Metro Detroit Lead Gen System

Working lead-gen tracker for a solo low-voltage / structured cabling business in metro Detroit, with an adaptive lead-scoring algorithm that learns from real outcomes.

• Metro_Detroit_Lead_Gen_System.xlsx — the pipeline CRM. Tabs: Read Me (full algorithm writeup + sources, incl. how larger companies source/advertise for this work), Channel Performance (the adaptive scoring engine, 10 channels), Dashboard (live formulas), Pipeline (master prospect tracker, 32 seeded leads), Weekly Call Rotation, Events (in-person networking calendar), Competitor Intel (6 named metro Detroit incumbents, sourced).
• Outreach_Toolkit.docx — capabilities one-pager, cold-call scripts, follow-up email templates by channel, and an In-Person Event Playbook (what to bring, pitch, follow-up).
• AsBuilt_Documentation_Template.docx — TIA-606-structured as-built documentation to fill in and deliver with every completed job (cable schedule, labeling legend, test results, photos, sign-off).
• WEEKLY_LOG.md — dated log the weekly routine appends to (created on first run).
• QUICKLOG_ISSUES.md — created by the quick-log processor (below) when a dictated log email can't be confidently matched to a Pipeline row.

The scoring algorithm

Lead Score = Base Fit × Channel Multiplier, clipped to 1-10. Base Fit, Channel Multiplier, Channel Confidence, and Lead Score all live in the Pipeline sheet; Channel Multiplier, Channel Confidence, and Lead Score are live Excel formulas, not values the routine sets.

• Base Fit (1-10, column D) — set once per lead, company-level only: 0.35×Recurring Revenue Potential + 0.25×Deal Size + 0.20×Urgency Signal + 0.20×Access Ease (each sub-score 1-10, judgment call at lead-creation time).
• Channel Multiplier (column E) — INDEX/MATCH pull from the Channel Performance tab's Multiplier column for that row's Channel. Range 0.6x-1.4x. Recalculates instantly, no waiting for a routine run.
• Channel Confidence (column F) — INDEX/MATCH pull from Channel Performance's Confidence column: a plain-language label (No data, Low, Building, High) based on Evidence Share = Attempts / (Attempts + K) — how much of the Smoothed Win Rate is real data vs. the starting Prior. A high Lead Score on Low confidence is still mostly a guess.
• Channel Performance tab computes, per channel, from the Pipeline's own Status column: Attempts, Won, Lost, Raw Win Rate, and a Smoothed Win Rate via empirical Bayesian smoothing: (Prior×K + Won) / (K + Attempts), K=5. Multiplier = 0.6 + 0.8×Smoothed Win Rate. Evidence Share = Attempts/(Attempts+K) drives the Confidence label. Starting Prior Win Rate per channel is a documented, sourced assumption (see the tab's Source/Rationale column and the Read Me tab) — not measured data, and it's designed to be overridden by real results within a handful of attempts per channel. 10 channels: Property Mgmt, CRE Broker, MSP, TI Electrical/GC, Platform, Bid Board, Vendor Program, DC Integrator/Staffing, Association/Networking, Direct Corporate Prospect (new — a company with a specific, dated, sourced relocation/expansion signal you call directly; prior 0.16, above generic cold-outbound since the need is documented, not guessed).
• Service Match (column H) — tags each lead against Austin's actual service list: Fiber, Copper/Structured Cabling, Decommissions, Network Setup, Security Camera & Access Control, Cloud/IT Support. Judgment call, set once per lead (blue), based on what that specific company's environment plausibly needs — distinct from Likely Service Need (column I), the free-text one-line reasoning.
• Evidence (column J) — the specific, cited reason a lead was added: a job posting, a named project, a prequalification portal, a confirmed event, a news article about a relocation/expansion. This is the answer to "how do you know this company needs my services." Most of the original 22 seed leads are honestly labeled Category match only — no specific signal found; worth a discovery call to verify current need, because that's genuinely how they were sourced. A growing set have real citations — see "Concrete-evidence leads" below. Every new lead added from here forward requires a real Evidence citation, not just category match.

This is the actual "learning" mechanism: it's a live spreadsheet formula chain driven by Austin logging real Won/Lost outcomes, not something that only updates when the weekly routine runs.

Concrete-evidence leads (Direct Corporate Prospect channel)

Found via Crain's Detroit Business "Companies on the Move" coverage — the highest-signal source pattern found so far, and the model for future sourcing:

• F.H. Paschen Contractors — nearly tripling its Detroit office size (2026). Strong evidence: dated, specific, physical-office-expansion signal.
• Gardner White — relocating HQ from Warren to Bloomfield Hills (old Taubman building). Strong evidence: named household retailer, full HQ move.
• Eccalon — $71M / 800-job new Detroit HQ for a Maryland cybersecurity firm. Strong evidence, but likely too large-scale for a solo op to land directly — worth tracking to find out who's running the buildout rather than expecting to win it solo.

All three: the source articles are paywalled on crainsdetroit.com, so only the search-result snippet was verified, not the full article — read the full piece before calling to get exact addresses/timelines.

Events (in-person networking)

New Events tab + live Google Calendar sync. Two found so far:

• IREM Michigan — 9th Annual Sporting Clays Classic, Aug 20, 2026, Bald Mountain, Lake Orion, MI. Added to Google Calendar as an all-day placeholder (exact time not published — call 734-655-0268 to confirm before attending).
• BOMA Metro Detroit — Trade Fair: confirmed date was May 1, 2026 — already passed by the time this was researched (Aug 12, 2026). Not added to calendar as upcoming. Logged as a reminder to call BOMA now for vendor membership + next year's date so it isn't missed again.

The weekly routine now also searches for new in-person events (BOMA, IREM, and similar trade/networking events relevant to property/facility managers) and pushes confirmed dates to Google Calendar directly via the Google_Calendar MCP connector, logging its search method in the Events tab's "How Found" column and in WEEKLY_LOG.md.

Competitor Intel

New Competitor Intel tab: 6 real, sourced metro Detroit incumbents (T-Tech Solutions, Wolverine Low Voltage, CTC Technologies, Detroit Field Techs, Metro Detroit Network Installers, US Cabling Pros), each with size/positioning where verifiable, their claimed strength, and Austin's specific wedge against them. Two honest notes from this pass: (1) Wolverine Low Voltage's own site says they "partner with local businesses and individuals looking for work" — worth a call as a potential subcontract channel, not just competition; (2) Metro Detroit Network Installers and US Cabling Pros could not be independently verified as distinct operating companies (no reviews/presence beyond a listing) — treat as unconfirmed until checked further, exactly the same evidence discipline applied to leads.

Personalized draft emails (Gmail)

For leads with a real Evidence citation AND a real named contact + email, a Gmail draft (never auto-sent) gets created referencing that lead's specific situation instead of a generic template. First one: F.H. Paschen — confirmed via their own press release (primary source) that they moved into a new riverfront HQ (Talon Center) March 2, 2026, tripling their Detroit office, and are actively hiring PM/Superintendent roles. Named contact: Ken Swartz (VP, Detroit office), kswartz@fhpaschen.com — draft is in Gmail Drafts, needs [Your Name]/[Your Business Name]/[Your Phone] filled in before sending. The weekly routine now does this automatically going forward for Direct Corporate Prospect and TI Electrical/GC leads with real evidence + a real contact — never for "category match only" leads, since there's nothing specific to personalize. Drafts never change a lead's Status to Contacted — only Austin actually sending it (and logging that) does.

Quick-log processor (new routine, every 2 hours)

A second scheduled routine (separate from the weekly one) turns a quick voice-dictated note into a Pipeline update, so Austin never has to open the spreadsheet after a call:

1. Right after a call, dictate a note using your phone's native voice-to-text (built into the iOS/Android Mail app) into a new email: subject LOG: <company name>, body whatever you said, unedited (e.g. "talked to ken sounds interested wants a quote by friday mark contacted follow up friday").
2. Every 2 hours, the routine searches Gmail for unread LOG: emails, fuzzy-matches the company name to a Pipeline row, and updates Status / Last Contact Date / Next Action / Next Action Date / Notes (appending, never overwriting) based on the dictated content — conservatively, defaulting to Contacted if the implied status is ambiguous rather than guessing wrong.
3. If it can't confidently match a company name, it leaves the email unread and logs the ambiguity in QUICKLOG_ISSUES.md instead of guessing or creating a duplicate row.
4. Marks processed emails read so they aren't reprocessed. Commits and pushes only when it actually found and processed something — no empty commits.

Routine ID: trig_01XYxurFfX1PXgcKJddTjb6V, cron every 2 hours, Gmail MCP connector attached.

Contact data

Phone and Website (columns L, N) are populated by research when publicly available. Contact Name and direct Email (columns K, M, yellow) are usually gatekept by property managers/MSPs specifically and rarely public — they're expected to come from Austin after the first call, not from research. The weekly routine still checks each new lead's public "Team"/"Contact" page for a named contact or general email and fills it in when it genuinely finds one, but won't fabricate or guess one.

How larger companies source these leads (researched, see Read Me tab for full writeup + sources)

• Trade association membership (BOMA, IREM) — puts you directly in a room with property/facility managers, your exact Tier 1 buyer, via vendor membership and events.
• GC prequalification portals — large GCs (Barton Malow via BuildingConnected) proactively vet and list subs before they need them, rather than waiting for calls.
• Local SEO / Google Business Profile, case studies/portfolio content, manufacturer installer directories — documented in the Read Me tab as worth doing but not tracked as Pipeline rows (one-time setup tasks or require completed jobs first).
• Paid lead-gen agencies — exist and are used by larger firms, explicitly not recommended at this stage.

Weekly refresh routine

A scheduled cloud agent runs weekly (Mondays, 6am ET) against this repo, with Google_Calendar and Gmail MCP connectors attached (routine ID trig_01R1sVb99nWWSTVnrU86qaRV). Each run:

1. Reads README.md (this file) and the Read Me tab for full context.
2. Reads the Pipeline sheet: header row 5, data from row 6. Columns: A Company, B Tier, C Channel, D Base Fit, E Channel Multiplier (live — don't touch), F Channel Confidence (live — don't touch), G Lead Score (live — don't touch), H Service Match, I Likely Service Need, J Evidence, K Contact Name, L Phone, M Email, N Website, O Last Contact Date, P Status, Q Next Action, R Next Action Date, S Notes. Also reads the Events sheet: header row 5, data from row 6 (Event, Organization, Date, Location, Status, How Found, On Calendar?, Action Needed, Notes).
3. Researches metro Detroit for new, real signals — bid postings, GC prequalification openings, MSP hiring activity, DC vendor windows, and "companies on the move" relocation/expansion news (Crain's Detroit Business is the proven pattern — search it specifically) for the Direct Corporate Prospect channel. Also searches for new in-person trade/networking events (BOMA Metro Detroit, IREM Michigan, similar). Never invents a company or event, and requires a real Evidence citation for every new lead — a category match alone is not enough to add a row.
4. Appends new Pipeline rows only (never edits existing ones) starting at the first blank row in column A, following the Base-Fit-only rule (never sets Channel Multiplier, Channel Confidence, or Lead Score — copy the formula pattern from an existing row).
5. For any new event found with a real, confirmed date: appends a row to the Events tab (including the specific search method used, in "How Found") AND creates a Google Calendar event via the Google_Calendar MCP connector on Austin's primary calendar, with the source URL and a "verify before attending" note in the description if any detail (time, cost) wasn't confirmed. Never creates a calendar event for a date that's already in the past.
6. Logs a dated entry to WEEKLY_LOG.md: new leads by channel with their Evidence in one line each; new events found and whether they were added to the calendar; Channel Performance's current Attempts/Won/Lost/Multiplier/Confidence per channel.
7. Saves, commits, pushes to main. Never touches Dashboard formulas, Channel Performance formulas, or any user-filled Pipeline/Events column.

Uses openpyxl to edit the xlsx in place. Only appends real, sourced, evidenced leads and events — if nothing credible turns up in a given week, it logs that instead of padding the list.