explore the breadth of data, data quality is even more critical than data size. Gaps in B2B data quality and actionability are holding many business software providers back from maximizing revenue potential with 80% of B2B marketers blaming data quality for ineffective demand generation processes. (Demand Metric)
After working with 7 years of data across 600+ million connected CRM records and over 50 billion dynamic data signals, Radius’ team of data scientists has uncovered groundbreaking insights into just how difficult it is to maintain data quality.
On average only 70-75% of CRM data is accurate.
It is also critical to understand that the insights and prospects delivered by a predictive technology are only as good as the underlying data. Later in this series, we will discuss how the data collected and data science used to match CRM datasets can be used to drastically improve overall data health and enhance targeting capabilities.
Being able to define predictive analytics and understanding the role of data quality are key for forward-thinking marketers looking to adopt predictive. But, it’s equally important to understand how you can leverage predictive in the context of your business. We’ll cover specific predictive use cases and highlight the benefits in our next installment of the series – stay tuned.
In the meantime, read about PrimePay’s predictive marketing journey and learn how they leveraged Radius to optimize their Go-to-Market strategy.
When we commissioned Forrester Consulting to conduct a comprehensive survey of B2B marketing leaders to hear their views about predictive analytics, we were pleasantly surprised by how many were fami
With over 61% of B2B marketing leaders already implementing solutions, it’s clear that Predictive has progressed from a nascent marketing tool to an essential solution in the marketing technology (m
When it comes to competition, no two brands have had a more fierce, long-standing head-to-head battle than Coke and Pepsi. In fact, historically they’ve had many famous clashes, but one in partic