Appendix E. Assessment lenses, client dashboards, accounts, sentiment and website

E.1 Org assessment lenses

The Diagnose stage opens on an assessment dashboard that reads the 27 practices through four lenses. Lenses are a view over the same framework, not a second model: every practice belongs to exactly one lens.

• Score. A lens score is the weighted mean of its practice means, using the framework's practice and dimension weights. The confidence interval comes from respondent-level lens composites; a respondent counts only when they answered practices carrying at least 60% of the lens weight. Bands are the framework's bands.
• Peers. Each lens is placed against the middle half and median of the firm's own book: opted-in, completed engagements in the same industry group, never the current client.
• Dashboard. The index ring and band; one tile per lens with its ring, band, difference from the peer median, a bullet against peers, and its practices as small bars; lens by site as a heatmap (differences from the company, suppressed cells hatched); lens by level as profile lines; the lowest practices; links into the existing deep-dive pages.
• Lens deep dive (/diagnose/<lens>). Hero with score, band, interval, share of explained variance and peer bullet; practices with intervals and peer medians; the lens by level; practices by site; what people say (themes and client-safe quotes); moves from the pack mapped to the firm's offerings. The Structure lens adds the shape of the organization from the directory: headcount by level, span of control and layers by site against the peer span.
• Colour. The four lens colours are a validated categorical set (light and dark). They always appear with a label, never alone.

E.2 Portfolio, client dashboards and project switching

• Firm view. The Overview has two views, For you and Firm. Firm shows active engagements, fees in flight, people in fieldwork now, the size of the book, the portfolio as a board by stage, what the book is saying and the firm workspace. For you opens with quick actions and the working engagement’s four lenses and sentiment.
• Projects. A board by stage (default) or a list. Every card shows the client, package, due date, risk and progress.
• Switcher. The project name in the sidebar and in the page header is a switcher. It lists every project with its five-stage track, so a consultant can see where each one is and jump between them from any page.
• Client dashboard (/project/<id>). One button for the next step, then every action in three groups: Plan (write the brief, create or import the survey), Run the survey (preview, invite people, launch, track responses) and Results (see results, sentiment, build the report, share with the client). Each shows where it stands. Below: the stage tracker, and results at a glance (index, four lenses, sentiment) or the snapshot for the stage the project is in.
• Per-project work. Each project keeps its own brief and survey. Switching project stores what you were editing and brings back the other project’s; a new project starts from a blank brief and the framework’s questions. Pages that need a population, fieldwork or responses show a plain “no data yet” state for projects without them, with the next useful step. In the prototype only Ferrovane carries responses.

E.3 Accounts and the firm name

• Sign in. People choose their account on this device or add one (name, work email, role, job title). The default account is India Goureau, Partner. Switching person keeps you on the same page, so a draft and its approval can be shown side by side. There are no passwords in the prototype; accounts live in the browser.
• Who did what. Approvals, notes, hypotheses and insight histories are signed with the person who made them.
• Client previews. The account menu previews the portal as the client sponsor or a client manager, and returns to your own view.
• Firm name. Set in Settings (or when the first account is created). It appears on the survey, the portal and every export; when blank, client surfaces show only the client’s name and exports say “Prepared for ”.

E.4 Sentiment

• What is scored. Every open-text comment, sentence by sentence, from −1 (very negative) to +1 (very positive). A comment is positive above +0.05 and negative below −0.05. Net sentiment is the positive share minus the negative share, from −100 to +100.
• How. A word list with valences from −3 to +3, a negation rule (“not fair” flips and softens), intensifiers (“really good”), dampeners, and a contrast rule (after “but” counts more than before it). Scores are normalised with x / √(x² + 15). The method is deterministic and every score can be explained word by word; nothing is sent to a model.
• Where it shows. A Sentiment page in Diagnose (overall mix, by lens, by topic, by site, by level, most positive and most negative comments, and a check against the scores by site), a chip on the assessment dashboard, a panel in each lens deep dive, a tile on the client dashboard and Overview, and an answer in Ask.
• Privacy. Groups with fewer comments than the comment threshold (10, or 15 in works-council mode) are hidden. Quotes come only from sites that clear it, and carry the level only when the site-and-level group clears it too.
• Prototype data. Comments are generated from each respondent’s own answers and coded themes; the scorer never sees how a sentence was chosen. Across the sentence bank it reads 81 of 81 sentences as written, and across sites, sentiment tracks the health index (r ≈ 0.8).

E.5 Website

The public site at orgdiagnostic.com presents the product. It is a separate site and deployment from the app, which lives at app.orgdiagnostic.com with the specification at app.orgdiagnostic.com/spec/.

• Source. site/ holds plain HTML, CSS and JavaScript with one self-hosted font (Plus Jakarta Sans, OFL). It has no build step and is deployed on its own; links to the app use its full address.
• Media. Screenshots and films are recorded from the prototype itself. Films ship as H.264 MP4 with a WebM fallback, play only when on screen, and do not autoplay under reduced motion.
• Claims. Comparisons are by category of tool (employee survey platforms, survey and form builders, org design tools, decks and spreadsheets), never by named product, and carry a footnote. Simulated Microsoft 365 connections are labelled as simulated. No customer logos or testimonials until they are real.
• Motion. Scroll reveals, a hero that flattens as you scroll, a sticky five-stage story, count-up rings and a pointer glow. prefers-reduced-motion turns all of it into a static page; prefers-reduced-transparency makes glass surfaces solid.
• Hosting. Two Vercel projects from one repository: the app and spec from the root (vercel.json), the website from site/ (site/vercel.json). Each skips a deploy when a push only changes the other. Each half also has a wrangler.jsonc for Cloudflare Workers static assets.