CAPM for Agencies — Calibration Notes

Purpose

Use this if you want the implementation details behind the current Decision Cards build: the score mappings, the current calibration choices, and a few sanity-test scenarios run against the live calculator.

Run the current implementation here:

What Changed

Two calibration changes matter in the current build:

Layer 1 is now truly neutral at the midpoint. The systematic adjustment factor now maps: * 6 to 0.85 * 18 to 1.00 * 30 to 1.15

This makes the midpoint score behave like a real normal environment instead of quietly pushing the hybrid hurdle upward.

The B Corp overlay is now less aggressive by default. * the B Corp portfolio score is neutral at its midpoint * the portfolio modifier now ranges from 0.8 to 1.2 * each B Corp engagement point away from its midpoint is worth about 1.0 margin point before that modifier is applied

This keeps the overlay useful as governance without turning it into a large automatic pricing swing.

Current Formula Choices

Layer 1

The Layer 1 composite score is mapped into a systematic adjustment factor. In the current build, the range is intentionally moderate:

lowest environment: 0.85
midpoint environment: 1.00
highest environment: 1.15

That factor is then used in two ways:

in the pure approach, it sets the portfolio-wide hurdle directly
in the hybrid approach, it weights the engagement beta

Layer 2

The current hybrid implementation is:

Engagement β = Score / 21
Blended β = (Engagement Score / 21) × Layer 1 factor
Required Margin E(R) = R_f + Blended β × (R_m - R_f)

This is a midpoint-anchored calibration, not an endpoint-anchored one. A neutral Layer 2 engagement score maps to market-like β = 1.0. The low end does not collapse to β = 0, which intentionally prevents zero-risk pricing while preserving more headroom for difficult deals.

The decision thresholds are:

Go: proposed margin is at or above E(R)
Caution: proposed margin is within 3 margin points below E(R)
Stop: proposed margin is more than 3 points below E(R)

Caution Band Review

Using the current defaults (R_f = 10%, R_m = 22%), a low-risk deal at blended β = 0.50 clears at 16.0%, while a high-risk deal at blended β = 1.50 clears at 28.0%.

Case	Hurdle E(R)	Fixed 3-point band	10% proportional band	12% proportional band
Low risk (`β = 0.50`)	`16.0%`	caution from `13.0%` to `15.9%`	caution from `14.4%` to `15.9%`	caution from `14.1%` to `15.9%`
High risk (`β = 1.50`)	`28.0%`	caution from `25.0%` to `27.9%`	caution from `25.2%` to `27.9%`	caution from `24.6%` to `27.9%`

What this shows:

the current fixed 3-point band is about 18.8% of the low-risk hurdle and 10.7% of the high-risk hurdle
so the current rule is actually more forgiving on safer work and stricter on riskier work
a 10% proportional band tightens low-risk deals sharply and only slightly tightens high-risk deals
a 12% proportional band is closer to the current rule, but it starts widening caution on higher-risk work

Side by side, that means:

at -2.5 points below the hurdle, a low-risk deal stays Caution under the fixed rule but becomes Stop under both proportional alternatives tested here
at the same -2.5 points, a high-risk deal stays Caution under the fixed rule and under both proportional alternatives
at -3.0 points below the hurdle, a high-risk deal stays Caution under the fixed rule, becomes Stop under a 10% band, and stays Caution under a 12% band

Decision for now: keep the fixed 3-point caution band. It is simpler to explain, and it already applies a tighter leash to high-risk work than the proportional alternatives tested here. This should still be revisited once retrospective data exists.

B Corp Overlay

The current B Corp implementation uses:

Standard hurdle: E(R) from Layer 2
Impact-adjusted hurdle: E(R*) = E(R) + Impact Adjustment

Where:

B Corp Layer 1 midpoint is neutral
B Corp Layer 1 modifies scale from 0.8 to 1.2
B Corp Layer 2 midpoint is neutral at 12 on a 4–20 scale
each B Corp Layer 2 point away from midpoint is worth about 1.0 point before that portfolio modifier is applied

So the current automatic adjustment is:

Impact Adjustment = (B Corp L2 score - 12) × 1.0 × portfolio modifier

This is still heuristic. It is designed to make the trade-off visible and discussable, not to discover an objectively correct mission premium. It is also midpoint-anchored by design: neutral B Corp impact maps to no mission discount or harm premium, while more strongly aligned or more harmful work moves the hurdle down or up from that center.

Sanity-Test Scenarios

The scenarios below were run against the live form and checked against the displayed outputs.

Scenario 1 — Calm Small Agency, Straightforward Project

Inputs:

R_f = 10%
R_m = 22%
Layer 1 scores: all 2
Layer 2 scores: all 2
Deal price: $120,000
Estimated cost: $92,000

Expected behavior:

Layer 1 factor should be mildly favorable
the hurdle should stay below the portfolio average
the proposed margin should clear comfortably

Observed result:

Layer 1 factor: 0.93
Engagement β: 0.67
Blended β: 0.62
Required margin: 17.4%
Proposed margin: 23.3%
Gap: +5.9 points
Verdict: Go

This makes sense.

Scenario 2 — Elevated-Risk Market, Ugly Fixed-Price Build

Inputs:

R_f = 10%
R_m = 22%
Layer 1 scores: 4, 4, 3, 4, 4, 4
Layer 2 scores: 4, 5, 4, 4, 4, 4, 5
Deal price: $200,000
Estimated cost: $150,000

Observed result:

Layer 1 factor: 1.06
Engagement β: 1.43
Blended β: 1.52
Required margin: 28.2%
Proposed margin: 25.0%
Gap: -3.2 points
Verdict: Stop

This also makes sense. The deal is not absurdly underwater, but it still fails the stated threshold.

Scenario 3 — Mission-Aligned B Corp Work

Inputs:

same financial inputs as Scenario 1
B Corp Layer 1 scores: all 2
B Corp Layer 2 scores: 1, 1, 2, 2

Observed result:

Standard hurdle: 17.4%
B Corp impact adjustment: -5.4%
Impact-adjusted hurdle: 12.0%
Verdict: Mission-aligned

This is still a meaningful mission discount, but it is more defensible than the earlier, much larger automatic swing.

Scenario 4 — Harmful but Financially Attractive Work

Inputs:

R_f = 10%
R_m = 22%
Layer 1 scores: all 3
Layer 2 scores: all 3
Deal price: $210,000
Estimated cost: $150,000
B Corp Layer 1 scores: all 4
B Corp Layer 2 scores: 5, 4, 4, 4

Observed result:

Standard hurdle: 22.0%
B Corp impact adjustment: +5.5%
Impact-adjusted hurdle: 27.5%
Proposed margin: 28.6%
Verdict: Harm premium cleared financially — review mission trade-offs explicitly

This is a good example of the B Corp overlay doing what it should: making the mission cost visible without automatically forcing every bad-fit deal into a stop.

What To Watch For

One practical calibration note: agency-wide net margin benchmarks and project-level portfolio margin benchmarks are not the same thing. Industry averages may sit in the mid-teens, while a usable project-margin benchmark for pricing decisions may reasonably be higher.

The current Decision Cards build is strongest when:

the team agrees on what the score anchors mean
the same people score similar deals repeatedly
the agency compares required margin, proposed margin, and actual outcomes over time

It is weakest when:

users treat the numbers as statistically discovered truths
the portfolio baselines are guessed rather than observed
the B Corp overlay is mistaken for an official B Lab formula

Recommended Use

Treat these calibrations as defaults, not doctrine.

If you have enough historical data, you should revise them:

based on closed-won and closed-lost deals
based on margin slippage in delivery
based on how often Go, Caution, and Stop calls proved right in retrospect

That is the real point of the model: not numerical purity, but disciplined judgment that gets better over time.