Businesses have invested heavily in data analytics to understand consumers and to predict what they might do next. Smart companies want to take that a step further and use insights from predictive analytics to build what is being described as “predictive engagement”.
According to an Investigates study by the Which-50 Digital Intelligence Unit, in partnership with Genesys, “Predictive engagement is the result of modeling solutions based on historical and real-time data to predict customer journey outcomes. It utilises predictive analytics to help organisations drive more intelligent customer engagement”.
- Download Which-50 Investigates – The Shift to Predictive Engagement. Produced in partnership with Genesys
- A Peek Into The Future – Identifying The Benefits Of Predictive Engagement
Crucially, the authors argue, “It enables organisations to act when an action is required, not react after the moment has passed”.
So what does predictive engagement actually entail?
The report, called “The Shift to Predictive Engagement”, identifies techniques such as graph analysis, simulation, complex-event processing, neural networks, recommendation engines, heuristics, AI and machine learning to prescribe a preferred course of action.
These then inform strategic and tactical decision-making, operations and training, ensuring salespeople and marketers have access to the most accurate information to maximise sales. It can even be coded into systems to automate decision-making, according to the report.
At a practical level, companies that are able to master the art should then be able to map customer journeys across channels, personalising experiences to maximise conversions. Importantly, this applies to offline as well as online channels, making it especially relevant for marketing leaders in industries such as retail and financial services.
The report’s authors say the business leaders they spoke to describe predictive engagement as the intersection of science and art.
For instance, according to Megan Isherwood, Marketing Director at Digital Document specialist Conga, “we are in a march towards operational excellence. The organisation sees the value in what we are doing as we continually review, optimise and recalibrate.”
Predictive engagement is proving especially helpful for those wishing to implement genuine omnichannel marketing. That’s because it is effective at breaking down the walls that companies often build — for organisational or technological reasons — between those channels.
“Today, marketing channels are much more integrated, and we are better able to measure the handoff. It’s changed the dynamics of marketing teams because truly integrated campaigns can no longer just be lip service,” says Isherwood.
Other executives cited in the report argue that predictive engagement removes barriers to action and allows marketers to tailor buyer personas more closely.
“This leads to a more consistent customer experience across channels — which is a key goal of good CX strategy,” say the authors.
Of course, the art requires some mastery of the science — in this case, the science of predictive analytics.
Last year, Gartner argued that predictive analytics had begun its slide into the trough of disillusionment. The analysts said “overinflated” vendor claims and “excessively ambitious” early projects had produced the predictive analytics blowback.
However, Gartner expects the technology won’t spend much time in the trough. Its “rate of evolution and underlying value” will propel it quickly through Gartner’s hype cycle.
Not far behind it is prescriptive analytics. Prescriptive analytics is more advanced, allowing users to create rules for when predictions are met, thereby automating business processes.
The predictive engagement software market will reach $1.88 billion by 2022, with a 20.6 per cent CAGR from 2017, says Gartner. By 2020, predictive and prescriptive analytics will attract 40 per cent of enterprises’ new investment in business intelligence and analytics.
Recognise the pitfalls
But analysts and data experts the report’s authors spoke with warn of a complex market, ethical considerations and a skills chasm that will likely see many marketers fail to predict their next white elephant.
All the while, consumer understanding and sentiment toward data and collection is trending in a direction which will likely make marketers’ jobs harder.
In Australia, marketers are taking a cautious approach to predictive analytics — usually taking the lead from North America, where much of the technology is developed and talent resides.
“There are sophisticated areas [in marketing] but I don’t think they are commonly used,” says Ian Bertram, a Global Manager within Gartner’s Data and Analytics team.
“I think a lot of marketing people are still dipping their toe in the water around using some of that data,” he told Which-50.
What are marketers predicting?
Some marketers are already using data from customers’ facial expressions while viewing advertisements to form and predict purchase intent, according to Bertram. But he says most are expanding on traditional marketing metrics to determine the quality of advertising content and placement.
Media buying agency, Audience Group, implemented predictive engagement techniques last year. In the Investigates report exploring the technology, Audience Group director James McDonald explained predictive analytics allows the company to dynamically shift budget spend across market segments in real time, according to how they are performing.
“It makes those decisions on the fly, and better spends the budget on available audiences rather than over or underspending,” McDonald said.
“Rather than signing off individual channel budgets, now we go ‘right, it’s a $250,000 budget, these are our goals, this is what we want to do, and the technology goes out and spends our money. There are still humans involved in the middle of it, but the AI machine learning within each of the training desks allows us to better optimise on the fly.”
The Investigates report also quotes Karen Ganschow, lecturer at the Macquarie Graduate School of Management and former General Manager of Consumer Marketing and Customer Strategy at NAB.
Ganschow said the bank’s customer data is used to predict customer interactions and minimise unnecessary interventions.
“If you are only reacting, you are usually too late to have great effect,” she says. “Predictive tools allow you to intervene with customer events that precede the final step of a customer deciding to finally take their business elsewhere,” she says.
The technology is also used to predict and preempt customer churn, according to Ganschow. She says the machine learning algorithms often uncover insights that would otherwise be missed.
“Using predictive analytics helps you see the ‘early signs’ that indicate a customer is starting to question their relationship with your brand,” she says. “It helps you see the preceding events that start to make a customer question, so you can intervene before they have committed to taking their business away.”
Ganschow says the insights must be considered sensibly, and not all patterns are indicative of customer behaviour. “It is not a silver bullet, but being predictive and proactive increases the likelihood of positive outcomes,” she says.
Predicting customer churn has also been an early use case for NBA franchise, the Utah Jazz. Its marketing department is in the early stages of a customer experience overhaul, using data and analytics to improve fan engagement and ticket sales.
According to Jared Geurts, Senior Director, Marketing Analytics at the Utah Jazz, the predictive analytics use case standing out early is customer churn.
“We do quite a bit with who is and who isn’t going to renew season tickets,” Geurts told Which-50 at the ADMA Data Day event earlier this month.
“Because we make so much of our money from our season ticket base, which is really only about 3,000 people, for us it’s a huge problem if we lose customers. So a lot of our predictive work is around who we are at risk of losing.”
Geurts explained that the Jazz combines customers’ browsing data and sales data with first-hand responses from fans at games in a central data warehouse, then runs algorithms to predict a churn rate and identify those at risk.
Ultimately, Geurts says, it can improve the customer experience. But he too cautioned that it must be seen as only a part of a bigger marketing strategy.
Indeed, while some of the marketing applications for predictive analytics are proving valuable, the technology is far from a silver bullet — a message repeated by both practitioners and analysts.
It’s also a more complex and resource-intensive challenge than it may appear — or at least than vendors often present it.
Gartner’s Bertram said advanced uses of predictive analytics should be accompanied with appropriate domain and data expertise — a challenge in the already tight data science talent market.
Sophisticated analytics programs will involve several levels of data collection, ingestion, analysis and execution, according to Bertram. In sophisticated programs, data engineers and data scientists — both hot commodities — are needed at the ingestion and analysis phases, respectively, Bertram said.
At the execution phase — where marketers will likely become involved — a level of data literacy will also be required. Some “cowboy” organisations ignore this advice, Bertram says, and often pay for it down the line with poor performance or difficulty operationalising insights.
“Business users … can go out there and buy a packaged [analytics] app, slap it onto some data. But without the appropriate training to say ‘this is what the data means, this is how you interpret it’ — the storytelling bit — [they’ll be challenged].”
Just as important as data literacy are the ethical considerations throughout the analytics process, according to Bertram.
“The ethical consideration needs to be done at multiple points … Hopefully, organisations are now putting [data] literacy training and ethical training and governance training in all of those different points of that particular process,” he said.
“There’s a huge undertone of ethical considerations around a lot of this [data use].”
Craig Young is the President of the Association of Market and Social Research Organisations (AMSRO), an industry group with plenty of experience in data collection and use, with several members involved in marketing analytics programs.
He told the report authors the pendulum has swung on data collection and use, with a more informed public and, consequently, regulators demanding a higher standard from data practitioners.
He says for Australian organisations the data and analytics opportunity remains “massive” but in some instances, more restraint is needed.
“The promise and potential of big data, and the ability of companies to leverage big data and leverage predictive analytics to do interesting and exciting things with data, is increasing drastically.
“But I’d also say lagging that, but also moving quickly, are the public’s changing expectations and government’s legislative response to those changes.”
For marketers using predictive analytics in the name of personalisation and experience, the bar is higher now than ever before, with consumers less willing to trade their data, according to Young.
“As customers find out more about what is possible and what companies can do with the data that is held on them, I think customers are, in general, becoming surprised and … quite alarmed at what’s being done with their data.”
The risk goes up with analytics programs, Young says, because ultimate use cases are difficult to define initially. This becomes an inherent challenge to collecting informed consent. It also increases the potential of re-identifying individuals in data sets.
“Those data analytic techniques create an additional risk that didn’t previously exist with discrete survey data. And that risk is: you start to be able to create additional data by combining data from different sources. That then creates additional potential privacy issues.”
Young says public awareness and government privacy and data legislation are advancing, but still trail sophisticated analytics use cases.
“Things are being done I’m sure, and can be done, to manipulate and enhance individuals’ data in ways that are probably different to the initial purpose for which data originally had been constructed.”
For organisations with doubts or ethical questions about data use and analytics programs, Young’s advice is simple: slow down.
“That can be a very enticing prospect for the data scientists that are working with the data. But just because you can do things with data doesn’t mean you have permission to do it.”
While privacy legislation is notoriously hazy, Australia’s privacy principles provide a good place to start — particularly in regard to informed consent, according to Young. His organisation’s members abide by a deliberately more stringent code.
“If you don’t have informed consent from people to manipulate their data you really, ethically, need to be questioning ‘should I be doing this?’ And the answer is probably no.”
Marketers have spent much of the last decade building strong data analytics capabilities, and that has meant overcoming internal choke points like operation silos, while remaining alert to growing sensitivities of consumers about how their data is collected and utilised.
That trend is unlikely to change soon.
Instead, predictive engagement looks like the next natural consequence of the investments in analytics capabilities and technologies.
As the authors note towards the end of their report, “Predictive engagement is putting to rest the old Microsoft ‘paper clip’ pop-up approach — understanding that follow-up interactions have a huge impact on marketers’ ability to generate high-quality leads.”