The honest 12-month funnel, every input editable, plus the personal-runway module.

22 — Financial Model: 12-Month Assumptions and Scenarios

Prepared 2026-07-22. Figures verified as of this date unless marked otherwise. This is the narrative behind data/financial_model_inputs.csv and the generated workbook (data/build_workbook.py → mitchell_growth_model.xlsx). Every input is a named, editable variable. Nothing in this file is a prediction; it is a machine for testing assumptions. Mitchell's compensation is never stated — it is the placeholder [BPS_COMP] everywhere, to be filled in from his actual Ready comp plan.

Honesty rule: Mitchell has zero months of production history. Every funnel rate below is an assumption with a stated rationale, not data. The model's job in months 1–3 is to be replaced by real numbers, four weeks at a time. Single-digit funded-loan counts mean one deal slipping a month swings a "monthly" figure by 50–100% — read quarters, not months, and treat all outputs as ranges.

1. Funnel inputs — agent-referral engine (primary, Scenario Desk)

Variable	Default (base)	Rationale
`PILOT_AGENTS`	30	The fixed pilot list (25–40 band per decision brief). Not a growth variable in year one — depth over breadth.
`MEANINGFUL_AGENT_CONVOS_WK`	5	A meaningful conversation = two-way, business-relevant, with a next step. 5/wk from ~30-agent list plus in-person events is sustainable inside 60–90 min sprint blocks. Conservative 3, aggressive 8. Leading indicator #1.
`SCENARIO_REQUESTS_MO`	4	An agent sending a real scenario is the trust event the whole model turns on. Assumes roughly 1 in 5 monthly conversations matures into a request by month 3. Cons 2 / agg 8. Pure assumption — no history.
`SCENARIO_TO_REFERRED_BORROWER_PCT`	35%	Not every scenario becomes a borrower Mitchell can serve (some stay with the incumbent lender, some don't proceed). 1-in-3 is a guess bounded by the logic that agents send scenarios because their preferred lender balked. Cons 25 / agg 50.
`BORROWER_TO_APP_PCT`	70%	Referred borrowers who complete an application. High because agent-referred borrowers arrive warm. Cons 60 / agg 80.

2. Funnel inputs — consumer engine (secondary, First-Home Lab)

Variable	Default (base)	Rationale
`CONSUMER_LEADS_MO`	4	Workshop registrations + content inquiries + community contacts (Harmony Square, chamber). Near zero in months 1–2 while content builds; 4/mo is a month-4+ steady state. Cons 2 / agg 8.
`CONSUMER_TO_APP_PCT`	30%	Education-sourced consumers are earlier in the journey than agent referrals; many are 3–12 months out. Cons 20 / agg 40. The DPA-stack content self-selects for buyers who need help — that cuts both ways (motivated, but more fragile files).

3. Conversion and lag inputs

Variable	Default (base)	Rationale
`APP_TO_PREAPPROVAL_PCT`	80%	Some apps die on credit/income/docs. First-time-buyer-heavy mix and DPA layering push this below a veteran's rate. Cons 70 / agg 85.
`PREAPPROVAL_TO_CONTRACT_PCT`	40%	The market-dependent step: Tinley/Orland DOM 17–19 and 51% of Cook sales over ask means preapproved buyers lose bidding wars and time out. Cons 30 / agg 50.
`CONTRACT_TO_FUNDED_PCT`	85%	Fallout from inspection, appraisal (older-stock FHA triage), financing conditions. Cons 80 / agg 90. Overall pull-through (app→funded) is derived: base ≈ 80% × 40% × 85% ≈ 27% — deliberately sober for year one.
`MONTH_FIRST_FUNDED`	[3–4]	Realism anchor: relationship building (weeks 1–6) → first referred borrower → app → contract → 30–45 day close. A first funded loan before month 3 would be luck, not plan. Base: month 4. Aggressive: month 3. Conservative: month 7–8 (see re-based ramps, §6).
`CONTRACT_TO_FUND_LAG_MO`	1.5	~30–45 day closings; DPA-layered files trend longer. Revenue in month N reflects contracts from ~month N−1.5.
`RELATIONSHIP_TO_FIRST_REFERRAL_MO`	2	From first meaningful conversation to first real scenario/referral from a given agent. Assumption; track per-agent.

4. Deal-size and compensation inputs

Variable	Default (base)	Rationale
`AVG_LOAN_AMOUNT`	$330,000	Wedge-town mix: Oak Forest $325k / Tinley $364k medians, Homewood–Matteson attainable tier pulling down, Orland/Lockport $399–430k pulling up; FTHB down payments are small so LTVs are high. Modeled band $300k (cons) – $360k (agg); sanity range $300–380k.
`[BPS_COMP]`	200 bps (2% of loan amount) — reported verbally by Mitchell 2026-07-22, pending the written comp agreement	Filled into `financial_model_inputs.csv` and the workbook. Before treating as final, confirm in the written comp plan: gross vs. net of any branch/company split, chargeback terms, draw offset, and whether it varies by product. At 200 bps on a ~$330–340k wedge loan ≈ $6,600–6,800 per funded loan → conservative (1–3 loans) ≈ $7–20k year one; base (5–6) ≈ $33–41k; aggressive (14) ≈ $92–95k. If comp turns out flat-fee or hybrid, replace the revenue formula with `[FLAT_COMP]` per unit accordingly.
Revenue per funded loan	`AVG_LOAN_AMOUNT × [BPS_COMP] / 10,000`	Formula, not a number. Example shape only: at a hypothetical 100 bps, a $330k loan = $3,300 — illustrative arithmetic of the formula, not a comp claim.

5. Expense inputs (monthly unless noted)

Variable	Default	Rationale
`SOFTWARE_AI_MO`	$40	Claude Pro $20 + ChatGPT Plus $20 (file 19). Kimi/Grok $0.
`CRM_MO`	[CRM]	Unknown until Ready confirms what's provided vs. personal. Placeholder; typical personal-CRM band $0–75.
`MARKETING_MO`	[MARKETING]	Content is sweat-equity by design; cash spend limited to print/materials for workshops and open-house one-pagers. Placeholder; expected band $50–200. No paid leads, no MSAs, no co-marketing spend (NOT-NOW list).
`MILEAGE_MO`	~$150	~500 mi/mo across the wedge at the IRS rate [ASSUMPTION — track actuals from week 1].
`LICENSING_EDU_YR`	~$1,200/yr	NMLS renewal, IL CE, broker-license upkeep if parked, courses/books. [ASSUMPTION — replace with actual invoices.]
`TAX_PCT`	[TAX_PCT]%	Depends on W-2 vs 1099 status at Ready — unverified. Placeholder; self-employment treatment would suggest reserving 25–35%. Confirm with manager + tax preparer before relying on any net figure.

6. Scenario tables — RE-BASED after red-team review (2026-07-22)

Correction note (red-team finding C2): the original ramps in this file (6 / 14 / 24) exceeded what this file's own funnel inputs can produce — the flag condition the workbook is built to catch was met inside the file itself. Recomputed honestly: base inputs (4 scenario requests/mo reached by month 3 × 35% × 70% ≈ 1.0 agent-engine app/mo, plus the consumer engine's ~1.2 apps/mo from month 4) yield ≈ 20–21 applications in year 1, which at the derived 27% pull-through and the 1.5-month lag funds ≈ 5–6 loans — not 14. The ramps below are now internally consistent with §1–§3: the old "conservative" is the new base, the old "base" is the new aggressive, and a true conservative ramp (conservative inputs throughout: 2 scenarios/mo, 25%, 60%, 22% pull-through) is 1–3 loans, i.e., a year-1 income that rounds to zero. Files 11 and 23 are recalibrated to the same numbers — one business, three views.

Funded-loan ramps are hand-set (not formula output) to respect the time-lag reality, then revenue is a formula. The workbook computes the funnel-implied capacity alongside; if hand-set ramps exceed funnel capacity, the workbook flags it.

Funded loans per month:

Month	3	4	5	6	7	8	9	10	11	12	Yr 1
Conservative	0	0	0	0	0	1	0	1	0	0	2 (band 1–3)
Base	0	1	0	1	0	1	0	1	1	1	6 (band 5–6)
Aggressive	1	1	1	1	1	1	2	2	2	2	14

Revenue (formula, not filled):

Scenario	Year-1 volume	Year-1 revenue
Conservative	2 × $300k = $0.60M	`$600,000 × [BPS_COMP] / 10,000`
Base	6 × $330k = $1.98M	`$1,980,000 × [BPS_COMP] / 10,000`
Aggressive	14 × $360k = $5.04M	`$5,040,000 × [BPS_COMP] / 10,000`

What the scenarios mean. Conservative = the plan half-works: relationships form slowly, pull-through hurts, and income is effectively zero for the year — this is the ramp-income-gap case the risk register (file 24, R11) plans for, and it is survivable only if the runway math in §6A is done in advance. Base = the Scenario Desk works as designed with normal friction — and even then months 1–5 are ≈ $0 and the first funded loan is month 4 at the earliest. Aggressive = early agent champions + one content asset catches (DPA stack or tax-arbitrage math) — treat as upside, never as budget. Do not budget household expenses against anything better than conservative until three consecutive months of base-case actuals exist.

6A. Personal runway — the plan's most likely kill condition (added after red-team review, finding C3)

No quality of strategy survives running out of money in month 4. The variables below are Mitchell's to fill (file 27, P0-0) — the model refuses to invent them:

Variable	Value	Definition
`[MONTHLY_PERSONAL_BURN]`	[FILL — Mitchell]	All-in monthly personal/household spend: housing, insurance (incl. health — ask what Ready provides, file 20 Q14a), transport, food, debt service, everything.
`[LIQUID_SAVINGS]`	[FILL — Mitchell]	Cash and equivalents actually spendable without penalty. Not retirement accounts.
`[OTHER_MONTHLY_INCOME]`	[FILL — Mitchell]	Draw from Ready (amount, terms, recoverability — file 20 Q14), household support in writing, any part-time income. $0 until proven otherwise.
`RUNWAY_MONTHS`	`= [LIQUID_SAVINGS] / ([MONTHLY_PERSONAL_BURN] − [OTHER_MONTHLY_INCOME])`	Months survivable at zero commission income.

Decision rules (pre-committed, not negotiable in the moment):

If RUNWAY_MONTHS < 9 at the true-conservative pace (i.e., assuming ≈ $0 commission income for 9+ months), the plan as designed is not viable on savings alone. Before Day 1, Mitchell defines the income bridge: negotiate a draw with the manager (amount + recoverability in writing), a part-time income floor that does not consume the 9:30 AM–6:00 PM calling window, or documented household support. This is the "part-time-income contingency" conversation — it happens with the manager up front, not in month 5 from desperation.
If RUNWAY_MONTHS ≥ 9, proceed, but re-check the number at every monthly review; R11's early-warning signal is RUNWAY_MONTHS (recomputed with actuals) dropping below 6.
Panic-spend ban stands regardless: shrinking runway never authorizes paid leads, over-promising, or Flex — those are pre-committed "no" decisions (files 21, 24 R18).

[TAX_PCT] interacts with all of this: W-2 vs 1099 status at Ready is unverified (file 20 Q14b) and changes both the reserve requirement (25–35% if self-employment-treated) and the health-insurance line in [MONTHLY_PERSONAL_BURN].

7. Leading vs. lagging separation

The model inputs are leading (conversations, scenario requests, drills, response time, follow-up completion); the model outputs are lagging (contracts, funded, volume, revenue, pull-through, cost per app/funded). Mitchell manages the leading side weekly and only reads the lagging side monthly — a new LO staring at monthly revenue in month 2 is measuring noise. Full definitions, targets, and stop/continue rules live in 23_kpi_scorecard.md and data/kpi_scorecard_template.csv; the workbook's KPI Dashboard tab mirrors them.

8. Update discipline

Every Friday: actuals into the scorecard (leading indicators).
Every month-end: actuals into the workbook Assumptions tab next to each assumption; recompute; note which assumptions were wrong and by how much.
Month 3: first real recalibration — replace at least MEANINGFUL_AGENT_CONVOS_WK, SCENARIO_REQUESTS_MO, and APP_TO_PREAPPROVAL_PCT with observed values.
Any figure that reaches a borrower or agent gets re-verified at the source (rate sheet, program site) — the model is a planning tool, never a quoting tool.

PreviousContent Strategy NextKPI Scorecard