34 protocol points
across six audit cycles.
Qahwa Coffee Group is the founder's own specialty coffee brand — built from scratch in Cape Town as a mobile vendor, scaled to a permanent location, then restructured for international green-bean distribution. Currently dormant. The Qahwa website was designed and built by the founder, then audited under the Master Execution Protocol over six sequential cycles in May 2026. Opening score: 63.5%. Closing score: 97.5%. This is the engineering transcript — what was found, what was changed, and what the methodology looks like when applied transparently to the founder's own surface.
What 63.5% actually meant.
The pre-audit state of the Qahwa site was not a brand in trouble. It scored 63.5% — meaning roughly two-thirds of execution surfaces were already protocol-aligned, but the failing third was concentrated in the highest-leverage dimensions: Translation Engine, Editorial Restraint, and Authority Architecture.
A site built by the founder, with strong intent but the same drift pattern observed across most specialty operator surfaces: substance present, signal dispersed. Documented sourcing language overshadowed by marketing-tier vocabulary. Documentation chain implied but not displayed. The MEP exists precisely because this pattern is so consistent across the category.
The opening audit identified ten ranked failures. The remediation sequence was structured so that fixing the first would compound into the next four — which is how a 34-point delta becomes possible across six cycles instead of six months.
10 ranked failures identified. 4 concentrated in the lowest-scoring dimensions (Translation Engine 49%, Editorial Restraint 58%, Accessibility 59%). These became remediation priority 1 through 4.
The ten execution failures,
ranked by impact-per-week.
Evidence layer absent above the fold.
Hero positioned the brand as premium but failed to ship a single piece of measurable evidence — no origin certificates referenced, no specific cupping scores, no documented supply chain markers. The most expensive page on the site was selling on adjectives.
Conversion architecture indistinct.
Three different primary CTAs across three sections, all competing. No defined buyer next-step. A specialty importer surface was offering the cognitive load of a SaaS marketplace — exactly the editorial drift the protocol exists to catch.
Visual discipline drift in typography hierarchy.
Six different sizes of "heading" deployed across the homepage, no clear h1/h2/h3 logic. Buyer subconsciously read this as "they don't know what's important here" — which then transferred to "they don't know what's important about the coffee."
Hospitality layer missing on first contact.
No accessible "next step" for a buyer who wanted to enquire without committing to a 60kg order. No sample request mechanism. No documented response time. Premium operators don't make their buyers guess what's possible.
Trust architecture under-specified.
Geographic locations mentioned in passing (Dubai, South Africa) but no documentation of cooperative relationships, no named producers, no specific shipping partners. Trust signals were implied, not stated.
Page-builder dependency breaking component governance.
Site rendered through an Elementor 4.0.8 page-builder layer that generated inline CSS with hardcoded hex values, broke the token taxonomy, and bypassed semantic HTML on interactive elements. Component registry compliance was structurally impossible until the builder dependency was removed.
Information architecture flat.
All content tiers competing for the same visual weight. The "available lots" page (the operational core of the business) read with the same priority as the journal teaser (informational supplement).
Performance hooks unoptimised.
Parent theme injecting 280KB of unused WooCommerce + emoji + block library CSS on every page. Time-to-interactive lagging by 1.4 seconds on 3G. Premium buyer doesn't tolerate slow.
Engineering excellence below documentation threshold.
No semantic HTML structure on hero. ARIA labels missing on navigation. Skip link absent. Performance acceptable but accessibility audit would have failed multiple WCAG 2.1 AA checks.
Self-audit compliance not built in.
No internal mechanism for re-scoring drift. Any improvement risked regressing within a quarter because the protocol existed in the operator's head, not in a documented standard the next hire could inherit.
Six audit cycles.
Two days.
Opening audit and remediation of Translation Engine + Authority Architecture.
Site re-audited end-to-end. Failures #1 and #5 above remediated first because they were prerequisites for the others — adding documented sourcing language plus authority surfaces (Sub-Identifier, Proof Line, Evidence Strip) to the hero changes how the entire downstream editorial pattern is read by the buyer.
Ticker introduced as evidence-tier surface. Documentation chain (ICO Certificate of Origin, Phytosanitary, Cooperative Cupping Report) named explicitly. Geographic specificity tightened from "East Africa" to "Muyinga Province · Kavugangoma Washing Station."
Editorial Restraint + Wholesale Readiness remediation.
Three competing CTAs collapsed to one primary path (Available Lots) with one secondary (Sample Request). All other surfaces became downstream of this single buyer journey. The site started behaving like a structured commercial proposition instead of an information menu.
Direct correspondence channel introduced as the hospitality layer. Response time documented (24 hours). Sample request mechanism wired with explicit non-commitment language. Buyer cognitive load reduced from "where do I click?" to "do I want to start the conversation?"
Atmospheric Integrity + Protocol Compliance restructure.
Typography hierarchy collapsed from six heading sizes to three (h1/h2/h3 with strict logic). Display and body typeface pairing locked. Italic emphasis restricted to a single editorial pattern. Brand surface started reading as one coherent voice instead of six competing ones.
Information architecture re-sequenced. Available Lots (operational core) elevated to primary navigation tier. Catering page fully migrated from Elementor 4.0.8 to a governed page-catering.php child-theme template — eliminating the largest single deficit in the audit series and bringing the page under full token-system and component-registry compliance. Journal repositioned as proof-of-method rather than competing for buyer attention with the offer.
Engineering Excellence, Performance, Mobile Fidelity, Accessibility.
Parent theme suppression hardened. Unused WooCommerce + emoji + block library CSS dequeued. Semantic HTML refactored across all six page templates. ARIA labels added to nav and main landmarks. Skip link implemented. WCAG 2.1 AA target hit.
Brand coherence locked at the constitutional layer. Self-audit mechanism built in via the MEP v1.5 Part 17 Self-Audit block (visible at the top of every code file). Every future change has to score itself against the same 10 dimensions — drift can no longer be invisible.
Ten dimensions.
Before, after, delta.
| Dimension | Before | After | Delta |
|---|---|---|---|
| 01 · Protocol Compliance | 0% | 62% | 0 |
| 02 · Atmospheric Integrity | 0% | 71% | 0 |
| 03 · Editorial Restraint | 0% | 58% | 0 |
| 04 · Authority Architecture | 0% | 65% | 0 |
| 05 · Translation Engine | 0% | 49% | 0 |
| 06 · Engineering Excellence | 0% | 72% | 0 |
| 07 · Wholesale Readiness | 0% | 68% | 0 |
| 08 · Mobile Fidelity | 0% | 61% | 0 |
| 09 · Accessibility | 0% | 59% | 0 |
| 10 · Performance | 0% | 70% | 0 |
A 34-point composite improvement across 10 dimensions, delivered through six sequential audit cycles in May 2026. The protocol now exists in code, in copy, and in the operator's daily workflow — not just in the consultant's head.
The drift. 82.7%.
Four months after closing at 97.5%, the Qahwa website was re-audited under the same 10 MEP dimensions. The score had dropped to 82.7%.
What caused the regression: The Available Lots page was rebuilt in Elementor rather than the custom child theme, introducing competing class architectures, split typography systems, and mixed WebP delivery. Content grew without schema markup. Authority gaps — no founder visibility, no named testimonials, no press coverage — remained unaddressed.
What this proves: The protocol achieves the ceiling. Governance is what holds it. Without ongoing protocol stewardship, drift is inevitable. The site didn’t break — it slowly unwound. That slow unwinding is the argument for the retainer.
What we'd do differently.
Build the self-audit mechanism first, not last.
The Part 17 self-audit blocks at the top of every code file were added in the final cycle. In hindsight, building this mechanism first would have accelerated every subsequent audit — each cycle would have written into a standardised template instead of being re-invented. Future engagements: the self-audit gate is Audit 1, not Audit 6.
Self-audited pace was operator-tolerable, not optimal.
The compressed audit series worked because the founder was simultaneously operator, subject, and auditor — decisions could be made instantly across sessions. Most engagements should not compress this hard. The standard 6–8 week Remediation gives a real client time to integrate changes into their team's workflow, which is where compounding gains come from.
More artifact, less narrative.
The MEP v1.5 documentation produced during this audit series is dense — a 78-check master register, ten dimensional checklists, a Field Guide. Future engagements would split this into two layers: a short operating protocol the client uses daily, and a longer reference protocol kept in the appendix.
If the protocol I'm offering to other operators couldn't move my own brand from 63.5% to 97.5%, it wouldn't be worth charging anyone for. And if the score drifts back to 82.7% without governance, that proves the ongoing need for the protocol — not just the one-time remediation. The transcript is published, the methodology is reproducible, and the proof is transparent.
The recovery. 98.3%.
Act Three proved that without governance, drift is inevitable. Act Four proves the protocol can reverse it. In September 2026, the Qahwa website was re-evaluated under all 10 MEP dimensions across six sequential audit cycles. Every finding was actioned. Every improvement was measured.
What the protocol found: personal email undermining authority, missing meta description, zero WebP images, no per-lot conversion paths, accessibility gaps in alt text and ARIA coverage, no explicit image dimensions causing layout shift. Each issue was ranked by dimensional impact and fixed in sequence. The score moved in one direction.
Dimension Breakdown · Audit #12
From 82.7% drift to 98.3% recovery across five audit cycles. Eight dimensions at 100%. The protocol achieves the ceiling. Governance holds it. And when governance returns, the ceiling is recoverable.
The drift to 82.7% proved that the protocol is necessary. The recovery to 98.3% proves it works. Every improvement was measured against the same 10 dimensions, the same 78 checks, the same methodology. Nothing was assumed. Everything was verified. That is what the MEP delivers — not a feeling that your website is better, but a number that proves it is.
What would your score be?
Submit your website. I’ll personally evaluate it against the same 10 dimensions and 78 checks you’ve just read about. Same methodology. Same transparency. Every engagement scoped individually.