— Field notes from the operating floor

Case stories, not slides.
Essays for the people who ship.

Anonymised but unvarnished accounts of what we actually built for clients — and the patterns we keep seeing when operators try to make strategy survive contact with delivery.

— Case stories
6 published
— Field essays
6 published
— Industries
8 covered
— Format
5–12 min read
— 01 / Story Index

Twelve pieces, six industries, one operating thesis.

Internationalisation
Case story·10 min

Compressing country roll-outs from quarters to weeks.

A configuration-driven internationalisation blueprint that turned market launches into a data exercise.

— Commerce · Aftermarket
AI-Native
Case story·11 min

Building an AI-native product organisation that actually ships.

Rewiring discovery, delivery, and analytics into a single AI-native execution system.

— Commerce · Strategy
IoT · Telco
Case story·9 min

From innovation lab to enterprise revenue: the IoT blueprint.

How an in-house lab inside a global telco became a productised commercial engine.

— Telco · IoT
Automotive
Case story·8 min

Connected automotive at OEM scale: one platform, many fleets.

A streaming and personalisation platform standardised across four global OEM teams.

— Automotive · OEM
Logistics
Case story·8 min

Digitising last-mile: planner, driver, returns.

Operations as a product. Route planning, driver apps, returns — on one digital backbone.

— Logistics · Ops
Replatform
Case story·9 min

From legacy webshop to in-house commerce platform.

A replatform that didn’t stop the business — and an internationalisation blueprint that came for free.

— Nutrition · Commerce
Essay
Essay·7 min

Configuration over code: the test for whether your platform is finished.

If a new country, payment method, or campaign needs an engineer, the system isn’t done.

— Architecture
Essay
Essay·6 min

Career frameworks that move people — not just promote them.

Roles, skills, and levels designed to make growth legible. Three-times-a-year micro-reviews.

— People & Org
Essay
Essay·5 min

Radical candor, the operating-floor version.

Less a virtue, more a meeting cadence. What candour looks like once you remove the LinkedIn aesthetic.

— Leadership
Essay
Essay·7 min

The accountability scoreboard: when every release scores itself.

A +1 / −1 ledger per release, public post-mortems, and what happens to a team’s quality when the score is visible to everyone.

— Engineering culture
Essay
Essay·6 min

Verification over assumption: engineering a zero-hallucination culture.

Most production incidents trace back to one pattern — assuming instead of checking. The discipline of “I don’t know, let me verify.”

— AI-Native · Quality
Essay
Essay·7 min

Weighted evidence: a decision protocol that kills the HiPPO.

Consult every discipline, bring three sources, weight each by relevance, clear a threshold. Decisions by evidence — not by the loudest voice.

— Leadership · Decisions

Compressing country roll-outs from quarters to weeks.

A configuration-driven internationalisation blueprint that turned market launches from an engineering project into a data exercise — and made multi-market growth a repeatable cadence rather than a heroic launch.

— At a glance
— SectorB2B + B2C Commerce
— ScopeMulti-market launch model
— ReachDouble-digit market footprint
— Duration3 quarters
— SurfaceStrategy + Engineering

— SituationA growth thesis that the platform couldn’t keep up with.

The business was growing on the back of internationalisation. Each new country, though, took multiple quarters — localisation, payments, taxes, returns, carrier integrations, every one of them an engineering project. The roadmap was clogged with country-specific exceptions and the org was burning out launching markets instead of compounding on them.

The diagnosis surfaced a structural problem: every assumption about a single home market was hard-coded somewhere. Internationalisation wasn’t failing because of capacity. It was failing because the platform had nationality baked into it.

— ApproachExternalise the assumptions, ship a blueprint.

We started by cataloguing every national assumption in the platform — tax, payments, language, currency, carriers, returns, regulation, even copy. Each of those became a configuration surface, with feature flags and per-country defaults. A new launch became a checklist run, not a sprint.

  • Configuration surface: tax, payment, locale, currency, returns
  • Per-country feature-flag matrix tied to release management
  • Launch operating model with explicit go/no-go gates
  • Embedded carrier and partner integrations as data, not code

— ExecutionOne launch, one cadence, one team.

We ran the first three markets ourselves under the new model, with the local commercial teams as in-room partners. By the third launch, the engineering involvement was minimal — the work had moved into product configuration and operations. The fourth and fifth markets were run by the in-house team without us.

— OutcomeFrom quarters to weeks — without heroics.

Country roll-outs collapsed from a quarter-scale programme to a few weeks of focused operational work. The platform stopped fighting itself. Engineering capacity, formerly spent on per-country exceptions, was redeployed to compounding features that benefited every market simultaneously.

More importantly, the operating model survived without us. Markets four through twelve were launched by the in-house team on the same blueprint, with quarterly check-ins rather than embedded oversight.

— Signature outcome
Launch cadence compounding without engineering bottleneck.
— Cycle compression
Quarters → weeks
— Footprint
Double-digit markets, single platform
— Hand-off
Owner-operated within one cycle

Building an AI-native product organisation that actually ships.

A product organisation rewired around AI as load-bearing infrastructure — not as a feature surface — with discovery, delivery, and analytics integrated into one execution system aligned to the company strategy.

— At a glance
— SectorPremium e-commerce
— SurfaceOperating model + AI
— TeamsMulti-team, single org
— CadenceQuarterly outcomes
— SponsorC-suite

— SituationAn AI initiative trapped in workshop mode.

The organisation had run multiple AI pilots. None of them had survived contact with the operating model. Discovery happened in one place, feasibility in another, analytics somewhere downstream. AI lived in slides and demos, never in the release branch.

The leadership knew the answer wasn’t another tool. It was a different operating model — one where AI sat inside the workflow, not next to it.

— ApproachSix growth pillars, one execution system.

We redesigned the roadmap planning to integrate User Research, Technical Feasibility, Product Discovery, Data Analytics and Delivery into a single end-to-end execution system — explicitly aligned to the company’s strategic pillars.

  • Six growth pillars with clear owners and measurable outcomes
  • AI tooling embedded into Figma, analytics, and release pipelines
  • Customer Effort Score and automated feedback wired to roadmap
  • Competitor monitoring as a continuous research function
  • Career frameworks and 3×/year micro-reviews to sustain the model

— ExecutionBehaviour, not posters.

The hard part wasn’t the architecture. It was making sure the values showed up in how teams actually planned, built, and shipped — under deadline. We embedded inside the leadership team, ran coaching programmes and workshops, and protected the cadence until it became default.

— OutcomeCulture as behaviour, growth as a side effect.

The product culture that emerged contributed directly to double-digit YoY business growth and consistent KPI overachievement across all teams. AI tooling stopped being a quarterly demo and became how design pipelines, analytics, and feedback loops worked.

Retention and team progression both improved measurably, with the career framework and micro-review cadence absorbing what used to be hallway promotion politics.

— Signature outcome
AI as load-bearing infrastructure, embedded in delivery.
— Growth contribution
Double-digit YoY uplift
— Operating model
One end-to-end system
— Sustainability
Owned by in-house team

From innovation lab to enterprise revenue: the IoT blueprint.

An in-house innovation lab inside a global telecommunications group, turned into a productised commercial engine with cross-functional self-managed teams across multiple European hubs.

— At a glance
— SectorTelco · Enterprise IoT
— SponsorExCo · Global enterprise
— TeamsCross-functional, multi-hub
— HubsLondon · Düsseldorf · Berlin
— SurfaceIdeation → release

— SituationAn innovation lab without a commercial engine.

The lab had been generating credible IoT concepts but struggling to convert them into recurring enterprise revenue. Roadmaps slipped, ExCo sponsorship cooled, partner conversations stalled. Innovation was happening; productisation wasn’t.

— ApproachA KPI-driven operating model.

We formulated business cases directly with ExCo members, established a KPI-driven operating model across multiple self-managed cross-functional teams, and aligned partnerships with the broader group’s infrastructure strategy.

  • Business cases co-authored with ExCo sponsors
  • Self-managed cross-functional teams with measurable outcomes
  • Discovery, analytics, and BI as continuous functions
  • Partnership model aligned to global infrastructure

— ExecutionIdeation to release, every quarter.

Two enterprise IoT products went from ideation to release management within the engagement, with release governance designed to work inside a global carrier’s constraints rather than around them.

— OutcomeRecurring revenue, in-house ownership.

The lab generated new revenue lines for both the German entity and the global enterprise IoT business. More importantly, it shifted from a project-funded experiment into a continuous productisation engine — with cadences, scorecards and accountability that survived leadership changes.

— Signature outcome
Innovation lab productised as a recurring revenue engine.
— Products shipped
Two enterprise IoT lines
— Reach
DE entity + Global CIoT
— Model
KPI-driven, self-managed

Connected automotive at OEM scale: one platform, many fleets.

A content streaming and personalised marketing platform standardised across multiple global OEM teams — releases, requirements and go-to-market run as a single coordinated motion across regions.

— At a glance
— SectorAutomotive · OEM
— ScopeIn-car + mobile
— ReachNorth America + Europe
— Teams4 global
— SurfaceB2C + B2B + OEM

— SituationA platform pulled in three directions.

OEM partners wanted a stable, frozen surface. End-consumers wanted continuous improvement. Safety and homologation wanted both. Four global delivery teams were ostensibly building one platform but operating off four different sets of requirements and four different release calendars.

— ApproachStandardise the motion, not the feature set.

We revamped the product development workflow to standardise requirements gathering, technical reviews, release management and go-to-market execution across the four global teams — without imposing a single feature set on every OEM partner.

  • Unified requirements and technical review process
  • Coordinated release management with regional flexibility
  • OEM partner cadence with explicit alignment rituals
  • Continuous research into driver safety and product usability

— ExecutionThe OEM partnership as a product.

Partner relationships moved from ad-hoc engagement to a formal cadence: shared roadmaps, joint go/no-go reviews, instrumented release artefacts. Driver safety and usability research drove a meaningful slice of the roadmap, raising satisfaction and reducing post-release defects.

— OutcomeFaster releases, fewer defects, higher loyalty.

Release cycle times shortened materially, post-release defects dropped, and customer satisfaction with the in-car and companion-app experience rose — with measurable improvements in driver-safety perception. The platform reached global scale on a single coherent stack.

— Signature outcome
One platform across global OEMs, on a coordinated motion.
— Release cadence
Materially faster cycles
— Quality
Fewer post-release defects
— Scale
Global, multi-OEM fleet

Digitising last-mile: operations as a product.

A logistics business reorganised around digital products — route planner, driver application, returns portal, internal operations dashboards — with machine learning embedded where it materially improved productivity.

— At a glance
— SectorLogistics · Last-mile
— ScopeEnd-to-end transformation
— ReachNationwide expansion
— SurfaceB2C + B2B
— OutcomeNew investment round secured

— SituationOperations as a cost centre.

Last-mile operations were running on a mixture of spreadsheets, point tools and tribal knowledge. Customer experience varied widely by route; new market expansion required hand-built process replication; investors were sceptical about scaling.

— ApproachA product per workflow.

We defined a new digital product and operations strategy that turned each major workflow into a first-class product — with its own roadmap, owner, metrics and release cadence.

  • In-house e-commerce platform with subscription support
  • Route planner with ML-assisted optimisation
  • Driver application with operational dashboards
  • Internal planning and operations tools

— ExecutionFrom plan to product.

Each workflow was rebuilt as a product with explicit metrics: time-on-route, complaint rate, capacity utilisation. ML was introduced where it earned its place — planning, dispatch — not where it merely sounded innovative.

— OutcomeNew funding, lower opex, national reach.

The digital backbone supported a successful new investment round, a continuous reduction in operational cost, and a national expansion that the previous setup would not have supported. B2B and B2C product lines began to compound on a single platform.

— Signature outcome
Operations as a digital product, not a cost centre.
— Capacity
Material efficiency gains
— Funding
New investment round closed
— Reach
National expansion

From legacy webshop to in-house commerce platform.

A D2C nutrition business migrated from a Magento-era webshop to a self-developed, in-house commerce stack — with an internationalisation blueprint that came for free.

— At a glance
— SectorD2C nutrition
— ScopeReplatform + marketing stack
— MarketsDE · AT · CH · SE
— PositionCategory leadership · EU
— SurfaceD2C · subscription

— SituationA platform that couldn’t carry the strategy.

The business had outgrown its Magento-era webshop. Customer acquisition, supply chain digitalisation and analytics were all bottlenecked by platform constraints. Internationalisation was on the strategy — but the stack couldn’t carry it.

— ApproachReplatform with internationalisation built-in.

We led the transition to a self-developed in-house commerce stack — designed from day one to support international launches, segmentation, and end-to-end automation of payments and logistics state machines.

  • In-house commerce platform with international hooks
  • Complete digital marketing setup — GA, GTM, SEM, SEO, CRM
  • Customer segmentation and lifecycle tracking
  • Automated payment and logistics state machines
  • Real-time dashboards for fulfilment operations

— ExecutionMigrate without stopping the business.

The migration ran in parallel-run mode, with the new platform absorbing markets and product lines one cohort at a time. The marketing stack went live early and started informing the catalogue and merchandising strategy before the migration was complete.

— OutcomeCategory leadership, multi-market.

The strategic plan enabled the business to establish category leadership in the nutrition space in Europe — with a blueprint to launch into adjacent markets (Austria, Switzerland, Sweden) on the new stack.

— Signature outcome
Replatform as the on-ramp to international growth.
— Position
EU category leadership
— Markets
DE · AT · CH · SE
— Migration
Zero business interruption
— Essays

Field essays for operators.

— Essay 01Configuration
over Code

~7 min · Architecture
Filed under: Platform

Configuration over code: the test for whether your platform is finished.

If launching a new country, payment method, campaign or fulfilment partner needs an engineer, your platform isn’t finished. It might be working — it’s not done. The test is whether the next variant can be expressed as data instead of code.

Most internationalisation programmes fail this test in the first quarter. Most payments roadmaps fail it permanently. Most campaign platforms quietly fail it forever, and the cost shows up as opportunity rather than incident.

The platform you can configure is the platform you don’t have to staff for.

The discipline isn’t to write less code. It’s to be extremely deliberate about which axes of variation your platform supports natively — tax, locale, currency, carrier, payment method, regulatory regime, campaign rule, segment — and then refuse to ship features that hard-code along any of them.

The fastest expansion stories we’ve seen were not the ones with the biggest engineering team. They were the ones where the platform’s axes of variation had been correctly externalised before scale — and where operations, not engineering, owned the launches.

— Essay 02Career
Frameworks

~6 min · People & Org
Filed under: Leadership

Career frameworks that move people — not just promote them.

A career framework is supposed to make growth legible. Most do the opposite. They become rigid level rubrics that calcify into hallway politics, with promotion decisions made on tenure and a yearly conversation no one trusts.

The version that actually works has three properties. The roles are described in terms of the behaviour you’d see in the workplace — not abstract competencies. The levels are calibrated against real artefacts — not adjectives. And the review rhythm is fast enough to course-correct before resentment compounds.

A framework reviewed once a year is a framework no one believes.

Three-times-a-year micro-reviews replace the annual conversation with something that genuinely operates. Calibration happens with peers in the room. Promotion is a function of demonstrated artefacts, not advocacy. Retention improves because people can see whether they’re moving — and managers can do something about it before someone leaves.

The point of a framework isn’t to label. It’s to move.

— Essay 03Radical
Candor

~5 min · Leadership
Filed under: Culture

Radical candor, the operating-floor version.

Radical candor isn’t a virtue. It’s a meeting cadence. The performative version — the one that survives on LinkedIn — is mostly people saying the unkind thing first and then complimenting themselves for honesty.

The version that works inside teams looks duller. It’s a manager who debriefs each week with a structured set of questions, who names the same uncomfortable observation three meetings running until something moves, and who’s willing to be wrong out loud.

Candour is a habit measured in repetitions, not in slogans.

It’s also a system. You can’t expect candid feedback up the line if down the line has never seen it modelled. You can’t expect cross-team candour if the rituals to surface it don’t exist. Servant leadership and radical candor sound like values; in practice they’re a set of meetings, with the awkward parts left in.

— Essay 04The Accountability
Scoreboard

~7 min · Engineering culture
Filed under: Operating

The accountability scoreboard: when every release scores itself.

Inside our engineering practice, every release carries a number. A clean release that passes all quality gates and stays stable in production earns +1. A release that causes an incident, breaks the pipeline, or needs a hotfix earns −1. The score is visible to the whole team, alongside a written post-mortem for every negative.

It sounds harsh. In practice it does the opposite of what people fear. Because the score is attached to the release and the post-mortem is about the system, not the person, it removes the politics from quality. Nobody argues about whether something was “really” a bug. The pipeline failed or it didn’t. The number moved or it didn’t.

A visible score turns quality from an opinion into a shared fact.

The real value is in the post-mortems. Each negative release gets a structured write-up: what we expected, what actually happened, the first-principles root cause, and the specific prevention measure added to the pipeline. Over time the prevention measures compound — pre-commit hooks, contract tests, artifact validation — until an entire class of failure simply can’t recur.

The pattern we see again and again: teams don’t lack talent, they lack a feedback loop tight enough to learn from. A scoreboard, used without blame, is the cheapest such loop we’ve found.

— Essay 05Verification
over Assumption

~6 min · AI-Native · Quality
Filed under: Engineering

Verification over assumption: engineering a zero-hallucination culture.

When we run post-mortems across engagements, one root cause shows up more than any other — and it isn’t a missing skill. It’s assumption without verification. Someone assumed an endpoint existed. Someone assumed a tool generated a particular file structure. Someone assumed the framework behaved the way it used to. The code was fine; the assumption wasn’t.

This problem gets sharper, not softer, in AI-native teams. The same instinct that makes a model confidently hallucinate makes an engineer confidently guess. The discipline is identical in both cases: when you don’t know, say so, and then go and check.

“I don’t know — let me verify” is a senior sentence, not a junior one.

We make it a stated rule: never provide information without verification; never implement on an assumption; always read the actual file, test the actual artifact, consult the actual documentation. If something is unknown, the honest move is to say “I don’t know, I need to find the answer” — and immediately start looking.

It feels slower. It is dramatically faster, because the alternative is a ten-day-old broken pipeline nobody noticed, or an eval suite quietly validating the wrong thing. Verification is not bureaucracy. It is the cheapest insurance a team can buy — and the cultural foundation for trusting AI in the loop at all.

— Essay 06Weighted
Evidence

~7 min · Leadership · Decisions
Filed under: Operating

Weighted evidence: a decision protocol that kills the HiPPO.

Most product decisions are still made by the HiPPO — the Highest-Paid Person’s Opinion. It’s fast, it feels decisive, and it’s wrong often enough to be expensive. The fix isn’t more meetings. It’s a protocol that forces evidence to the surface before a decision is locked.

The version we use is deliberately mechanical. Before any significant decision, the owner consults across disciplines — architecture, backend, frontend, design, QA, security, data, research. They bring at least three independent sources of evidence. Each piece is weighted: high for a direct-expertise blocker or enabler, medium for a related consideration, low for a tangential note. A decision can’t proceed until it clears a points threshold proportional to its blast radius.

Seniority should set the bar for evidence, not substitute for it.

The effect is subtle but profound. The loudest voice no longer wins by volume; it wins by bringing better-weighted evidence, or it doesn’t win. Junior specialists with direct expertise suddenly carry “high” weight on their turf, regardless of title. And because the consultation is documented, every decision leaves an audit trail the team can learn from later.

It is, in effect, radical candor applied to decisions instead of feedback: the awkward step of writing down who you asked, what they said, and how much it counted — left deliberately in.

— Newsletter

A note every other Tuesday.
From the operating floor.

One field note, one playbook artefact, one pattern we’re currently watching across our portfolio. No filler, no roundups, no “industry news.”

~1000 operators · unsubscribe anytime