Alex Dean and the Trouble with a Finished Answer

Alex Dean wanted to get past the answer. Working with his future co-founder, Yali Sassoon, he had encountered an awkward feature of digital analytics: a company could watch the reports about its customers while struggling to reach the individual events underneath. There was information in the building. Getting it out required an expedition. For someone interested in how organizations change, that was a rather expensive arrangement.

The frustration became Snowplow in 2012. Its founding proposition put the underlying behavioral record within reach of the business that generated it. Companies could work with what happened on their websites and applications, then decide which questions deserved asking. A report is a useful convenience. It can also be a remarkably polite way of closing a door.

Dean, now Snowplow's CEO and co-founder, has spent the intervening years working on that door. His public career runs through consulting, advertising technology, cloud infrastructure and the mechanics of events. The continuity is unusually clear: before you can reason usefully about a customer's behavior, you need a trustworthy account of what the customer actually did.

An education in awkward questions

Dean attended the University of Cambridge from 1999 to 2002. Before Snowplow, his work included technology roles at OpenX and Deloitte Consulting's business intelligence department, alongside strategy roles at Fathom Partners and Keplar LLP. It is a career that crosses two conversational worlds: the people deciding what an organization should do, and the people discovering what its systems will permit.

That background helps make sense of the subjects he chooses to explain. His interests include functional programming, cloud architectures and organizational change. The technical machinery matters to him, but so does the institution attempting to use it. A company can buy software more easily than it can establish a common understanding of what its data means.

In a 2018 podcast conversation, Dean discussed Snowplow through collection, validation, architecture and its managed-service business. The scope is revealing. He is comfortable following the event from the website through the pipeline and into the commercial arrangement that keeps the pipeline running. His work asks the executive and the engineer to remain in the same conversation.

Please keep the evidence

Snowplow's early experiment let brands run an analytics engine in their own Amazon Web Services account and reach generated events in an S3 bucket. The attraction was the freedom to build on those records. Dean and Sassoon were leaving room for questions their software could not anticipate.

Think of a publisher that wants to understand a reading journey, or a retailer examining the steps before a purchase. These are illustrative situations, but they explain the stakes. A tidy total gives an answer to a defined question. Individual events, with useful context, let a team revisit the question, combine records and test another explanation.

The distinction places responsibility on the business, too. Someone must decide which behavior matters, how to describe it and what quality to demand. Owning the evidence creates work as well as freedom. Snowplow's proposition has always asked customers to take their data seriously enough to make those decisions.

The grammar lesson inside the software

By March 2014, Dean was writing about event context with the enthusiasm of somebody who had discovered a useful extra dimension. He used grammar to explain the problem: subjects, verbs and objects tell you who did what; context adds the circumstances. Time, place and manner belong in the record.

The numbers in that early explanation are pleasingly concrete. Of the 98 fields in Snowplow's canonical event model, he counted 57 devoted solely to context. More than half the model described the circumstances around the action. The apparently secondary information occupied most of the room.

A page can be the object of one event and the setting for another. That small observation carries a larger lesson about Dean's work. Meaning depends on relationships. Strip away the surroundings and even a perfectly recorded action may tell you less than you think.

The grammar analogy also makes his technical interests accessible. Anyone who has misunderstood a sentence because a crucial phrase was missing already has a foothold in the problem. Dean takes an infrastructure question and gives it a familiar shape.

Even the clock needs cross-examination

His September 2015 essay on time tackled another innocent-looking question: when did the event happen? Snowplow recorded multiple clues because the answer could be surprisingly difficult to establish accurately. The clock on a device and the time a collector receives a record describe different moments.

That is the sort of detail that rarely survives into a product slogan. It matters if a team is reconstructing the order of actions. A system that casually rearranges events can produce a convincing account of something that never happened in that sequence.

A year earlier, Dean and Frederick Blundun had drafted SchemaVer, a way to version data schemas according to their compatibility with historical data. Published in May 2014, the proposal distinguished changes that preserve compatibility from those that may disrupt some or all earlier records.

There is an understated persistence in these subjects. Time and definitions sound settled until someone has to implement them. Dean's published work repeatedly stops at that uncomfortable point and asks for a more careful answer. The glamorous part of analytics can wait a moment while the plumbing is examined.

“The time at which an event took place is a crucial fact for every event”Alex Dean, September 2015

A larger company, with the same difficult plumbing

On June 30, 2022, Dean announced a $40 million Series B led by NEA, with Atlantic Bridge and MMC participating. The round brought disclosed funding to $55 million at that point. The announcement described more than 200 customers generating a quarter of a trillion events a month.

Those figures marked a change in scale. Snowplow had announced partnerships with Databricks and Snowflake and opened a second headquarters in Boston. The funding was intended to support further development across analytics, customer data infrastructure and machine learning.

NEA's account of meeting the founders offers another perspective on the milestone. The investors had found businesses in their own portfolio already paying for Snowplow and running it for years. The product was doing its work underneath other companies' operations before the investment conversation began.

There is a particular satisfaction in that kind of discovery. Infrastructure earns its place by continuing to work while attention goes elsewhere. For Dean, the commercial challenge was to build a company around a capability whose success often makes it less visible.

The watercooler makes a guest appearance

Snowplow has long operated as a geographically distributed company. In July 2022, Dean described bringing more than 130 colleagues together in Dublin for Snowfest, a week of learning and connection. He celebrated the chance for a distributed organization to share a physical location.

His chosen detail was the watercooler: for that week, the conversations could happen around a real one. It is an appealingly ordinary image amid the language of cloud infrastructure. The people maintaining digital connections still value the chance to stand beside one another.

Dean credited the geographic spread of the team with giving Snowplow strength and diversity. The gathering made a different benefit possible: time together outside the daily pattern of remote work. The organization behind the event stream needed its own encounters, its own shared context.

The price of keeping the glass clear

The business story includes a consequential change. In January 2024, Sassoon announced that new versions of Snowplow's core pipeline technology would move to the Snowplow Limited Use License Agreement. The company described an evolution from commercial open source toward enterprise software with a source-available posture.

Code visibility remained part of the proposition. Usage conditions changed, including restrictions concerning highly available production deployments and competitive products. The announcement said tracking SDKs would remain Apache 2.0 licensed. It also acknowledged disruption for open-source users upgrading core components.

This decision belongs in any account of Dean's company. A philosophy of customer data ownership does not settle every question about software economics. Maintaining and developing the infrastructure requires a durable business. Snowplow's leadership chose a different licensing balance and explained it publicly. The distinction between accessible code and unrestricted use became something customers had to examine carefully.

An agent walks into the website

By May 2026, Dean's attention was turning toward the customer-facing AI agent. In an essay defining a customer context layer, he argued that useful agents require fresh behavioral information, resolved identity and context structured well enough to support action. A record arriving after the moment has passed cannot help with that moment.

His six principles covered quality and governance at creation, broad behavioral coverage, timely and compact context, ownership and composability, feedback from outcomes, and practical deployment. The ambition reaches beyond producing another report. Software may now use the information to decide what to do while a customer is still there.

Consider an assistant that needs to understand what a visitor has already tried. This is an illustration, not a reported customer case. Without that history, the assistant risks recommending the same unhelpful step again. The sophistication of its language does little to repair the missing record.

Dean's argument connects directly to his earlier attention to grammar and timestamps. Acting quickly raises the cost of a bad description. The old infrastructure questions follow the agent onto the stage.

The newsletter has other plans

Dean returned to his Source of Truth newsletter in May 2026 after nearly two years away. He planned a renewed focus on real-time decisioning, AI agents and the changing customer journey. Even the newsletter's name was up for reconsideration. He wanted to publish more before asking readers what should replace it.

That July, he approached customer data platforms through a less solemn frame: an East Coast versus West Coast rap rivalry. His argument concerned how vendors' early customers shape their products. The metaphors were doing explanatory work, with an occasional raised eyebrow.

He also reached for the technology tree in a strategy game: early engineering choices constrain what a company can build later. It is a useful glimpse of his public voice. An argument about data architecture can accommodate a joke without losing its technical point.

Fourteen years after Snowplow began, Dean is still examining the consequences of those early choices. The dashboards have company now. AI agents want to act, customers bring different expectations, and software must keep track of both. Somewhere underneath the conversation sits a modest requirement he has been working on since the beginning: keep a usable record of what happened.