Over the past three years, Spark NZ has advanced its approach to data-driven marketing with the development of a world-leading AI platform integrated into the company’s MarTech stack. This platform is now delivering 10x improvements in campaign outcomes. We speak to Matt Bain, Marketing Director for Spark NZ, to discover how the company has achieved these outstanding results and what the future holds.
With the vision of creating advertising that feels more like a service, and drawing from his extensive experience overseas, Matt Bain and his team found a MarTech solution that not only worked, but exceeded expectations.
This meant no more interrupting content to try and sell someone something and instead focusing on anticipating people’s needs so the company only sends messages when they are relevant to the individual or the household.
In a market such as telecommunications, Matt says it’s likely that only two percent of the whole country is actually ‘in the market’ at a time for a good or service that Spark can provide.
“Doing traditional advertising to a country where two percent of people are actually in the market is pretty inefficient. Ninety-eight percent is just promotional work. If you’re talking about brand consideration it still works at that level but once you get down to lower funnel impact it becomes really wasteful,” he says.
To overcome this issue, Spark had to look at its database through a series of what Matt describes as ‘lenses’, such as a product, customer and household lens.
“A lot of our services are relevant at a household level, so what is each individual households situation and needs?” he says.
Once all this data is in order and at scale, Matt says it’s all about how you predict when the customer is in the market. This is where machine learning comes in.
“Machine learning models basically run off the rich data we have around customers. And they are able to predict when the customer is in the market. They can predict with 80 percent accuracy if [you] are in market right now for a new phone.”
This means customers will only get a message if they have been deemed likely to be looking for a product or service that Spark can provide.
“Doing that, we are only targeting you when one of our models tells us you are in market. We generate a 10x improvement in conversion rates through the channels we are using.
“Companies like us have a lot of first person data and by analysing large numbers of customers at scale, models start to learn what a customer looks like and they start to learn the certain types of customers, given certain types of profiles. If I know your mobile phone is three years old, and you upgrade every three years and I know you have an Apple device then you might flag in our models as a really high-propensity individual.
“These are one-to-one segments. We do not create a segment and then target a segment, we target an individual customer. Usually that’s impossible, but with AI you can do it across millions of customers in real time. We call it an AI brain. It’s constantly scanning our entire base and looking for a customer that it flags to be in market, not randomly.”
This level of scalability is at the very forefront of what MarTech can achieve and beyond what is humanly possible. “The big shift is that no human can do that,” Matt says.
“It’s millions and millions of rows of data on a spreadsheet. The only way we can do this is with machine learning.”
Spark NZ has a number of models running across its customer base and Matt says this is where AI shines.
“Once you have 50 models for the whole of New Zealand all in real time and they have to decide which offer to send which customer, that’s when AI really comes to the party. It could be too that a customer or a household is in the market for multiple products at a time, something or someone has to decide which offer to send them.”
He says Spark was initially focused on lower funnel conversion – meaning the sales promotion work – which New Zealand sinks a lot of money into.
“Brands spend millions and millions on that and by connecting the return on that investment you create really massive impacts for your customers and in the business.”
Spark found that customers were more responsive if they targeted them less and customer satisfaction goes up at the same time.
“We are polluting them less, we send them less emails, we send them less app alerts. We drive up customer satisfaction at the same time as we drive up sales which makes intuitive sense.
“From a mid and upper filter perspective, like consideration and brand strength, the next step is to integrate all of your advertising into a platform like this.”
Currently Spark is the only company in New Zealand that has fully in-house programmatic and social search stack. All the company’s marketing is done by an in-house marketing team with in-house media buyers and production teams. This is so they can “quickly and efficiently spin up an optimised campaign,” says Matt.
Currently this is happening manually but Matt dreams of the day when this element of the process can be connected to AI as well.
“When you talk about MarTech most people are talking about the Adobe’s or Salesforce’s of the world but actually, your programmatic marketing stack is a dominant channel for digital marketing these days.
He says bringing this in-house creates a data driven marketing team.
“You have this big team working on data in real time on advertising and search. It’s as much a cultural shift as it is speed to market and efficiency as well. What that gives you potentially is a full funnel marketing capability that’s run by an AI system not by humans. “The cool thing about that is that it frees up humans to do much more creative strategic work.”
As many marketing jobs require quite repetitive day-to-day tasks such as creating emails, sending campaigns or writing media, Matt says AI is useful in that it can take over these menial tasks so “humans can focus on the much more fun exciting and valuable things”.
As for the other ways Spark NZ has incorporated MarTech into its tech stack, Matt says the company has “big global, best in class stacks” including Adobe, Salesforce and world-class data housing and data transformation cloud-based application Snowflake.
“What we found is by building our own proprietary AI capability across the top of all of that including our call centres and our stores, you actually get a multiplier effect of all those systems working together and that’s where we have now got to.
“We are doing proof of concepts where that’s all joined up and seeing 20x improvements in results which is kind of unheard of and more than I expected.”
This, he says, shows the power of bringing all these systems together. It’s here that privacy and data security become increasingly important.
When Spark was developing this system the team had to reconsider all the company’s privacy values and built a brand new state-of-the-art privacy framework that supported AI capabilities. Human capital and tools were then built around that.
“We created a privacy ambassador role in the business and we trained privacy ambassadors around our privacy principals and frame work,” Matt says.
“Then we created in our tools where any marketer is using data for a purpose they have to submit it for authorisation to make sure we only use data in a way that customers are happy for us to do so. The great thing about AI is that it can do anything but left up to its own devices it could create bias in those models and misuse data. We’ve got really stringent processes that stop any misuse of data.”
Matt says this isn’t very common in New Zealand but believes it is crucial for success and for retaining customer trust.
“No one in this country has a brain like we have so consequently no one has had to build a privacy framework like we’ve built.”
This article was originally published in the September/October 2022 issue of NZ Marketing. Click here to subscribe.
