0000 M · Dive log · archived descent

emetriq · Deutsche Telekom Group · 2019–2021

Machine-learning products at telco scale.

Two years inside the Deutsche Telekom group's data company, shipping ML-driven product MVPs, rebuilding how product gets designed, and ending up on the management board.

0300 M · Station log

Six stations,
two years down

  1. – 0400 M

    Station 01 · Machine learning

    AI-driven content buying

    Concept, design and MVP for new data- and machine-learning-based services for buying video-on-demand content for a major streaming platform. From raw viewing-market data to a decision tool the buying team could actually use.

  2. – 0700 M

    Station 02 · Measurement

    Second-precise TV-campaign measurement

    An MVP that tracked TV campaigns to the second across broadcasters and measured their impact on online conversion and other targets, connecting old-world TV to digital outcomes.

  3. – 1000 M

    Station 03 · Process

    Product process, rebuilt near the code

    A design and development process for all new digital products that left static artifacts behind: design tokens, a design system, prototyping with real code modules. Design that ships, not handoffs that drift.

  4. – 1300 M

    Station 04 · Feedback loop

    A continuous user-feedback engine

    User interviews, tracking observations, actionable KPIs, prototype testing: learn, improve, test, repeat. Installed as a standing cycle, not a one-off study.

  5. – 1600 M

    Station 05 · Go-to-market

    Internal service to SaaS

    With management and finance, built the go-to-market strategy to turn an internal service into a SaaS product. Then ran lean build-measure-learn cycles and adjusted the business model accordingly.

  6. – 1900 M

    Station 06 · Governance

    The boardroom

    Member of the management board for organizational development: new pay and new work, protected spaces for pioneering projects, the post-cookie era, and positioning the company beyond its original mandate.

2300 M

Contact

If you're serious about what failure prediction can do for your fleet, I'm easy to talk to.