Searching for Answers About Predictive Marketing for Brands
The term predictive marketing can conjure up thoughts of either complicated data science or mystical fortune-telling. But it’s really not that complicated or mysterious.
Prediction boils down to finding patterns in data, specifically patterns that let you calculate the likelihood of future actions or desired outcomes.
For example, if you have customers who purchased a product or service – like high-end bed linens or on-demand doctor services – then you can use data to find new people who are likely to buy those same products. We call these people likely to buy, net-new prospects. ‘Net-new’ just means people who’ve never heard about or been customers of your business.
Unfortunately, before predictive marketing, the word ‘prospect’ was a flimsy term that usually pointed to a random person, who perhaps was a certain age or gender. ‘Psychographic’ criteria were added over time, in an attempt to account for things like people’s activities, opinions or possessions. But, marketers couldn’t effectively process demographic, psychographic and other data all together – in toto – so no one really knew if these prospects would love or buy a product.
Today, the word prospect is becoming a more accurate and powerful word for B2C marketers. True predictive-based, net-new prospects are people new to your business and who are going to love and buy your products. For customer acquisition marketers, real net-new prospects are what dreams are made of. Predictive marketing lets you find and acquire them.
To help further demystify predictive marketing, here are a few other misconceptions—
The Top 5 Myths Of B2C Predictive Marketing
Building Predictive Models Takes A Long Time
This was the case at one time, however advances in technology like machine learning and cloud computing have dramatically reduced the time it takes to build powerful and accurate models. With the right software, you can now build models on a daily or even hourly basis.From data preparation, to model building, to scoring prospects for campaigns – the process can be streamlined for better customer acquisition.
You Need Data Scientists, Statisticians or Analysts To Build Predictive Models
Predictive models can now be created easily in the cloud. It doesn’t require technical subject matter experts. Any marketer can build a solid model themselves in minutes, which is great, because there is still a human element and marketers are best suited for the job.A software platform can build a model (or a data scientist, if you have time and money to waste), but knowing the customer base and marketing strategy will enable the proper reasoning behind building a model in the first place.
Marketers can quickly grasp what they need to know about how modeling works. It has become an accessible discipline like content marketing or digital advertising. As you plan, build, and implement predictive models, your familiarity with predictive details like ‘tiles’, ‘scoring’, and ‘likelihoods’ increases over time.
More importantly, marketers know what to do with model results. They can rapidly run acquisition campaigns to predictive-based prospects and feed results back into the modeling software for continued campaign improvements and optimization.
Building Predictive Models Is Expensive
In-house, manual data science is expensive due to the team and resources required. Meanwhile, outsourced, manual model building can cost many thousands of dollars per model – that’s just for building and doesn’t cover the cost of implementation. Predictive software platforms are cutting overhead by using machine learning and algorithmic prospecting to automate the building of custom, sophisticated predictive programs. The result is better modeling and implementation than previously available, yet at a reduced cost.
Transactional Customer Data Is Sufficient To Build Predictive Models
Brands are rich in data. Or are they? Transactional customer data is an advantage, but by itself won’t let you find net-new customers, who have yet to purchase your product. You also need demographic, financial, and behavioral characteristics to determine who your best customers will be, beyond previous spending that occurred. Predictive software platforms are able to enrich existing customer transaction data with hundreds of other valuable pieces of information.
Likewise, third-party behavioral and community data is a vital resource that allows you to reach net-new customers and avoid ‘red oceans’. (Red oceans are where companies for limited pools of customers reducing them all to competing on price.) Predictive software processes external data to open up immense pools of relevant prospects, new to your brand, who will look and act like your best customers.
Predictive Models Don’t Work / They Aren’t Better Than Univariate Targeting
Targeting prospects based on a single variable or imagined personas are expiring solutions. Marketers have relied on simple targeting, because there wasn’t a better option. When trying to hone in on future buyers, it was better to identify at least one, or a few, meaningful variables that ostensibly indicated something important about customers or prospects. This has limitations in the real world. People are more complex that simple persona-based modeling or univariate targeting can account for.
As marketers, we’re used to making assumptions about our ideal customer and potential buyers. We’ve grown used to attributing traits to segments like “Eco-Conscious Moms,” based on anecdotal evidence or severely fragmented data, which – when put to the test – doesn’t correlate or predict future purchase behavior. It’s not our fault. Marketers chose the best criteria available, but have been missing the many hundreds of variables and characteristics that give a true and dynamic view of customers and can identify best prospects.
Try it – run a campaign with predictive-targeting versus a random or univariate selection and check the results. With a control group, you’ll clearly see the power and ROI of predictive prospecting.
This article was originally published on the Predictive Analytics Blog.