There’s so much choice available that customers can pick and choose who they buy from and where, when, and how it happens. They want to discover, research, evaluate, and purchase on their preferred channel. Give them that option, and they’re more likely to choose you. That’s the whole point behind the multi-channel approach.
Customer analytics has always been an important part of B2C marketing, but B2B marketers have historically not been anywhere near as well equipped or informed, according to Palo Alto Networks CMO, Rene Bonvanie.
“For the longest time, when we talked about marketing analytics, we meant to say the analytics of the market – surveys or customer interactions on a purposeful basis,” the enterprise cybersecurity’s marketing leader tells CMO. “We’d go out to customers, speak to them, and on the basis of these findings, we’d draw conclusions about persona, buying intentions and so on. That shouldn’t go away, but the analytics I’ve been interested in is much more predictive and less jaded by the opinions in most cases, of a few people.
“In a company with 10,000 customers, why is it good enough to speak to 100 and draw conclusions about the whole? Especially if you have multiple interactions on a daily or weekly basis with these customers.”
Bonvanie is the posterchild for using data analytics at scale to understand customers, and has spent the past eight years at Palo Alto Networks building a marketing analytics function that not only gleans insights off the back of data, but can operationalise this for more successful customer engagement.
His ultimate ambition is not just predictive analytics but to perform “prescriptive analytics”, where he can provide best actions the rest of the organisation should do in order to convert customers.
The first step, Bonvanie says, is to collect data on every interaction through every channel with customers.
“From there, you start to come up with recommendations, patterns, correlations and potentially start to do predictive analytics – what should their or my next move be?”he explains.
“Ideally, and what has been eight years in the making here [at Palo Alto] is: Can I go to prescriptive analytics? What I mean by this is knowing someone in the company should do this, and assigning a percentage of change to that interaction. For example, if you call that customer now, your chance of success is X per cent. Or, the converse - don’t go in now, because the chance of success diminishes.
“That’s just the example of calling, but there’s other things marketers can do, such as using this for context, offers and so on.”
Getting to prescriptive analytics
Bonvanie’s background lends itself to a data-driven marketing bent. He is a scholar in economics and mathematics, and started his career at one of the first commercial database companies, Ingres. He spent the next 20 years marketing database products, building the analytical technologies, data collection, data interpretation and visualisation products for Ingres and then Oracle. He also ran Oracle’s database business for seven years encompassing engineering, product management and marketing.
Upon joining Palo Alto Networks, Bonvanie had the opportunity to set marketing up from scratch in advance of the release of the vendor’s inaugural product.
“It was the first opportunity in my career to set up a marketing organisation with a very strong analytical footprint,” he says. “I knew once we reached scale, we could highly benefit from that.”
For Bonvanie, the first thing is to collect the right data on customer interactions.
“Classically that has meant purchases, and you expand that to opportunities and leads. But there’s much more,” he says. “For example, the average company will be in contact with a person for many reasons – exploration, consideration, negotiation, transaction, help and support. All of these reasons can be put into a model. Then you can start to collect outcomes from these interactions. For example, did those conversations lead to something?
“You can also start to understand environmental elements. For example, size of company, the team, size of their purchase, what is the competitive playing field. Then you can also start to mine what else they are considering.”
Bonvanie says these signals can be ingested and used to make actions for other activities, such as coming up with list models, or building groups to start to see what customer affinity exists and what the relative importance is of certain parameters.
Data-driven marketing in this sense requires skillsets not typically brought together, Bonvanie continues. On the one hand, you need data scientists who can build and test these models. You also need people that can do micro actions on these findings within minutes, such as augment a media buy or engage in segment marketing or microtargeting.
“That’s a different skillset – it’s much more transactional than we see in building long-term programs for long-term consumption,” he comments.
Across Bonvanie’s 150-strong marketing team worldwide, 10 per cent are in analytics and focused on figuring out models to make them operational. These are then automated into predictions, scores on activities or insights by a team of 15 staff. The rest of marketing runs programs off the back of these.
“It’s one thing to have lots of data, and analytics, but if you don’t operationalise that all you have is lots of data and lots of analytics,” he says. “That’s useless, because you need to operationalise so you can trigger actions that cause people to buy more. That’s all we want to do in the end – have people buy more.”
The problem with many marketing interactions today is that they do the opposite, Bonvanie claims.
“The reason Australia and many other countries have anti-spam laws is because marketers went crazy,” he says. “We now know there’s a very high negative correlation between the amount of contact and the appropriateness of the message, and the likely outcome. In other words, you need to know when to say things, what to say, and when to shut up.
“As marketers, we’re not used to knowing when to shut up, or exactly when to trigger things. We trigger things because we want to trigger them. What data science does is the exact opposite – it makes you much more thoughtful. And in the eyes of the customer, if you do it right, they won’t believe you’re contacting them at that moment because they were just about to do something.”
Getting the right metrics
According to Bonvanie, the reasons so many marketers are still struggling to do this are three-fold. The first is the mission of marketing in companies and the metrics used.
“In many companies, if you asked the CMO if there is one metric they feel responsible for, oftentimes they struggle,” he says. “In my case, it’s always been the pipeline number. The number of leads or contacts are all weak proxys to the numbers, which is the pipeline we need to make our quota.”
Bonvanie says pipeline is a leading indicator of success in the next fiscal year and can be predicted based on activities.
“It’s less about the amount of money I spent on marketing, and more about the effectiveness,” he says. “For example, if I measure the number of leads, which most CMOs will say is important, it’s the direct result of the amount of money you’ve spent. It doesn’t tell you much about effectiveness of leads or the activity.”
The second thing marketers suffer from is a massive data management challenge.
“There’s lots of data in lots of places and lots missing. They don’t have the historical data, or if they do, it sits in multiple systems,” Bonvanie says. “It becomes a massive IT project and that’s not very encouraging, because most CMOs are not IT specialists, or come from a database background, they’re more branding, advertising and communications. For them to define what these models should look like is a challenge.
“But this is changing rapidly, and I see many CMOs with almost as big, if not bigger, IT budgets than anyone else in the company.”
This is the case at Palo, where Bonvanie has more IT projects than any other department in the company.
The third reason for the struggle is skillset, and Bonvanie admits this is the one that will take the longest to overcome.
“The classic consumer CMO, with a strong brand and comms background, understands this very well, but for the B2B CMO, it’s not been their daily fare,” he says. “There’s an incredible hunger, but they don’t know or where to start.”
From CMO to chief digital officer
The status quo is changing, however. Over the next few years, Bonvanie expects to see the CMO evolving to chief digital officer, managing the majority of data concerning customers. That also means owning a bigger part of the IT budget than the CIO. And that’s what’s causing the relationship angst between CMO and CIO.
“The agility marketers want from the system and self-service they want to profess oftentimes doesn’t sit well with the IT department,” he says. “But that relationship has to be rock solid, for the simple reason of compliance.
“Ultimately, systems used to make data available and operational will mean the CMO is a bigger user and buyer than the CIO. The CIO goes to the background for those systems, other than making infrastructure available if needed.
“The other trend is that CMOs are now buying IT systems… that all run in the cloud. We’re wall-to-wall Adobe from a marketing perspective, for example, and it’s the same for analytics, data sources – they’re all in the cloud. This makes us very agile, but there are also challenges as it’s very confidential data sitting in other places. But I would always take those compliance challenges over not having the software.”
Examples of data in action
As an illustration of how data is being used to drive marketing at Palo, Bonvanie points to the vendor’s annual user conference and efforts to get the right people in attendance. To do this, the group implemented a score on everyone in the database and their likelihood of coming to the event. From this, it undertook specific marketing actions, including a concierge service for the highest ranking people based on its score.
“These people hadn’t shown up to our conferences before. It was around the way they interacted with us and characteristics of the companies they worked for, plus the opportunities,” Bonvanie explains. “That model scored significantly better than random, and even better than our own estimations.”
A second example is opportunity management and in particular, the point where a potential customer has done a technical evaluation of Palo’s products.
“In any given quarter, about 60 per cent of opportunity we have sits in that stage,” Bonvanie says. “There’s so much of it, it could potentially drown the sales reps, as they wouldn’t have any idea which ones have the highest likelihood of success… But we don’t need all of that to succeed. In that blue ocean, we look at which ones are the true gems we should absolutely get. You only see they’re wildly different when you start to do correlation analysis on these opportunities and predictive modelling.”
This kind of data-driven scoring has been done over the past 18 months and has a reliability of over 97 per cent, Bonvanie says.
“The relationship we now have with sales team is they see these almost as prescriptions, and will ignore the opportunities that don’t appear on their list to pursue,” he adds.
Advice for getting ahead with dataBonvanie offers several key pieces of advice for those struggling to get their data analytics in order:
Focus on lead scoring: “We look at things we do in marketing and see which have the highest propensity to involve into an opportunity,” he says. “There are techniques out there to do that; marketing automation vendors bring them too but there are also third parties that do that. Most marketers are focused on lead gen, so this is a natural next step.
Start to understand conversion rates in pipeline: “It’s straightforward math, but later on, it’ll start to educate more sophisticated models as to things that worked, and those that didn’t.”
Forget about demographics, who are these people and companies: “Marketers typically buy lists of names of supposedly important people in companies. These are not useless, but they’re not very informative. From an analytics perspective, you need a lot more data that’s behavioural and environmental to do advanced modelling.”