Skip to content
XConversionLab

The intellectual core

The studio runs on an operating system. In public.

Internally we call it PortfolioOS: a written constitution, a thirteen-stage lifecycle, and a set of frameworks that make honesty structural instead of aspirational. This page is the method, unhidden — because a systems studio that keeps its systems secret is asking to be taken on faith.

Three frameworks carry the weight

Evidence levels

Every claim we publish is tagged measured, observed, or judgment — declared at the research stage, before anyone knows what would be convenient to claim. Fabrication isn't forbidden; it's structurally impossible.

Decision logs

Significant choices are recorded the day they're made: context, options, decision, reasoning. Case studies are assembled from these logs, which is why they read like engineering and not like marketing.

The nature taxonomy

Client, Internal, Concept, Research, Experiment — every project carries its label on every card, in every hero, and inside the structured data machines read. Concept work is an asset only when it can never be mistaken for commissioned work.

The thirteen-stage lifecycle

Every project moves through the same recorded stages. Each stage ends with a filed artifact and a signed checklist — no artifact, no advance.

  1. 01Lead

    Every serious conversation gets a permanent ID before any work begins. Dead leads keep their record — three years of intake forms teach more about who buys and why than any consultant.

  2. 02Discovery

    Structured conversations about the actual business — revenue mechanics, the funnel as it really is. Ends with a written problem statement the client confirms in their own words.

  3. 03Research

    Evidence-gathering, proportional to stakes, with an honest declaration: is this project's evidence measured, observed, or judgment? That ceiling on future claims is set now — before anyone knows what would be convenient.

  4. 04Strategy

    The project's thesis: the written argument connecting evidence to approach. Design doesn't start until the argument is agreed, because revising paragraphs is free and revising pixels isn't.

  5. 05Planning

    Sitemap, content inventory, and a scope ledger with an explicit out-list. The out-list prevents more disputes than the contract does.

  6. 06Design

    Tokens first, then screens — every project gets its own miniature design system in its own aesthetic. Rejected directions are filed, not deleted; they're future patterns.

  7. 07Review

    Internal review against checklists, then client review anchored to the strategy document — feedback resolves against what we agreed, not against taste.

  8. 08Development

    Built under performance and accessibility budgets set before the first line of code. Every gotcha is written down the day it happens.

  9. 09QA

    Verification, not exploration: functional, accessibility (WCAG 2.2 AA), performance, and content passes — with results filed. Nothing is silently accepted.

  10. 10Launch

    A checklist, executed. Every legacy URL redirected, every form tested to its destination, the repository tagged. Launches should be boring; boring is what prepared looks like.

  11. 11Capture

    Within 72 hours: the photograph of record. Full screenshots at defined breakpoints, key flows recorded, everything dated. The client will change the site; this window never reopens.

  12. 12Case study

    Assembled from the filed record — decision logs, research, QA reports — never from memory. Outcome claims wait for outcomes; we publish in two releases when measurement needs time.

  13. 13Retrospective

    One structured hour: what worked, what didn't, and at least one pattern or failure extracted into the studio's permanent knowledge base. This is the stage agencies skip — and skipping it is why deliverable shops stay deliverable shops.

Stages 11–13 in blue are the ones the industry skips. They're where the compounding lives.

When the rules don't answer, six tests do

Decisions the system doesn't already answer run through this order — higher tests beat lower ones, and the winner is written down so the decision never needs making twice.

  1. 1

    Honesty. Does any option involve an invented, exaggerated, or ambiguous claim? Eliminate it. This outranks everything, including commercial pressure.

  2. 2

    Ten years. Which option is still correct in 2036? Prefer it.

  3. 3

    System. Does this create a reusable rule or a one-off exception? Prefer the rule.

  4. 4

    Reader. Which option most respects the time and intelligence of whoever consumes the result?

  5. 5

    Maintenance. Which option costs least to keep true? An unmaintained feature is a lie on a timer.

  6. 6

    Founder's tiebreak. If all tests pass equally, the founder decides — and the decision becomes permanent infrastructure.

The method is public because it's inspectable in the work itself — every case study in the register was produced by exactly this machine.

Inspect the evidence