What you need to know:

• Compare Club, Southern Cross Audio, and the University of Tasmania have persevered and ultimately prevailed in their respective customer data platform journey all of which are now several years deep.
• First the good news. Though three very different types of businesses opted for different solutions (A composable CDP,  Salesforce Data Cloud, and an Australian-first Oracle implementation, the common factor is, happily, success. Acquisition costs are falling, engagement is up, and day-to-day customer engagements are much more stitched together making it easier to extract behavioural insights.
• And the work needed to get the data flowing through the pipes puts all three organisations in good stead for the next generation of AI-driven use cases that will soon inevitably emerge.
• But like all complex IT implementations, there were plenty of challenges, including many that were flagged in our original series of investigations – not least among these the troubles of data integration. 
• These aren’t just technical issues, it’s essential to have all the right people at the project table to deliver the best outcomes.
• Lessons learned? That’s the kicker, these brands typically learned more about themselves than the vagaries of middleware implementation.
• Read last week’s piece for more insight from brands, publishers and vendors.
• Listen to the podcast here.

The University of Tasmania’s plans to optimise digital advertising took something of an unexpected detour when their DMP provider Oracle announced in the middle of a CDP implementation that it was exiting the adtech business. Over at Compare Club, Rich McPharlin decided to ignore one major CDP industry trend – handing back control of data architecture to the IT department while embracing another, the rise of composable CDPs. And SCA set itself the goal of better orchestrating and segmenting data for both its advertising and audience customers with a platform that was more attuned to ecommerce environments than media channels.

Such are the challenges brands face when they look to implement customer data platforms, still one of the hottest techs in the martech stack.

As Mi3 reported last week, the CDP market in Australia remains strong, despite problems like bill shock, API ‘blowouts’, and the inevitable disappointment of contract renewals when decisions come to be based on the realities of often complex implementations, rather than the sizzle of those early sales decks.

In this week’s Mi3 Audio Edition, we interrogated those topics in conversation with Courtney Geritz, Director, Marketing for the University of Tasmania, Cam Strachan Head of Data and Analytics for SCA (and in whose studios we recorded the podcast), and Rich McPharlin, the chief customer officer of Compare Club.

CDP rationale

The good news is that in each case the now maturing CDP programs are delivering against their initial business case rationale.

McPharlin says they were certainly chasing intangible benefits like speed to market and an acceleration in data activation but he prefers to anchor the conversation with tangible outcomes.

“When we decided to take on the composable CDP solution we bought we had a very simple business case. We spend a lot of money on marketing, and we have a very large sales team, and it’s a pretty simple process for customers end to end,” he says.

What we wanted to do was use some of our existing data and information to better find the right customers at the right time and to serve them more appropriate messaging based on what we already knew about them.”

The payoff? Compare Club was able to reduce the cost of acquisition by about nine and a half per cent on one of its key segments. “When you consider marketing to be our largest cost of the business by far at tens of millions of dollars, stripping nine and a half per cent off that very quickly hits the bottom line and makes the CFO very happy.”

For the University of Tasmania, the goal ultimately was about stripping complexity out of a very complex organisational environment with lots of different data sets, says Courtney Geritz, Director, Marketing for the University.

“The CDP really enabled us to synthesise all of this rich information from a number of different systems so that we could customise the user journey from end to end. And what that looked like for us in real terms, was a huge saving in our cost for acquisition.”

In the last year alone the University has achieved a reduction of 20 per cent which Geritz describes as “incredibly significant when you think about the time and energy and investment that you spend on attracting these students and then ultimately getting them through to enrolment and into their current studies.”

She told Mi3, “It’s really been invaluable in making sure that we can actually create communications for them in real-time that are tailored to their needs.

For SCA’s head of data and analytics, Cam Strachan, the CDP project was ultimately twin-pronged.

“We are a multi-brand publisher and equally, we have a marketing funnel as well.”

As a marketing tool, “The CDP allows us to use targeted data for acquisition optimisation [into master app, Listnr]. Equally, as they’re in our environment listening to our content, we want them to stay, so we have use cases around churn reduction, and we use it for machine learning, and also for engagement as well through personalisation and content recommendations”.

As a publisher: “The CDP allows us to segment once and use them both for the marketing use case and the commercial use case.”

Among the current use cases, “SCA uses the first-party data for our Advertiser Match, and we also use it for advertiser targeting in our audio and visual display ad server as well. So it’s very valuable to have that common layer of segmentation that we can use across multiple use cases.”

Problems worth solving

Marketers at University as Tasmania were battling with five discrete data systems, and trying to triangulate them down to a single student across the myriad data streams across applications like the CRM, and the email marketing platform, but also including some CX sensitive applications like the student management portal.

According to marketing director, Courtney Geritz: “We were creating disparate profiles for each of our students and we were never really getting a holistic look in terms of that end-to-end experience. And I know this sounds fluffy but ultimately, the thing that we’re always thinking about is, ‘How do we want them to feel at the end of the process?’ We want them to feel valued, and we really want to make sure they don’t have to enter their data and re-prosecute their inquiry time after time.”

As anyone with a kid at university knows, anxiety and drama are always on the curriculum, or as Geritz more elegantly puts it, “It’s a very significant decision in anyone’s life and we really want to treat it as such.”

The CDP helped solve the problem by allowing the university to synthesise all of the information about a student in its systems and create what Geritz calls a Golden Record.

“We can then use the behavioural insights about our students to customise communications and to meet their needs, and [we can] create interventions as well throughout their study journey. So we did look at other solutions”

She says the Uni assessed a variety of approaches including a data warehouse and also what she describes as “amplifying our current CRM”, but ultimately, none of those provided the marketing team with the real-time value that the CDP did in allowing it to asses behavioural insights in real-time, and proactively change the way that it communicates.

The new approach also let the team examine “the way that we market and, of course, look at the affinities of our students so that then we could create our lookalike audiences, and target those students that have that higher propensity to come and study.”

SCA’s Strachan meanwhile flagged a learning that is common in many CDP projects – underestimating the integration requirements with the development and product teams. (In fairness that’s common to a lot of Saas projects!) 

According to Strachan, “Obviously, when you’re integrating into operational systems you need to slot into the timeframe of other development that’s needed as well.”

He stressed the importance of understanding and working with those teams, and to do that early in the process.

The other issue he flags is that of people and capabilities.

“With some of these newer technologies and the newer views of data, you need data-driven and digital-driven folks in the business.”

That’s an insight that has seen an evolution in the way the data analytics team communicates with the business.

In the early days, information and insights were shared with everyone, but now there is more focus on key ‘power’ users.

For Compare Club a key insight that led to change according to McPharlin “…was that there was a small group of customers that we talked to who were very loyal, who we knew a lot about. We could take that information and activate those customers far more effectively.”

That can happen through first-party platforms like email, SMS, or even phone calls or it can be a case of finding the right segments to target on third-party platforms, like Facebook, or Google, or other digital ecosystems.

“We hypothesised that we could make our marketing investment far more efficient. And so ultimately, that’s what we did.

Just as important was the speed to value. “It was a relatively quick turnaround for us, because we had a real idea as to the size of the prize up front, and we’re focused on that one specific use case. We’re able to get the program across the line very, very quickly and realise that value almost instantaneously.”

Challenges

When Mi3 first took a hefty dive behind the CDP landscape in mid-2023, one of the most common difficulties to surface out of the customer community was the difficulty brands face integrating systems.

CDP implementations are replete with difficult in-between moments marketers must navigate, and it’s much messier and more complicated than the slick sales pitches let on: Fights over data ownership, significant technical integration problems, and serious limitations with all those out-of-the-box connectors the vendors like to brag about create risks to project deadlines, and in the worst cases, career advancement. Sometimes just identifying who has credentials and getting them to share can take months.

Geritz, McPharlin, and Strachan are mercifully free of “the middle” but they experienced the kinds of problems their peers have previously flagged.

According to Strachan, “The integration nuances were some of the challenges. It’s all well and good to say you’re going to connect all the data coming in and then push it out to the different channels. But the reality is, each of those [channels] has their own nuance as well, and you’re dealing with a lot of different data sets coming in, some coming in real-time, some coming in batch, some being fired off from different systems as well.”

That’s when you start to appreciate the challenges of orchestration he says.

“You can run into challenges with web data being slightly different from app data, things like that. CDPs are generally an integrative technology, so you’re always integrating with the execution engine or with the data sources themselves.”

He says these aren’t always challenges you can plan for. “However, you do need to have a pretty strong architecture as well. And one thing I do like about the Salesforce Data Cloud is that it comes with its own data model. Regardless of whether you know what type of CDP you [will] choose, you ultimately have to have a data model, and you really have to understand that data model.”

He says this is an important issue for business users, because “You don’t want your data engineers or data analytics [people] to run your marketing campaigns.”

You should also expect to find yourself basically having to shoehorn some data into that data model as well, he cautions.

If he had his time again? “I think probably spend a little bit more time with architecture understanding what the data actually looks like. You can say you have one person to multiple devices, however, the way each of those devices is named and so forth can be different. So you’re always running into data integration challenges as well.”

The good news is you can inoculate yourself against the worst of these problems, Strachan suggests.

“Above all, I think we’ve done a great job in terms of the governance of that data. And it’s definitely helpful to have a SaaS model in terms of governance because it really helps for that user access and locking down data that you don’t want people to see as well.”

Geritz likewise lived through the perils of CDP integrations and came out the other side, if not unscathed, at least better educated.

“We had five different disparate systems, the ideal for us was to actually stitch them all together. So some of them were based in real time. Some of them were batch, as Cam had mentioned. So the goal for us was having that holistic view in one single place.”

Prior to the CDP implementation, the student experience was highly fragmented from end to end.

“The challenge was we had five different profiles for a single student as well. Not only were they represented five different times in five different systems, but they may have been receiving communication and marketing from us up to five different times.

“The message fatigue and the disparity between one end of the business to the other was really palpable.

Eventually, it was all stitched together although she acknowledges “There is still some work to be done.”

“The more systems we ingest, obviously, the more complex it becomes. But for us, it was about making sure that the model was sizable.”

It took a lot of trial and error to get it right and to make sure everything lined up.

According to Geritz, “It’s always about the data integration. And when I think about the number of systems that we were working in, it does give me a little bit of PTSD now, thinking about just even mapping the fields and bringing that all together in a really coherent way.”

“And of course, when you’re working with so many different stakeholders. Everybody’s data or every piece of that journey, is the priority. I think upfront one of the things that we really had to work on was bringing the right people to the table. I can happily say I have become best friends with our IT team because that was critical for us in balancing the needs of marketing, but also balancing the needs of our student data and how we would ingest that.”

She says a rule-based approach to ingestion helped when it came to which data was going in and how it was cleansed. 

“But we also found that we needed to really establish a CDP technical expert in that space as well because that was the piece that really brought it together by working with different stakeholders across the network and then tangibly piecing it together.”

McPharlin’s experience seems to have been a little easier, despite taking on the potentially extra burden of a DIY CDP, as composable solutions are sometimes called.

He said, “What I love about the composable CDP is we can choose the best in market solutions for every component of the CDP. If you have to use a single off-the-shelf solution, you’re going to make compromises here and there across the board.”

“Our original biggest challenge was that we run two Salesforce instances, and the background there is that we have two legacy organisations that merged. But we never really did the work to merge the Salesforce instances, and so our energy and health insurance business live in one [instance], and our mortgages and our life insurance business live in the other.”

There is a lot of value to be gained by bringing those two data environments together but it was not a a seamless process.

“Everything down to customer statuses, for example, is notated and annotated entirely differently. So while the CDP implementation itself was very simple, it was the upstream work that happened before that, which we were able to tackle in an iterative process that was most important.”

Compare Club worked with an ELT (Extract, Load, Transform) provider named Fivetran, and then built all the custom modelling in Snowflake. Hightouch meanwhile is used to get the data out and across to systems like Facebook, Google, and Braze where campaigns are activated.

“One reason we love this process of the composable CDP is we don’t suffer from the vendor lock-in issues that some organisations have where every single part of the stack is owned by the same third-party vendor. If they get a bit greedy, they can jack up the prices on you materially.”

He says the composable approach helps the firm to get really strong pricing points within a very short period of time. Plus, “We can switch out our ETL provider, we can switch out our CDP, or we can switch out our database without getting those barriers.”

Teachable moments

Having navigated the vagaries of vendor choice (or ignoring it altogether) having tamed the silos and worked the integrations and having ultimately generated the kinds of returns that make the troubles feel worthwhile, we asked each of our guests what they learned.

None of the most important lessons were about the CDPs themselves.

For Compare Club’s McPharlin, whose approach to data analytics feels like an art form, it’s important to remember that marketing has always been a balance of art and science.

“I think marketers over the last five years have gone a bit of a science binge. I started as a performance marketer, paid search, paid social, etc. And the way you won in the old days in that landscape is you were the most fine-tuned. You squeezed every lever, you turned every knob to get the last 10th of a percent out of everything.”

That’s where machine learning is leading us, he says.

“We as performance marketers need to put more marketing back into our tool kit, particularly as these tools and the technology become even more commoditised. What we’re finding is that the accessibility to them is far more open, and therefore their distinction and their creation of difference in an organisation is less. We need to use them as time savers and efficiency creators so we can focus on who our customers are, what they think, what they need, how we communicate that to them, and how we serve them most effectively.”

For SCA’s Strachan, the issue was not necessarily the issues that get raised all the time, like skills and capabilities, which he acknowledges nonetheless are important.

Instead, “It is around the interpretation of the information. Some of these systems are geared towards e-commerce. We’re a media business, so we just had to change, probably more from a technical nature the way the data is stored. But it hasn’t especially changed the way we’ve behaved in terms of the use cases that we’ve been able to drive and that’s been a good outcome. But probably the difference between us as a media company versus an ecommerce – a little bit more extra work to understand the data model and how it will be used by the business users.”

And finally for Geritz, “The biggest thing for us is the quality of the student experience. It was really important to show that in metrics because ultimately, that’s what the organisation is looking for. It sounds all nice on paper, and it sounds good in theory, but what does it actually mean for the student?”

She gives an example of how students engage with emails. Previously there was a huge amount of drop off because students were receiving so many different communications from so many different areas to so many different profiles.

“Over the last 12 months, that email open rate has increased by 20 per cent unique click-through by 10 per cent but most importantly for us, the unsubscribe rate has dropped below 23 per cent so everything now is above the industry benchmark. Ultimately when you put that in front of the C suite, that’s what counts,” says Geritz.

“It was really important to make sure that that narrative revolved around how it actually looks and feels for our student, because obviously you can break it down into dollar values, but ultimately what we want for the student is for them to be more engaged throughout the life cycle of their studies with us.”

Ultimately it’s not about the tech. It’s the customer, stupid.