Let’s say Team A is your favorite basketball team. Over the past few seasons, it has absolutely owned its rival Team B on the court, to the tune of 18 wins in the last 20 games. Going into the upcoming season, you feel pretty confident about Team A’s chances to get off to a fast start since its opening game is against Team B. Predictive analytics would tell us that you’re good for a win 90% of the time. So, trust the numbers, right?
But hold on a moment…
In the off-season, Team B shocked the sporting world and signed LeBron James, Kevin Durant, and Stephen Curry. No one is quite sure how this new “super team” will play together, but one thing is for certain: it’ll play a lot differently than last year.
Trusting the numbers suddenly becomes a little trickier.
Looking at the Problem
For marketers, the trouble with predictive analytics starts when an unknown variable (like a trio of significant newcomers to your audience) enters the equation or when you don’t have enough data (or standardized data) to analyze. Both those situations are difficult to plan for. Even if you’ve anticipated the unknown, you’ve still never seen it before, making it hard to adjust for what its true effect will be.
Without the right data in place, what are you really predicting?
A new Gartner research report concluded that B2B technology marketers must have some form of predictive lead scoring to prioritize leads from inbound channels.
As marketers become more proficient at creating accessible content across all media, predicting which behaviors signify purchase intent becomes more difficult. More than ever, content consumption is a misnomer of purchasing desire. If the report is to be believed and predictive analytics is a must-employ tactic, marketers then need to understand its scalability limitations and to avoid the analytics pitfalls of the past.
Not too long ago, marketing analytics came from within your CRM database. Names, job titles, and company roles told the story of prospects and buyers—and that was about it.
All that data resided inside the marketer’s own database. It was his or her job to collect and make sense of it. This myopic process created scores of incomplete personas and superficial information.
Marketers knew broad, summary-level data was often inaccurate and lacked deep insights (not to mention the rapid degradation in validity and quality of that data). You may have noticed the use of past tense here, but this phenomenon is still very much alive.
Why What Worked Isn’t Working
Buyer life cycles all have different lengths. You might choose to replicate the shorter sales cycle without having the right information in the first place. Maybe you started tracking it too late, perhaps only at the end. That’s dangerous.
Once a marketer has even the smallest amount of data, he or she tends to make the hasty decision to assemble a sales playbook by what’s working. Marketers end up basing the playbook off the data the sales team has been collecting along the journey, even though sales is motivated only to capture whatever is necessary to the deal at the time.
As a result, marketing makes highly impactful, sweeping decisions based on fundamentally flawed information.
Blaming the Data
The new data revolution has combined structured data with so-called unstructured data.
Unstructured here means it is information you can’t glean by combing through a company’s website. That still doesn’t make it easy to capture data that may be scattershot across emails, social media feeds, phone call transcripts, Skype exchanges, and who knows where else. There are still too many discrepancies to call predictive analytics 100% scalable.
For example, if a team is motivated to sell a particular product, it may squeeze someone into a sales cycle for this product, even if the person isn’t quite a fit. Automatically, the data is wrong. That’s why data science is done by really smart people, usually from a third party not subject to internal biases (or, at least, less so).
Cutting-edge B2Bs are scrapping linear prediction models in favor of real-time insights that start at a desired outcome and work backwards, taking into account all variables along the way.
A perfect example of this is a restaurant using push notifications to alert a passerby of a drink special on a hot day, even if that specific individual has never been to that restaurant before.
The future of analytics will combine data in such a way that the end result seems obvious from the get-go, almost as if a clever trap has been laid before potential leads.
Not Quite Caught Up
The mentality behind predictive analytics is sound. When done right, it can help to cut down on wasteful content from your marketing team and to find ideal prospects for your product.
But even as some companies begin to embrace true data science by combining Big Data on top of traditional CRM information, we’re likely years away from most marketers catching up.
Why? Because true data science involves a lot of research and experimentation. There are no easy fixes or absolute outcomes.
Moreover, a true admission of unknowns is scary for most marketers.
Predictive analytics has a natural ceiling in terms of scale that will remain in place until data governance, collection, and hygiene becomes much, much better. In that same maturity curve, marketers must better understand how to use non-traditional information.
The good news is the brightest minds in the business always figure it out, which is a predictive analytics way of saying it will happen again.