Interview: Rahul Todkar, head of data and AI, Tripadvisor

Planning the journey

Todkar, who previously held senior data leadership positions with LinkedIn and Charles Schwab, says one of his key achievements at Tripadvisor is a data strategy for the business focused on three core elements.

The first – foundations – is focused on the technological underpinnings that enable the data strategy. The second element is driving incremental growth that’s measured through clear metrics. The final element is to embrace innovations that can lead to monetisation.

“We’ve made tremendous progress across all three areas of the strategy,” he says. “One of the big initiatives is moving to the cloud via Snowflake. We had data in our on-premise environment, and we’ve moved all of that over to Snowflake in a cloud environment. That shift has been a massive game-changer.”

Tripadvisor wanted to bring all its enterprise datasets together in one location. The company uses the Snowflake AI Data Cloud for Travel and Hospitality. The firm uses the platform to provide service redundancy and data protection while ensuring compliance with privacy regulations, such as the General Data Protection Regulation (GDPR) in the EU.

Todkar says getting solid data foundations in place via the Snowflake technology has made it much easier to create innovative customer services. The platform helps the company to unify enterprise data and analytics across products, marketing, operations, customer insights and finance operations. Todkar says that a single source of truth is crucial for modern businesses, particularly ones such as Tripadvisor, which collects huge amounts of information.

“Any company with massive data collection will end up at some point – if they’re not thoughtful from the start – with significant data sprawl. That’s exactly where we were,” he says, suggesting the company had as many as 10,000-plus data pipelines and 30,000 to 40,000-plus datasets.

Todkar says the decision to work with Snowflake pre-dated his arrival at Tripadvisor. However, the company takes a consistent approach to technology. He says senior professionals explore the solutions on the market and use proof-of-concept studies to find the best solution for the business challenge.

The team was impressed by Snowflake’s capabilities for its initial use case and continues to expand the platform’s footprint to additional areas. Todkar advises other digital and data leaders to work with a trusted partner who can cover as many facets as possible.

“Converge very quickly rather than diverging,” he says. “Look for consolidation of technologies around some critical key partners.”

Exploring new opportunities

The technological foundations his team is putting in place mean they can support the second core element of the data strategy – incremental and measurable business growth, including via emerging technology, such as AI.

Todkar says the business has successfully launched generative AI use cases in production. That’s a big difference from many other firms. He recognises CIOs often talk about GenAI as a concept and technology their organisation is exploring, but real-life exploitations can be thinner on the ground.

Most companies struggle to move GenAI projects into production, according to Deloitte. The consulting giant says that 70% of business leaders have moved 30% or fewer of their experiments into production. Tech analyst Gartner, meanwhile, predicts at least 30% of GenAI projects will be abandoned after the proof-of-concept stage by 2025.

That’s not the case at Tripadvisor, explains Todkar. “We’ve had some things in production for nine months, including our Trip Planning tool,” he says. “In terms of our engagement and some of the biggest strategic successes, we’re seeing big user adoption.”

“We had data in our on-premise environment, and we’ve moved all of that over to Snowflake in a cloud environment. That shift has been a massive game-changer”

Rahul Todkar, Tripadvisor

The company’s Trip Planning tool uses large language models and generative AI to make personalised suggestions to travellers. Todkar says the technology stack powering this tool includes customer data held securely in the Snowflake platform, off-the-shelf commercial models from OpenAI and internally created recommendation engines.

“All of us want to travel. However, when we travel for business, leisure or with our families, we know one of the common pain points is planning a trip and getting the best guidance. That’s a challenge for all of us. There’s often a lot of painful conversations and planning before you go on a trip. It’s an active and evolving planning exercise,” he says.

“That’s where Trip Planning helps. With generative AI and large language models, we take the information on our users, use that insight in a privacy-compliant way, and combine that data with additional information we collect. We then present and recommend the trip-planning solution to our travellers thoughtfully and contextually.”

Todkar says his team is looking at other uses for GenAI, including summarising text-based reviews on the website. “If you are a consumer of any information, you often don’t have the time and mental capacity to read through all this text,” he says.

“More things are coming. However, I am thrilled to share that we’ve already launched services and done so much work across our three strategic data pillars during the past 18 months.”

Supporting great experiences

As a modern data leader of a technology-enabled business, Todkar keeps one eye on the future. He searches continually for innovative ways to apply data and give his firm a competitive advantage. However, that horizon-scanning process doesn’t occur in isolation from the rest of the business.

All his team’s innovative ideas are tied tightly to business objectives – and 24 months from now, he’d like further progress in the three core pillars of his data strategy: foundations, growth, and innovation, adding: “I would love to see massive progress in these areas.” Todkar turns first to foundations.

“On data-specific issues, like storage, compute and access, we have made lots of progress,” he says. “We’re thinking a lot more now about machine learning operations [MLOps], large language models, and how to create a robust and scalable infrastructure for deploying AI and machine-learning applications. Hopefully, in 24 months, we’ll have seen some solid progress in that area.”

As his team continues to hone the firm’s data platform, Todkar anticipates further growth and innovation. “I would love to see the data organisation taking advantage of all these innovative technologies,” he says. “But we must link that innovation to the objective of driving business growth by launching multiple AI and analytic applications.”

In the longer term, Todkar believes his team can use emerging technology to create highly interactive, multimodal user experiences. He says ChatGPT has already helped to reframe user experience as bidirectional and dynamic rather than a static, one-way information exchange – and there’s more change to come.

“If you think about search, which has traditionally been about using Google, you type in certain things and get results,” he says. “It’s a very one-directional experience. It’s also not interactive enough, which is why ChatGPT – and other GenAI tools, such as Google Search Experience – are helping to create an interactive experience. I think travel planning and search could be similarly re-thought.”

Reaching the destination

The big challenge many digital leaders face when trying to innovate is ensuring everyone across the organisation sees the advantages of an investment in new technologies.

As an experienced executive, Todkar has expended considerable time and effort during his career to get people excited about data-led change. He says that one approach is the key to success – concentrating on business outcomes.

“I’ve done many transformations and always start with the big ‘Why?’” he says. “I ask people, ‘Why are you trying to do this?’ I tell my teams to spend lots of time explaining why they want to do certain things. Fall in love with the problem, not the solution.”

While focusing on business outcomes is crucial to success, Todkar says too many technical teams still fall in love with the solution: “They’ll say, ‘Hey, I’m going to do this with the latest large language model. I’m going to go and pilot it.’ And I’ll ask, ‘To what end? What are you trying to do with that model? What’s the business benefit?’”

The good news, says Todkar, is that justifying an investment in data technologies is much easier once you’ve got people on board. “Once you define the business benefit upfront and explain the exact problem you’re trying to solve, everything else becomes straightforward,” he says.

“It’s the same with digital transformation. Once people understand what you’re trying to do, and you define that solution with the right set of cross-functional stakeholders, then the transformation – and everything else associated with that process – falls into place.”