Task Methodology — Audit & Improve an Existing Website¶

Updated: 2026-06-02

Library: WDM Methodology Library Type: Task methodology Composes on: wdm-website-build.md — the manager methodology for WDM's overall operating reality. This task methodology is the procedure WDM runs when the trigger fires; the manager methodology is the framing it runs inside.

1. Purpose & when to use¶

Use this when an existing site needs to be scored, fixed, and improved against the outcome that matters for its entity type. This is the procedure WDM ran on Iron Mesa USA (composite 53/100, three production bugs fixed in place, top-10 ship shortlist, owner-input notepad) — generalized so it works for any of the six classifications.

The method does NOT assume "maximize sales." It maximizes the primary conversion event for the site's classification, which is defined per type below. For two of the six types (Nonprofit Informational, and partly Nonprofit Donation) the goal is not revenue at all.

Trigger patterns¶

Recognition signal: Brandon (or another principal/manager) says any of:

"Audit [entity] site"
"Run the audit on [entity]"
"Score [entity]'s website"
"Execute the audit document on [entity / all websites]"
"Improve / fix up [entity] site"
"Where do we stand on [entity]'s site?"

The entity may be ATG-side (SEQ, STE, CTP, SP, IMA, Cactus Diesel, Superior Service), partner-side, or external. The trigger applies to any live site WDM has access to.

First action: Open this document. Classify the site against the six types in Section 3. Begin the 11-step procedure in Section 5.

What you don't do:

Don't ask which audit format to use — this one is canonical
Don't ask for a budget envelope for the audit itself — auditing is in-scope by default
Don't fabricate dollar-figure revenue estimates per recommendation (see Section 6)
Don't ship invented "facts" about the entity to fill in placeholder content — capture in the owner-input notepad and proceed on items that don't require owner input (Section 5, step 10)

2. The universal audit core (every site, every type)¶

Score each dimension 0–100 with evidence. These are the same ten dimensions for every site; only their weight changes by classification (Section 4).

Performance — Core Web Vitals (LCP, INP, CLS), mobile load speed, image optimization, render-blocking resources.
Technical SEO — crawlability, metadata, structured data presence, internal linking, sitemap/robots, canonical correctness.
Accessibility — WCAG: contrast, focus states, semantic HTML, keyboard nav, alt text, form labels, error accessibility.
Mobile UX — touch-target size, readability, sticky contact, no broken popups (Android Chrome blocks them — use inline modals), no horizontal scroll.
Information Architecture — nav clarity, findability, depth-to-key-info, menu organization.
Visual / Design consistency — typography hierarchy, color system, component consistency, whitespace, brand alignment.
Trust fundamentals — real contact info, business legitimacy, SSL, NO placeholder/broken content (Formspree placeholders, [TODO] brackets, lorem ipsum all score this to zero).
Content quality — clarity, scannability, no filler, answers the buyer's actual questions.
Conversion mechanics — primary-CTA presence/placement/wording, friction on the primary action, secondary-action clarity.
Analytics / tracking — is the primary conversion event actually measurable? If you can't measure it, you can't improve it.

3. Classification lenses¶

For each classification: the primary conversion event, the entity-specific checks (beyond the universal core), and the anti-patterns that disqualify a site for that type.

High-Ticket B2B Sales — (e.g. Iron Mesa USA — truck sales)¶

Primary conversion: qualified inquiry — phone call, quote request, or "check availability" on a specific unit. High consideration, long cycle, no online checkout, no card processing.
Entity-specific checks: per-unit detail pages with full specs + multiple real photos + honest condition disclosure (critical for salvage rebuilds) + VIN; financing-partner visibility; before/after on rebuilds; sticky tap-to-call on mobile; inventory freshness (sold units removed or marked); Vehicle schema; related-units rail.
Anti-patterns: stock photos; no condition disclosure; price hidden with no "request quote" path; contact buried; stale "sold" inventory.

Used-Item E-Commerce — (e.g. Cactus Truck Parts — used parts from own dismantling)¶

Primary conversion: add-to-cart → checkout on an in-stock used item; secondary = fitment inquiry.
Entity-specific checks: fitment/compatibility data; condition grade shown; per-item real photo of the actual unit (one-of-each inventory — not a representative image); explicit "this exact part" vs "representative" labeling; qty-1 / auto-delist-on-sale handling; Product + Offer schema with itemCondition: UsedCondition; strong search/filter for SKU sprawl.
Anti-patterns: stock photos on used parts; no fitment data; sold items still listed; no condition grade; weak search.

New-Item E-Commerce / Lead-Gen — (e.g. Superior Parts new imports; Superior Water Truck Parts)¶

Dual-mode: transactable items checkout (Stripe); high-value/configurable items go to quote/lead capture. Audit must handle both paths.
Primary conversion: completed checkout for transactable items; quote request / lead capture for high-value commercial items.
Entity-specific checks: category depth + search/filter; bundles, frequently-bought-together, cross-sells (legit AOV plays here, unlike trucks); shipping-cost transparency; B2B / fleet account path + quote-request workflow; email capture with incentive; reorder/subscription for consumables; Product + Review schema.
Anti-patterns: treating a quote-driven catalog as pure transactional; no fleet/B2B path; hidden shipping cost; no email capture.

Nonprofit Informational — (e.g. NewPath Foundation info presence)¶

Primary conversion: intent capture + institutional credibility — newsletter signup, program/volunteer inquiry, contact. Not revenue. Audience includes grantors and partners, not just the public.
Entity-specific checks: 501c3 status + EIN visible; impact/outcomes data; clear program descriptions; leadership/board credibility; press/partner logos; newsletter capture; NGO/Organization schema.
Anti-patterns: bolting ecommerce conversion tactics onto an info site; donation-begging on an info page; opacity about finances or outcomes; fake urgency.

Nonprofit Donation — (e.g. NewPath donation flow)¶

Primary conversion: completed donation, with recurring gifts weighted higher (the LTV analog).
Entity-specific checks: dead-simple donate flow (this is the "checkout" — minimize steps); suggested amounts + recurring-default option; impact-per-dollar framing ("$X provides Y"); tax-deductibility statement + receipt clarity; mobile-optimized donate; donor trust signals (Charity Navigator / GuideStar if applicable); processor/security trust.
Anti-patterns: long multi-step donate flow; no recurring option; vague use-of-funds; donate CTA buried; desktop-only donate UX.

Service Business — (e.g. transport/service arms)¶

Primary conversion: inquiry / quote / booking — call, quote form, service request.
Entity-specific checks: clear service-area definition; licensing / insurance / bonding shown; reviews/testimonials; quote or booking form; sticky tap-to-call; LocalBusiness schema; proof of work / case studies.
Anti-patterns: no service-area clarity; missing licensing/trust signals; slow or buried contact; no proof of work.

4. Classification scoring matrix¶

Composite = weighted average of the ten dimension scores, using the weight column for the site's classification. Weights are tiers — 3 = high, 2 = medium, 1 = low — chosen to be honestly approximate, not falsely precise. WDM may tune per site.

Formula: composite = Σ(dimension_score × weight) / Σ(weight), result on a 0–100 scale.

Dimension	HT-B2B	Used-EC	New-EC/Lead	NP-Info	NP-Donation	Service
Performance	2	3	3	2	3	2
Technical SEO	2	3	3	2	1	3
Accessibility	1	2	2	3	2	2
Mobile UX	3	3	3	3	3	3
Information Architecture	2	3	3	2	1	2
Visual / Design	2	2	2	2	2	2
Trust	3	3	2	3	3	3
Content Quality	2	2	2	3	2	2
Conversion Mechanics	3	3	3	2	3	3
Analytics / Tracking	2	2	3	2	3	2

The weighting is the whole point: a trust failure costs a High-Ticket B2B site far more than it costs an info site, and an SEO gap barely dents a donation-page score. The composite tells you where the bleverage is for this entity type, not in the abstract.

5. The audit procedure¶

Classify. Identify the site's classification from the six. If it spans two (common for New-EC/Lead-Gen), audit both paths.
Snapshot. Archive a pre-fix snapshot of the live site before touching anything. (Iron Mesa pattern.)
Inventory. Crawl pages; record current state of each.
Score the ten universal dimensions 0–100, each with concrete evidence — not a vibe.
Apply the classification lens — run the entity-specific checks and flag anti-patterns from Section 3.
Compute the weighted composite using the classification's column in Section 4.
Write findings per dimension: problem → who it costs and roughly how much friction (tier, not a fabricated dollar figure) → exact fix → production-ready code where applicable.
Fix production bugs in place — placeholders, broken forms, [TODO]s, missing hero CTA — and log each fix. (Iron Mesa found and fixed three this way.)
Publish a ranked ship shortlist (top ~10), tiered: Quick Wins (1–7 days) / 30-day / 90-day, ranked by impact tier × effort.
Open-inputs notepad. Capture everything that needs the owner (cert facts, founding story, logos, photo paths, financing partners) in D:\claude_www\notepads\<site>.md. Proceed on queued items that don't need owner input.
Report composite + per-dimension scores + shortlist to Brandon. Keep the pre-fix snapshot archived.

6. On "revenue impact" estimates¶

The source prompts that inspired this demand a dollar figure per recommendation. Don't fabricate them — invented precision is worse than an honest tier. Rank by impact tier (High / Med / Low) × effort. A real dollar figure is only stated when there's a real number behind it (e.g. measured conversion rate × known traffic × known order value). Otherwise the shortlist ranking carries the prioritization.

7. Composition & deployment¶

This task methodology composes on top of:

wdm-website-build.md — the manager methodology that frames WDM's overall operating reality. Default assumptions (Tailwind, Cloudflare Pages, Supabase, design tokens, single-file HTML preference) inherit from there.
The six classification methodologies — one per site type (HT-B2B, Used-EC, New-EC/Lead, NP-Info, NP-Donation, Service). These supply the per-classification "primary conversion event," entity-specific checks, and anti-patterns referenced in Section 3 above. Status as of 2026-06-02: the six classification docs do not yet have authored audit-lens sections; the detail in Section 3 of this document is currently the canonical source. Authoring those per-classification sections is queued as a follow-on (per mc1-code 2026-05-30 supplement).

Deploy path in the library: methodology/wdm/tasks/audit-existing-site.md.

8. Failure modes & non-fabrication rules¶

The audit fails honestly if any of these get violated:

Fabricating dollar figures. Per-recommendation dollar estimates without measurement behind them invent precision the audit can’t defend. Use impact tiers (High / Med / Low) ranked by effort.
Fabricating facts about the entity. Years-in-business, certifications, customer logos, response times, warranty terms, financing partners, employee counts — if it’s not on the live site or confirmed by the owner, it doesn’t go on the new site. Capture in the owner-input notepad (Section 5, step 10) and proceed with what doesn’t require owner input.
Treating the universal core as the only thing that matters. A site that scores 90/100 on universal dimensions but fails its classification’s conversion event is not a successful audit. The classification lens is the work; the universal core is the foundation underneath it.
Auditing without snapshotting. Step 2 of Section 5 is non-negotiable. Pre-fix snapshots in _archive_v0_<stamp>-<purpose>/ are how rollback works if a fix goes wrong.
Closing the audit without an owner-input notepad. Most audits surface 5-10 items that require owner input (facts, photos, decisions). If the notepad doesn’t exist at the end, the gated work is invisible and gets re-discovered on the next audit cycle. The notepad is the durable artifact.