Can You Build Your Own Journey Analytics Solution with AI?

A Curious PM's AI Experiment

Hi 👋🏻, I'm Trevor. If you don't know me, I'm a naturally curious product manager, and lately I (like everyone else) have been fascinated by what's really possible with AI development tools. The best way to really understand something is to build with it, right? So I figured — what better way to test these tools than by building something in a domain I have some experience in?

While my day job is very much focused on building the next big thing in HR analytics with my team at UKG, I thought it'd be cool in my free time to see what's possible in a domain I got to know during my time at Adobe: customer journey analytics.

Years ago, I had the privilege of being the lead product manager to create a brand new analytics product at Adobe called Customer Journey Analytics. It was the culmination of work from hundreds of engineers, built on top of Adobe's Experience Platform. We borrowed and improved upon many of the ideas from Adobe Analytics over 20+ years, and grew the CJA business to serve many of the Fortune 500 and beyond. It was certainly one of the highlights of my career thus far.

Since the launch of CJA however, data and analytics technologies have changed quite a bit — most particularly around AI (which wasn't really the thing it is now). AI has changed not only how problems can be solved, but it's also proving that software development can be done more cheaply and quickly if you know what you're doing. I've observed that what was once only possible through the work of hundreds of engineers over months can in some cases now be done much more quickly by far fewer.

And while a lot of the AI talk out there in the B2B software space is often more bluster than substance, it still begs the question: Could you create your own journey analytics solution with AI coding tools? I figured if anyone were to take a stab at it, why not me? And even if it doesn't work out, there's a ton to learn from the process.

I also thought it'd be helpful to share what I learn along the way for anyone else out there wrestling with their own analytics tooling — or really, anyone curious about what AI-assisted development can actually accomplish in the data and analytics space.

So... where to begin?

What Is a Journey Analytics Tool?

In order to create a journey analytics tool, we have to understand what makes them unique and special within the analytics universe. If I had to distill down what makes such a tool special and distinct from generic BI infrastructure, I would say that it helps users accomplish a chain of capabilities no other type of analytical tool does well all at once:

Ingesting, organizing, and coalescing large volumes of event data from many sources.
Stitching the various forms of user identities together into a cohesive "person", and applying those identities retroactively to the data in an ongoing way.
Simplifying what would otherwise be otherworldly complex SQL queries into something more manageable (e.g. sessionizing data, persistence of variables at the user/session level, segmentation and sequence analysis, etc.)
Providing a journey-specific semantic layer on top of your data (e.g. creating actual customer journey dimensions and metrics from your data)
Performing queries at speeds that are prohibitive for a general-purpose data warehouse.
Providing a deep analysis user experience that goes beyond the basic BI dashboard.
AI agents armed with the mountain of context required to do journey analysis that generic natural-language-to-SQL solutions can't do on their own.
Doing all of these tasks while maintaining customer data security and honoring data privacy laws.

This is a tall order, even for companies with hundreds of engineers - but I actually believe building such a thing with AI (and an intelligent choice of existing technologies) is possible. But, I don't want to stop there; simply matching what is out there already isn't enough. There are many problems that have yet to be solved in this space that I also want to attempt to tackle.

Where Journey Analytics Tools Still Need Work

To my satisfaction, there are a number of areas that have not been adequately solved in the journey analytics space:

Most companies don't want to make independent copies of their data into another vendor - companies want to keep their data centralized in a single place to maintain a single source of truth where privacy and data corrections can be handled centrally.
Journey analytics tools are notoriously difficult to setup, mostly because they require you to stand up entirely new data collection (i.e. new "tags" on your pages). For companies that already have their data flowing into their warehouse, deploying yet more tags turns into a superfluous, tedious task.
There is not a competitive journey analytics tool out there (to my knowledge as of this writing) that allows companies to only pay for what they actually use (as opposed to having to sign expensive multi-year contracts) - making sophisticated analysis out of reach for smaller companies on a tighter budget.
Based on my experience, AI capabilities are typically a bolt-on afterthought used primarily for checking boxes in customer RFPs, not a foundational aspect of the entire user experience from setup to analysis.

So, if I'm going to actually attempt this, I'm going to set the bar high and try to solve these problems too. And while you may (rightfully) laugh at me for attempting to do this, you might also be surprised just how far I get!

So What's Next?

In upcoming posts, I'll be diving into how I actually decided to go about creating this thing, and I'm inviting you to follow along! I'll be sharing my thoughts on:

Choosing the right architecture - Why I picked Airbyte and Clickhouse for this use case, and how to connect it all together with the right databases and APIs.
Syncing data from your warehouse - Using Airbyte to sync data between a warehouse and ClickHouse, and how to use their APIs to make that work seamlessly.
Journey queries in ClickHouse - How to write sessionization, pathing, dimension/metric manipulation, and sequence queries that don't take forever to run.
Building an identity stitching solution - How to use the replacing merge tree capabilities of ClickHouse to accomplish ongoing, retroactive identity stithcing.
Creating a semantic layer for dimensions and metrics - How to make your data more accessible without having to remember a bunch of SQL definitions.
Using AI to simplify and streamline the experience - How to use Claude AI to do all the tedius configuration and setup for you, so you can get to working with the data.
Building a user experience designed for analysts - Why a user experience designed for analysis is inherently different than just building dashboards.

And if you'd like to see an early sneak peak demo of what I've cobbled together, you can watch it here:

Early prototype of the Trevorlytics query interface

Stay tuned — this should be fun.

And if any of this resonates with you — whether you're dealing with your own journey analytics challenges, curious about the technical approach, or just want to play with my prototype yourself — I'd genuinely love to hear from you. I learn best by talking through problems with others, and honestly, building something like this in a vacuum isn't nearly as interesting as building it alongside people who actually care about solving these problems.

Reach out on LinkedIn.

Trevor Paulsen is a data product professional at UKG and a former product leader of Adobe's Customer Journey Analytics. All views expressed are his own.