‍"With Signals, we can understand the deeper context of what a user wants from the product and facilitate an experience that delivers it to them. That's something we've wanted to do for years, and now we have the infrastructure to make it happen."‍

Anup Purewal, Chief Data Officer, DC Thomson

Background

With a mission to help people discover their family history, FindMyPast serves millions of genealogy enthusiasts through its extensive collection of over eight billion historical records, more than 70 million newspaper archives, and a suite of family tree-building tools. Owned by DC Thomson, the product supports a diverse user base ranging from casual hobbyists tracing a single ancestor to dedicated researchers exploring extensive lineages across generations.

Driven by advances in AI, the rise of hyper-personalized digital products, and growing demand for seamless experiences, users expect products to adapt to their individual needs in real time, rather than delivering generic, one-size-fits-all interactions. Genealogy presents unique personalization challenges: every family tree is different, motivations vary widely, and user journeys rarely follow predictable patterns. A user researching the Titanic in the newspaper archive has entirely different needs than someone methodically building out their maternal line. These journeys don’t generalize neatly, making basic segmentation and one-size-fits-all personalization ineffective. Anup Purewal, Chief Data Officer at DC Thomson, explained:

‍

 "There is no proper aggregation available for family history. Everyone's families are unique, and the motivations for each user are relatively unique as well." ‍

Anup Purewal, Chief Data Officer, DC Thomson

‍

FindMyPast had long relied on Snowplow's Customer Data Infrastructure (CDI) to power server-side analytics and behavioral data collection. But the product and engineering team needed a way to extend that data foundation with real-time behavioral intelligence to more deeply understand and respond to users in the moment. Attempts to build a solution in-house quickly stalled after much debate, leaving the project on the shelf for more than a year. 

When Snowplow introduced Signals, a real-time customer intelligence system for product and engineering teams, the FindMyPast team saw an opportunity to bridge their gap without taking on more bespoke engineering work. By combining their existing Snowplow event data foundation with Signals, the team could finally operationalize real-time user context across experiments, journeys, and future in-product interventions.

Challenges

FindMyPast was struggling to personalize product experiences in one of the most complex behavioral environments imaginable. Every user’s family tree is unique, motivations vary widely, and research paths rarely follow predictable patterns. At scale, this created a fundamental challenge for the product team: millions of users moving through entirely different journeys, often within the same session, without the real-time ability to deeply understand and respond to what they were trying to achieve.

FindMyPast relied heavily on batch-based segmentation, which wasn’t designed to capture the constantly shifting context of a genealogical search. Its existing approach relied on basic triggers such as counting how many family members a user had added to their tree or tracking simple events like record views. But these methods couldn't capture the nuanced, in-session context needed to understand what users actually wanted in real time.

Anup highlighted the gap: 

‍

"We could track when someone added a person to their tree and trigger something off that. But it's been much harder to get client-side events into a stream that we can actually act on in real time."

Anup Purewal, Chief Data Officer, DC Thomson

‍

This lack of real-time context created several interconnected problems:

1. Inconclusive experiments: A/B tests frequently produced flat results because cohorts contained users with vastly different motivations, such as first-time tree builders, deep-branch explorers, and casual newspaper browsers. Without the ability to sub-segment based on deeper intent, it was impossible to identify which personalized interventions actually moved the needle.
   ‍
2. Frustrated customers: Product changes intended for new users often alienated existing customers, and vice versa. Tom Getgood, Principal Engineer at DC Thomson, explained: "We'd run an experiment targeting new users, and then existing users would complain asking why we'd put something irrelevant in front of them."
   ‍
   
3. Avoidable user churn: FindMyPast could trigger interactions based on explicit actions, like adding someone to a family tree. However,  subtle signals of user struggle, such as repeated clicks, back-and-forth navigation, or aimless searching, remained invisible in real time. Without capturing and acting on these implicit behaviors in real time, the product couldn’t proactively intervene when users were confused or stuck, missing critical opportunities to guide them and prevent churn.
   ‍
4. Stalled engineering efforts: Internal debates over architecture—whether to use client-side or back-end events—consumed engineering cycles without producing results. As Tom put it:
   

‍

‍

"If we'd tried to build this ourselves, we'd probably still be debating architecture diagrams. It would have taken a lot longer."

Tom Getgood, Principal Engineer at FindMyPast

‍

What became evident was that FindMyPast needed a solution that could deliver real-time user intelligence without the lengthy build cycles and maintenance burden of a custom-built system.

Solution

To tackle its real-time personalization challenges, FindMyPast adopted Snowplow Signals, Snowplow’s real-time customer intelligence system. Signals provided the engineering team with the infrastructure they needed to compute and serve user attributes in real time, combining streaming behavioral data with historical warehouse data, and deliver them via API to downstream systems. This generated  a level of real-time personalization that was previously impossible.

The decision to adopt Signals rather than build in-house made perfect sense. Anup explained:

‍

 "We'd wanted to do contextual, real-time personalization for a while. Signals gave us the infrastructure to make it happen without building everything from scratch."

Anup Purewal, Chief Data Officer, DC Thomson

‍

Critically, Signals integrated with FindMyPast's existing infrastructure rather than replacing it. This gave the team control over their data and logic while eliminating the need to build real-time computation from scratch.

The engineering team integrated Signals into their existing stack using Kafka to produce user profiles consumable by any service, unifying workflows and architectures across engineering, product, and marketing teams. Tom described the architecture:

‍

‍

"Once we've built a behavioral profile from the aggregated attributes in Signals, we produce that to Kafka and any service can consume it. We push profiles to Iterable to power email journeys, and we've connected services for Trustpilot and Refiner, our customer feedback tool."
‍
Tom Getgood, Principal Engineer, DC Thomson

‍

Signals also addressed one of FindMyPast’s biggest pain points: noisy cohorts and inconclusive experiments. By identifying sub-segments of users with shared intent, such as first-time tree builders versus deep-branch explorers, Signals allowed the team to run experiments that actually reflected the impact of their interventions, rather than being diluted by irrelevant users.

The integration with Iterable proved particularly valuable for cross-team collaboration. FindMyPast feeds real-time user profiles from Signals directly into Iterable, enabling the CRM team to build sophisticated journey logic based on live behavioral attributes. As Tom explained: 

‍

"As soon as Signals produces a user profile, we send it straight to Iterable. From there, the CRM team can build journeys that show specific banners or tiles to certain users and not others.”

Tom Getgood, Principal Engineer, DC Thomson

‍

FindMyPast deployed two initial use cases: 

• Trustpilot reviews: Signals identified "power users" based on real-time engagement patterns, ensuring review requests went only to users genuinely enjoying the product, improving both the scores and quality of reviews.
• Search onboarding email personalization: FindMyPast launched an experiment to personalize the onboarding journey for users who signed up via search, adapting emails to their in-session behavior and experience level. Signals provides the user context to Iterable, which uses the data to only send emails to the right cohort of users. The next step is for Iterable to use Signals data to inform the journeys it is orchestrating on the FindMyPast site.

The speed of deployment exceeded expectations. Tom noted:
‍

 "Having a product like Signals solved a lot of internal debates. We weren't wasting time building custom infrastructure or arguing about architecture; we could focus on what we wanted to achieve."
Tom Getgood, Principal Engineer, DC Thomson

‍

Signals also enabled contextual, in-product interventions that had long been envisioned, such as detecting user confusion and surfacing guidance precisely when needed. Beyond immediate use cases, the team also sees Signals enabling more advanced personalization like onboarding for returning “reanimated” users.

The Snowplow team's responsiveness during early adoption proved valuable. Tom noted: 

‍

"We've never been told that something was a stupid question. They've been really open to feedback, and it's felt like a genuine partnership throughout."
‍
Tom Getgood, Principal Engineer, DC Thomson

‍

By combining real-time behavioral intelligence with a flexible integration approach, Signals gave FindMyPast the ability to finally understand users in the moment, reduce noise in experiments, and deliver the personalized experiences that its users expect.

‍

Results

FindMyPast’s adoption of Snowplow Signals has completely modernized the product’s personalization capabilities. It can build real-time user context into each user’s product experience.

‍
Faster Time-to-Value

FindMyPast dramatically reduced the time needed to ship personalized experiences. The search onboarding experiment—a complex, multi-touchpoint email journey powered by real-time user attributes—moved from concept to production in just six weeks. Tom noted: 

‍

"A month and a half to get this live is not bad at all. If we'd tried to build this ourselves, we'd probably still be debating architecture diagrams."
‍
Tom Getgood, Principal Engineer, DC Thomson

‍

Higher-Quality Customer Reviews

By identifying power users in real time, based on engagement with advanced features across multiple sessions, FindMyPast transformed its Trustpilot strategy. Rather than requesting reviews from all users, only users who are genuinely enjoying the product will receive a request. 

As a result, the average review score increased by 20% following the Signals implementation. They also noted a marked shift in review quality. Anup explained:

‍

"It's not just the score improving. People are now writing about specific features they enjoyed, not just complaining about price or leaving generic praise. That translates our product's value much better to prospective customers."

Anup Purewal, Chief Data Officer, DC Thomson

‍

Faster, More Relevant Communications

With Signals providing real-time user context for onboarding, relevant emails now reach users in 60 seconds, rather than the 4-hours it took before. Related recommendations emails, which suggest content based on recent activity, arrive almost immediately rather than waiting for batch processing cycles. This is extremely valuable for first-time genealogy enthusiasts diving into their complex family histories. 

This transformation represents the latest evolution in FindMyPast’s CRM maturity. Anup noted: 

‍

"Over the last couple of years, the CRM team has moved from sending everyone the same onboarding email to doing more sophisticated profiling—and we've seen click-through rates improve as a result. Now we're getting people coming back, taking free trials, converting to subscriptions, and engaging more with the site."

Anup Purewal, Chief Data Officer, DC Thomson

Reduced User Churn

Signals has unlocked churn mitigation strategies that were previously not possible to execute. One particularly powerful capability is detecting implicit user confusion in real time. Previously, FindMyPast could track major events like record views or tree additions. However, it struggled to capture the nuanced behavioral signals that indicate when someone is lost or frustrated. As Anup explained:

‍

"Previously, we could track the big main events like when someone registers a search, views a record, or adds someone to their tree. But we couldn't capture when users are clicking around in what seems like a really confused manner. Now we are able to detect that confusion in real time, and that's exactly when we will show them help or guidance."
Anup Purewal, Chief Data Officer, DC Thomson

‍

This represents a fundamental shift from reacting to explicit behaviors, like abandoned family trees or trial cancellations, to recognizing subtle, context-dependent patterns that signal frustration or unmet intent. 

These implicit behaviors emerge from multiple micro-interactions and are difficult to predefine in traditional rules-based systems. But with Signals computing complex behavioral attributes in real time, FindMyPast can now intervene at exactly the right moment to guide users in the most helpful direction. 

Bridging Data Science and Production

Beyond engineering, FindMyPast's data science team is now building predictive models based on data from Signals. This includes churn prediction and conversion propensity using Signals' real-time attributes. Using Signals, FindMyPast has bridged the gap between model development and production deployment, enabling its data scientists to train models on warehouse data with confidence they'll work identically on live streaming data.

‍

Looking Ahead

FindMyPast's new approach to real-time personalization is already showing significant impact and the roadmap ahead is ambitious.

The next priority is to deliver personalized on-site messaging, using Signals data to power dynamic banners and contextual prompts within the product itself. The team is already planning to extend the Signals integration with Iterable beyond email to on-site experiences. 

The product and engineering team is also focused on personalizing experiences for reactivated users, a significant cohort who return to the product after extended absences. With Signals, the FindMyPast team can personalize their experience based on where they left off and what's changed since. The team can also test ways to reduce this cohort going forward by keeping users actively engaged from the very start. 

Longer term, the company sees Signals as foundational to its AI strategy. Data science teams working on content discovery are exploring how real-time user context can surface the most relevant records and newspaper articles, not just what's generally popular, but what matters to that person at that moment. The goal is to combine the depth of FindMyPast's historical archives with the precision of real-time behavioral intelligence.

Looking onto the horizon, Anup summarized:

‍

 "We're getting to a point where we can understand what a user wants from the product and facilitate them getting it. That's something we've wanted to do for years—and now we have the infrastructure to make it happen."
Anup Purewal, Chief Data Officer, DC Thomson

‍

With Snowplow, FindMyPast has transformed its ability to deliver personalized in-session experiences at scale. What once required lengthy internal builds and architectural debates now takes weeks from initial concept to production. As the company expands Signals across new channels and use cases, the foundation is set for a truly adaptive, AI-powered genealogy platform.