What you need to know

• Twilio, which owns the Segment customer data platform (CDP) - the most popular CDP in Australia by installed based, considered exiting the global CDP market due to disappointing numbers post-acquisition.
• The company purchased Segment in 2020 for US$3.2 billion during a CDP boom, but integration challenges saw declining tech valuations in 2022 and 2023.
• New CEO Khozema Shipchandler ordered a strategic review that determined the importance of the CDP in the context of rising AI and conversational commerce.
• Thomas Wyatt, appointed president of Twilio Segment, emphasised the need to integrate Communication platform-as-a-service (CPaaS) and CDP technologies to enhance customer experiences.
• The strategic review revealed areas for faster integration and execution to improve product offerings.
• The decision to stay in the CDP market aligns with the increasing importance of generative AI and the need for businesses to orchestrate customer interactions across multiple channels. It also aligns to Twilio's pitch around conversational commerce.
• Twilio recently introduced the Personalised Virtual Agent, which combines traditional IVR functionality with data from the Segment CDP to enable real-time customer interactions.
• And Wyatt says the integration of personalised content leads to improved customer experiences and outcomes, such as effective call deflection and tailored recommendations

In February this year Twilio, owner of the Segment customer data platform - which is the most popular CDP in terms of installed based in Australia - seriously investigated abandoning the CDP market. Despite buying Segment in 2020 in the midst of a CDP boom, and at the start of the most aggressive acceleration of digital transformation ever thanks to pandemic lockdowns, the numbers just weren't adding up.

That may have something to do with the US$3.2 billion price tag and the fact tech valuations took a pounding in 2022 and 2023 generally.

But the strategic review of the CDP business ordered by new CEO Khozema Shipchandler a month into his tenure arrived at a very different conclusion to that reportedly pushed by activist investors. With AI and conversation commerce on the rise, a CDP is the core infrastructure in a world where data orchestration is central to commercial success.

Mi3 spoke with Thomas Wyatt at the recent Twilio Signal conference in Singapore. He was appointed by Shipchandler to the role of president of Twilio Segment in March, once the strategic review was completed, and tasked with overcoming the inertia holding the CDP back and driving the business forward. It helped that, as a former customer, he brought a real-world perspective to the role.

He started by saying the vision for bringing Communication platform-as-a-service (CPaaS)  and CDP together three years ago is still the right vision today (CPaaS is Twilio's core business), since these these types of technologies will converge to deliver much better customer experiences.

"We may have been a little before our time in terms of the way customers were thinking about consuming the functional services that that platform could provide. And Segment independently was growing at a relatively strong clip as was the communications business during the pandemic," Wyatt told Mi3.

That said, the review exposed some of the issues that led to the strategic review. "There were some areas where we could have integrated faster in terms of the product execution. For instance, how the architectures could come together to unlock more of these bundled solutions," he said.

"One of the lessons was the notion of trying to take a joint value proposition to market without the core building blocks of the foundation being fully integrated. That was a challenge for us and we had to find a more efficient way to execute that."

The decision to stick with the CDP also reflected some big changes in the market, starting with generative AI.

"Two or three years ago that whole paradigm of thinking about rebuilding applications, of agentic frameworks, thinking about proactive, prescriptive interfaces that are different than workflows, was even a real conversation," Wyatt said.

To take advantage of technologies like generative AI, having a data foundation is critical. "What Segment is really good at is the ability to collect data from multiple sources, to create identity resolution from it, and then activate it upstream for personalised communications or call centre use cases or marketing use cases," Wyatt continued.

"Now that we've been able to build those capabilities in a more integrated way, our go-to-market motion is evolving as well, and our customers are starting to pull us in that direction."

Those customers, according to Wyatt, recognise the need to orchestrate across multiple channels and to personalise experiences on top of that. "Over the last year we've really optimised and improved our innovation and to ultimately bring those things together."

Conversational commerce

Apart from generative AI, the other key consideration for Twilio in remaining in the CDP market is the company's belief in the growing importance of conversational commerce.

"Customers want to engage differently with brands. They log into your website and they may have questions about a complex purchase and want to understand more about it. Or maybe they have a returns issue or something, or another problem that needs to  be addressed immediately," Wyatt commented.

It's the kind of use case that led the company to recently introduce its Twilio Personalised Virtual Agent. It is an example of the critical role of the CDP in business for Wyatt.

"[The agent] is a combination of a traditional IVR functionality and the unified profile from the Segment CDP. We've collected the real-time data warehouse and communication data from all the past, and [data on the engagements with an end user. When you combine that with a large language model you're able to have real-time conversations, for instance, pre-sales trying to get people over the line, or post-sales helping them solve a problem once they've had one."

Likewise with call deflection, Wyatt said. "We've seen it over and over with customers, but what's interesting is they're actually getting pretty positive call deflection, even with models that are just trained on basic FAQs or simple content that might be policy around a particular return, for example.

"But when you start combining it with the personalised content from profiles and data warehouses and all this other stuff, then it becomes much richer. You can be more prescriptive in recommending the next thing somebody might want to do, or get them to do it for them in advance. And that's directionally where we are headed."