St. Pierre Dashboard Field Guide

What "a dashboard" means in this company

A dashboard is one web page that answers one question about one part of the business. Most of them read live from "the brain" — a Postgres database (a big spreadsheet-in-the-cloud) that stores everything the business knows.

A good dashboard is glanceable: you look, you learn one thing, you either act or move on. If a page needs a manual or ten minutes of study, it is doing too much. Some pages in the current list are actually documents or one-off reports pretending to be dashboards — a big part of this audit is separating those out.

The six rooms, one sentence each

• 💸 Finance — Can the business survive long enough to earn its first client cash? (hero: runway)
• 📣 Marketing — Turn creative volume into qualified leads at the lowest possible cost. (hero: CPQL)
• 📞 Sales — Move every real opportunity to the right: lead → booked call → paid client. (hero: RPL / cash collected)
• 📦 Fulfillment — Once a client pays, are we delivering the guarantee so they stay and renew? (hero: clients on guarantee pace)
• 🏗️ Projects — Are we building the machine, and can we keep the lights on while we build it? (hero: runway)
• 🔁 Loop Engine — Run the week as a loop: pick the targets, do the work, grade against them, adjust. (hero: weekly hit-rate)

Your routine — when to open what

If you only did this, you'd be running the business correctly. Daily surfaces are a five-minute morning loop. Weekly ones are a Sunday/Monday review. On-trigger ones you open only when a specific thing happens.

Mini glossary — the words that show up everywhere

Master table — all 27 surfaces

The recommended end-state

Per room: one canonical live dashboard, and everything else moves onto a labeled shelf so "the dashboards" list only contains things you actually monitor.

The three shelves the strays land on:

• Reports shelf — dated, generated files (Sunday Sales Brief, Target Lock, Execution Pack). Produced on a schedule, not monitored live.
• Docs / SOP shelf — playbooks and reference (Install SOP, Week-1 Runbook, Infra Backlog, Builder Onboarding).
• Tools shelf — interactive utilities you act inside, not read (Attribution Timeline, Profit First model, the :8765 sales cockpit).

Cross-room redundancy — the same data shown many times

Four overlapping KPI systems exist

Plus two catalogs with no data feed: sp_metric_registry (51 draft metrics) and the Sprint-5 target tables (all current_value = 0).

Tracking health of the live 25 (System A)

Live with data (13): MRR ($31,450 vs $42K target), live-clients (40), new-deals, creatives-launched (566/wk), qualified-booked-calls (0s), calls-on-calendar, show-ups (0), live-qualified-calls (0), closes (0), cash-collected ($3,004/wk), new-convos, booked-appts (7/wk — the $200 money metric), at-risk-clients (1).

Empty — missing feed (12): net-new-mrr, churned-mrr, churned-clients, cpqbc, no-show-recovery, ltgp-per-call, pct-clients-on-guarantee-pace, client-facing-errors, satisfaction-response-time, time-to-kpi.

Silent collector bugs (2): sur (show-up rate) and close-rate both have a source query defined but land zero rows — they look like real zeros but are actually broken collectors.

Data recency: daily 06-26 → 07-17, weekly roll-up 06-22 → 07-13. The HTML was last generated Jul 11 (the render step isn't scheduled — the data is fresher than the page). Nearly every target carries target_is_guess = true — placeholders, not Akash-ratified.

Akash's 12-core directive vs what's tracked

Covered live: new ads (creatives-launched), qualified convos (new-convos), pending calls (calls-on-calendar), completed calls (show-ups/live-qualified — currently 0), onboarded (new-deals), live clients, inspections booked (booked-appts), MRR (partial).

The five holes — core items with no home in the scorecard:

RECOMMENDATION — one core KPI, then three per department

Under it, at most three KPIs per department. Green = tracked today, amber = partially, red = not tracked.

Line-graph readiness — recommendation #1

Akash's standard: every kept dashboard should open with a 12-week line-graph hero. Today that's mostly impossible — the history doesn't exist yet.

• Only ~13 metrics have any history, and only ~4 weekly points each (MRR / live-clients / at-risk have just 2 weekly points).
• Cash / runway has 1 snapshot total — finance_metric_snapshots holds a single row of net_cash + runway_weeks, so no burn/runway line can be drawn yet.
• The old weekly_kpi_snapshots table has 387 rows back to Jan 2025, but they're stale EyeFly codes — useless for current SP graphs.

Recommended hero line series per kept dashboard (what each should trend once history accrues):

The room in plain English

This room answers one question: can the business survive long enough to earn its first client cash? Cash flow is oxygen, and right now the business is breathing through a straw.

Per the Cash OS directive (Jul 18), personal and business money are managed as one pool for operations — so the room blends them for survival, while keeping them separate for profit-and-loss and tax.

The hero number is runway = net cash ÷ weekly burn. Today the story is stark: business cash is $2,826.18, which is $376.03 below the draft cash floor, and the account reaches $0 in about 20 days at the current burn of $139.95/day. No St. Pierre client revenue is live yet; this month leaned on a $3,000 Fundbox loan draw — a warning sign, not income. The room's daily job is to point at the one move that brings cash in before the floor breaks.

Room KPI recommendation

Core company KPI this room feeds: Runway (days/weeks of cash before $0).

• 1. Runway to $0 (days) — tracked (point-in-time). Lagging. Reads 20.2 days to $0 / 0 days to the draft floor.
• 2. Trailing-30-day burn ($/day) — tracked. Leading. $139.95/day.
• 3. SP cash collected (sprint) → CPQL / RPL — partial. The number renders ($0.00) but it's a proxy, and CPQL/RPL are computed on no finance surface today.

Reconciliation: the KPI synthesis confirms Cash OS metrics live only in finance_metric_snapshots (1 row each, no history). The numbers exist for today, but there is no time-series to trend them yet.

Consolidation verdict

Confusing bits to fix (the wife test)

• Runway / cash floor / burn — define on first use in plain words on the tile.
• DRAFT vs ACTUAL / LOCKED tags — a newcomer won't know the floor is a not-yet-approved placeholder, so the red "BELOW FLOOR" partly reflects a guess. Say so.
• UNKNOWN striped grey cards look broken — add a legend: "Grey = no data source yet. We show UNKNOWN instead of a fake $0."
• Acronyms everywhere — MRR, AR, CAC, CPQL/CPQBC, RPL/LTGP, P0, WIG, opex are all undefined on the boards.
• Two different cash numbers — $2,826 / ~20 days vs −$198.85 / 0 weeks. Pick one framing, label the other.
• "Sprint / self-bridge / warm-5 / tripwire" — internal jargon; add a tooltip each.
• "Proxy" and "financing excluded" — explain a proxy is a stand-in, and that loan draws aren't counted as income.
• "St. Pierre Digital" on the P&L console — banned name; should read stpierre.ai.
• Dead personal-finance link — delist it; a newcomer just hits a browser error.
• Stale dates — add a loud "STALE — last updated N days ago" banner when a board is past its refresh window (e.g. consolidated-pnl at 27 days).

The room in plain English

This room has one job: turn creative volume into qualified leads at the lowest possible cost. You pour ads in the top (hooks, headlines, formats); what should come out the bottom is qualified leads — real people who fit the foundation & waterproofing niche and are worth a sales call.

The single number that says the machine is working is CPQL — cost per qualified lead = ad spend ÷ qualified leads. Lower is better.

Right now the machine is mostly idling on purpose: 532 ads built and paused, waiting for Akash to flip them live. A tiny canary is actually spending (~$123 total, which bought exactly 1 lead). So today the room is less "watch numbers climb" and more "get the inventory launched cleanly." Strategy is cold direct-offer at broad targeting — the copy does the targeting.

Room KPI recommendation

Core company KPI this room feeds: New MRR — qualified leads are the front of the revenue pipeline; CPQL is the leading cost signal.

• 1. CPQL (hero) — not tracked. Spend lives in sp_ad_metrics_daily (last 07-14), leads in sp_funnel_events (to 07-18), but no join filters "qualified" and divides by spend. Boards show raw CPL ($123.07 for 1 lead), never a qualified cost. This is the room's biggest measurement gap.
• 2. New ads launched / week — tracked (~566/wk), but counts built, not launched-live.
• 3. Qualified leads / week — partial. Raw leads captured; no "qualified" flag yet.

Consolidation verdict

Confusing bits to fix (the wife test)

• CPQL / CPL / CPQBC look interchangeable but mean different things — add a one-line glossary on-screen.
• "St. Pierre DIGITAL" on Creative Radar (logo + every ad row) — fix at the Meta page-name source.
• Hub says "$0 / Zero ads serving" while every cockpit shows real spend + 1 lead — reconcile it.
• "Insufficient-data" (120 of 122) sounds broken — say "too little spend to decide yet" in plain words.
• Format suffixes (Postit, Napkin, Lock, iMessage, Safari, X) are unexplained templates — add a legend.
• "No live-write" = "this snapshot can't update itself." Say it plainly.
• "VSL apps" = applications after watching the video sales letter — spell out on first use.
• CBO / Advantage+ / ABO appear with no definition — define on first use.
• Stage badges (visitor / engaged / applied / call booked / customer) have no legend for the order.
• "9.9d" in red everywhere reads as failure; the "expected while parked" note should sit next to the scary count, not only at the bottom.
• Attribution Timeline opens blank ("Select a lead…") — add a prominent "click any lead on the left" hint.

The room in plain English

This is the "People pipeline" room — where a lead becomes a booked call becomes a signed, paid client. Its one job: keep every real opportunity moving right. Lead in → qualified → booked call → close → cash.

The room runs on one source of truth. The honest answer today (refreshed Jul-22): 5 real contacts, 1 genuine booked call (a concrete-leveling contractor), 0 closes, $0 cash collected — show rate computes 100% off that single call (n=1). Two filter bugs were fixed Jul-22 (a test-flag leak and a booking clean-run window) that had been mis-counting; the walkthrough screenshots further down are the Jul-19 capture and predate the fix. The hero tile literally reads "UNPROVEN — $0 CASH" and stays red until signed-and-paid clients > 0.

Everything else is test data or plumbing the room quarantines so it can't pollute the real numbers. 15 of 16 CRM contacts and most GHL opportunities are flagged test and excluded — the room's most important discipline: no phantom pipeline.

Room KPI recommendation

Core company KPI this room feeds: Cash collected (entity = St. Pierre) → MRR. This room is the last mile before money lands.

• 1. RPL — revenue per lead (hero) — not tracked. Both inputs exist (cash $0; leads = 52 Typeform apps OR 127 funnel events OR 5 real contacts) but no surface divides them. The declared hero exists on no screen — the single biggest gap, and blocked on the leads-denominator decision.
• 2. Real booked calls — tracked. Leading. Honest 1 as of Jul-22 (a concrete-leveling contractor) after the booking clean-run fix; the remaining Calendly events stay test-excluded.
• 3. Cash collected — tracked. Lagging. Honest $0.00 because the deals table is live and empty (0 rows), not missing.

Note: this board runs its OWN 24-metric contract ("METRIC-CONTRACT sc-0.1.0"), a narrower system than the company's 25-metric scorecard_weekly. That divergence should be reconciled so "close rate" means one thing company-wide.

Consolidation verdict

Confusing bits to fix (the wife test)

• RPL / CPQL — the room is named for RPL but the term never appears on either screen; define it and show it.
• UNPROVEN / "honest 0" / "unknown, never 0" — one sentence up top: "0 = we checked, it's really zero. unknown = no trustworthy source yet."
• DRAFT / LOCKED / PROXY / UNSOURCED / UNKNOWN badges — five state tags with no legend.
• is_test / codex seeds / "real=1 vs 16 raw" — say plainly: "we hide fake/test rows; only 1 of 16 contacts is a real human."
• STOP / SMS opt-out / DND / suppression — explain the customer legally can't be texted; reach by phone/email.
• GHL, CRM, Typeform, Calendly, P-1380, sp_ghl_messages backfill — tool/ticket names with no gloss.
• "Weighted pipeline" / "5% DRAFT" / "$49 draft weighted value" — label "unproven estimate — not a forecast."
• Deadline colors (red ≤2d / amber ≤5d) — make the band legend global, near the tiles that reuse the colors.
• Source freshness 66.7% / STALE 4d–8d — add: "yellow source = the number below may be a few days old."
• "Published-workflow risk = 3" — the scariest tile; spell out what the 3 surfaces are and what "prove unpublished" means.
• Two different scorecards (24 here vs 25 company-wide) — tell the reader which is authoritative.

The room in plain English

This room answers one question: once a client has paid, are we actually delivering what we promised? stpierre.ai sells contractors a guarantee — a set number of leads (and booked inspections) per month. Fulfillment turns "cash collected" into "client stays happy and renews."

Three surfaces: one real dashboard (the Client Delivery Board, one row per each of 40 active clients) and two step-by-step playbooks (the 6-week install SOP and the Week-1 runbook) that are documents, not dashboards.

The stated hero is clients on guarantee pace — exactly right. The problem: there's no data behind it. All 40 clients show "no guarantee set" because nobody entered each client's monthly lead target, and the board never joins in leads actually delivered. So the hero renders blank across the board. The one live metric that matters is booked appointments (inspections booked) — 7/week, the "$200 money metric."

Room KPI recommendation

Core company KPI this room feeds: MRR retention — specifically churn (churned-clients / churned-MRR).

• 1. % clients on guarantee pace (hero) — not tracked. Target field NULL for 40/40 clients; the board doesn't compute pace. Fixing it is the #1 recommendation.
• 2. Booked appointments / inspections per week — tracked (7/wk, the "$200 money metric").
• 3. At-risk clients — partial. Renders, but sources disagree (scorecard says 1, delivery board shows 2).

Note: inspections delivered (vs booked) is NOT TRACKED ANYWHERE — if the guarantee is written on delivered inspections, that feed must be built before the hero can be trusted.

Consolidation verdict

Confusing bits to fix (the wife test)

• Niche codes CS / CL / CF / FR are never defined — add a one-line key (e.g. CS = crawl space, FR = foundation repair; confirm the real mappings).
• "Guarantee pace" is blank for 100% of clients — the room hero reading blank needs a louder callout: "⚠️ Room hero has NO data — targets not entered."
• Health vs Retention contradict each other — both at-risk clients show GREEN health next to AT-RISK retention. Explain what each measures and which wins.
• "Open cx" is unexplained — should say "open support tickets."
• "Days" column has no header explanation, and some rows show "—" with no reason.
• "No update drafted yet" on every row looks like 40 pending to-dos but is the empty state of an unbuilt feature — build it or hide the column.
• CSM shown as "—" on many rows — is that unassigned or N/A? Many clients have no success manager, itself a finding.
• Knoxville churn date in the past (2026-04-15) still counts as active — reconcile.
• install-sop link is dead and week1-runbook title says "St. Pierre Digital" — both erode trust in the link list.
• The 2-client install concurrency cap is invisible on the delivery board — a newcomer would never know onboarding is capacity-gated.

The room in plain English

This is the "are we building the machine, and can we keep the lights on while we do it?" room. It tracks the Project pipeline — the builder's side, where every unit of work is a numbered outcome (a thing that ships), not a person. Akash owns people and money; this room owns the build.

The one board that earns its keep is the Cockpit — the daily "what needs attention today?" entry point that fuses runway, the day's work blocks, habits, and the big finish lines. The rest is noise: two "project board" surfaces (project-check and projects-deployed) are the same page rendered twice and both show zero projects (a data lockout, not an empty business), and three more (infra-backlog, builder-onboarding, execution-pack) are documents, not dashboards.

Room KPI recommendation

Core company KPI this room feeds: Runway (weeks) = net cash ÷ weekly burn. This room is where money gets spent to build, not earned.

• 1. Infrastructure completion % — not tracked. Lives only as hand-typed percentages in the infra-backlog SOW (7 layers averaging ~60%).
• 2. Active projects with a dated next action — partial. The projects table has a next_action column, but the board that reads it is RLS-locked and renders 0.
• 3. Numbered outcomes shipped / week — not tracked. Derivable from handoffs/deliverables, but nothing rolls it into a weekly count.

Note: the live 25-metric scorecard has no builder / Project-pipeline lane at all. And Cash OS runway has no history to plot (1 row) — the cockpit's runway line is a projection off one baked point until a snapshot series is wired.

Consolidation verdict

Confusing bits to fix (the wife test)

• The zero-project boards read as "the business is empty" — 0/0/0/0 with no error message. Needs a visible "data unavailable" state, or just fix the RLS.
• Acronyms with no legend — DoD, ICE, CPQL, RPL, SLA, QA, BOF, L0/L1…L7 layers, v_dash, RLS, DND/STOP.
• "Baked fa-1.2.0" / "finance room · baked" means nothing to a newcomer — label it "as-of snapshot."
• The money hero shows a scary state without context — "20 days," "$376 below floor," "hot." Add a one-line caption ("cash is tight; runway is the number to watch").
• Income next 30 days = "--" looks broken, not "not modeled yet." Say which it is.
• The broken "This week" embed shows a broken-image icon — fix or remove.
• Habits grid all grey with "0 / 240" reads as "failed everything" — it means "nothing logged." Style the empty state differently.
• Three documents sit in the dashboard list — move them to a clearly-labeled docs/library shelf.

The room in plain English

This is the room where the week runs as a loop: pick the bars (project), do the work (execute), grade yourself against the bars (reflect), and adjust (learn). Everything else is a department; this room is the machine that keeps them honest week over week.

It has six surfaces that split into three jobs. Two are the target machine (target-lock drafts next week's bars; review-panel reads this week's actuals). Two are top-of-house goal views (hq and north-star). Two are reflection views (accomplishments and disciplines).

The problem is sprawl and double-counting. review-panel and target-lock are two halves of one system. hq and north-star both want to be "top of house," and hq is empty. The room can collapse to three live dashboards plus one weekly report without losing a single question.

Room KPI recommendation

Core company KPI this room feeds: Runway (indirect) — this room keeps execution on the bars that produce cash. Its truest north-star is the Rolling Weekly Hit-Rate.

• 1. Rolling weekly hit-rate — tracked. Share of locked target bars hit this week. Reading 33% now.
• 2. Ritual adherence rate — partial. The run log isn't capturing rituals being done, so the number is understated (reading 0%).
• 3. Weekly throughput — tracked. To-dos finished per week. 611 to-dos across 714 sessions in 6 weeks (~100/week, ~163 hours).

Consolidation verdict

Confusing bits to fix (the wife test)

• WIG (hq) — Wildly Important Goal. Define on the page.
• RAG pacing (north-star) = Red/Amber/Green. Not defined anywhere.
• Rolling hit-rate (review-panel) — the share of this week's target bars you hit; ≥75% lets next week stretch +10%.
• "bar" means "the target number" everywhere — say "target (the bar)" on first use.
• ICE-ranked (hq) = Impact / Confidence / Ease. Unexplained.
• speed-to-lead (lead actions) = call/text a new lead within minutes.
• The rite names (Sunday Projections, Exception Router, Friday Reflection, Nightly Actuals) are jargon — add a legend mapping each to project/execute/reflect.
• Empty ≠ broken, and the two empties differ: hq (blank via RPC) reads like the RLS lockout (plumbing); north-star ("No structured goals") is genuinely unpopulated (data entry). Each empty state should say which it is.
• Stale/contradictory labels: review-panel's header says "current week Jul 6 – Jul 12" (a week stale) and shows three metric counts (15 / 14 / 60). disciplines grades F off an admittedly-empty log.
• Banned company name on two boards: accomplishments subtitle "St. Pierre Digital" and hq footer "EyeFly-HQ-Dashboard." Replace with stpierre.ai.
• "~" times (accomplishments) are estimates scaled from partial clock data (134 of 714 sessions) — legend the tilde near the big 163.2h number.

Reminder: nothing above has been changed. This Fix List is the recommended backlog — approve the rows you want, in the order you want, and the Mechanical ones can be executed without further decisions.