Analytics Engineering: Internal Risk vs. External Rigor
Internal analytics: speed and tolerance for risk
Internally, analytics teams can usually move fast. If a model breaks or a dashboard shows an odd number, there is often a chance to explain, fix, or correct it after the fact. You can talk directly to users, understand context, and recover quickly. This environment encourages experimentation and rapid iteration, even if it comes with occasional mistakes.
Jack describes this as a space where teams can accept more risk. Tests might be lighter, releases more frequent, and assumptions more flexible. The priority is insight and learning rather than perfection.
Product analytics: rigor and responsibility
That mindset changes completely when analytics become part of the product. At Fresha, analytics are something customers pay for and rely on to make real business decisions such as staffing, pricing, and opening hours. When data is wrong or late, there is no analyst in the room to explain why.
This pushes analytics engineering closer to traditional software engineering. Jack talks about heavier testing, stricter deployment processes, coordinated releases with application teams, and even timing changes around customer usage patterns. Pressing “merge” suddenly carries real weight.
One key shift is scale. Internal queries might run occasionally, but product analytics must handle large volumes of concurrent requests. Fresha’s team had to rethink architecture, ingestion patterns, and cost control, moving toward near real time data freshness and CDC based pipelines to meet user expectations across time zones.
The semantic layer as the missing link
A major theme in the episode is the semantic layer. Jack argues that many analytics problems are not about raw data but about meaning. Different teams define concepts like “user” or “active customer” differently for valid reasons. Trouble starts when those definitions are opaque or inconsistent.
Rather than forcing a single definition, Jack advocates for making definitions explicit, codified, and queryable. The semantic layer becomes the place where business meaning lives, enabling consistency while still supporting multiple perspectives. This is especially important as natural language and AI driven interfaces become more common.
What this means for analytics engineers
As analytics engineering evolves, the role is shifting away from just building tables and dashboards. It is becoming more about structure, contracts, and clarity. Teams that take on product facing analytics gain a deeper appreciation of software discipline, while product teams gain access to richer, joined-up data.
Jack’s advice for teams starting this journey is simple: start small, move slowly, and build trust. Product analytics expose edge cases you never expected, and most of the work happens after the first model is built.
Dig Deeper 🎧
Listen: Analytics Engineering: Internal Risk vs. External Rigor
📺 Watch: https://youtu.be/XbhtO5HeFt0
👤 Connect with Jack Doherty on LinkedIn
🌐 More episodes: Matatika.com/podcasts