product-skill/ PRO/ Pricing & Packaging PRO module

PRO · Pricing & Packaging

Price is discovered from the customer, never derived from your costs — and three methods have to agree before you lock it.

The base file names the tools; this module runs them. EVC sets the ceiling, Van Westendorp finds the band, Gabor-Granger finds the revenue-max point inside it — and here they converge on one defended number. All figures illustrative; every rule of thumb directional — re-test on your own data.

The whole spine on one axis

PMC floor$120

PSM optimal~$180

G-G max $200 · list$199

intent floor~$275

PME ceiling$320

EVC ceiling$450

Revenue-max on the demand curve, inside the acceptable band, with ~2× EVC headroom for upsell.

The spine · don't skip the order

Five steps, always value-based, never cost-plus.

Price is discovered from the customer, not derived from your costs (Ramanujam, Monetizing Innovation; Campbell, ProfitWell/Paddle). Run the steps in order — each bounds the next.

1 · CeilingEVC

Value ceiling WTP can reach.

2 · BandVan Westendorp

The acceptable range.

3 · PointGabor-Granger

Revenue-max inside the band.

4 · CaptureTiers + fences

Architecture that captures it.

5 · ConfirmCohort model

Proof it's incremental.

Value, not cost

EVC is the ceiling, not the price.

Capture only ~10–25% of the quantified differentiation value — leave the rest as the customer's surplus or you kill adoption.

Band vs point

PSM bounds; Gabor-Granger picks.

PSM measures stated perception, not intent or volume — it never picks the price alone. Never ship a price on PSM alone.

Illustrative

Every number here is a worked example.

The figures triangulate to teach the method. Rules of thumb are directional / point-in-time — re-run the study on your own segment.

Steps 1–3 · find the number

Ceiling, band, revenue-max point.

Nagle & Holden · Strategy & Tactics of Pricing

1 · EVC — the value ceiling

Economic Value to Customer is the ceiling WTP can reach — it is not the price. Capture ~10–25% of the quantified differentiation value (directional) and leave the rest as surplus.

EVC = price of next-best alternative + monetized value of your differentiation − switching cost

Fill-in worksheet:

alternative $____ + differentiation $____ (hours saved × loaded rate · error cost avoided · revenue unlocked) − switching $____ = EVC $____/period

Judgment: a large EVC–WTP gap means communication is your lever — the value isn't believed — not the number. Sell the differentiation before you raise the price.

Peter van Westendorp · 1976

2 · Van Westendorp PSM — the acceptable band

Survey the target segment only (≥~100 responses, directional). Four questions about your specific product, plotted as cumulative curves; read the four intersections.

Too cheap — at what price so low you'd doubt the quality?
Cheap / a bargain — at what price is it a good deal?
Expensive — at what price does it start to feel expensive but you'd still consider it?
Too expensive — at what price is it so high you'd never consider it?

Too cheap ↘ A bargain ↘ Getting expensive ↗ Too expensive ↗

Point	Intersection of	Reads as
PMC · $120	too-cheap × expensive	Floor — below this, quality doubt costs you sales
OPP · $180	too-cheap × too-expensive	Price where resistance is balanced / minimized
IPP · $210	cheap × expensive	The "normal / expected" price; median WTP; where a market leader often sits
PME · $320	cheap × too-expensive	Ceiling — above this, too-expensive dominates

Range of Acceptable Pricing = PMC → PME ($120–$320). OPP and IPP sit inside it. Read the distribution, not the average — a bimodal too-expensive curve = two segments hiding in one survey; split and tier to each.

Say it out loud

PSM measures stated price perception, not purchase intent or volume — it bounds the band, it does not pick the price. That's Gabor-Granger's job. Never ship a price on PSM alone.

André Gabor & Clive Granger

3 · Gabor-Granger — demand & the revenue-max price

Inside the PSM band, test discrete price points: "at $X, how likely are you to buy?" (binary, or top-2-box of a Likert) across 4–6 prices. Build the demand curve = % accepting at each price.

Revenue index(P) = P × acceptance(P) → the price that maximizes it is your revenue-max price

Price P	$150	$200	$250	$300
Acceptance	70%	55%	38%	22%
Revenue index	105	110 ← max	95	66

Revenue-max ≈ $200/mo. Assumes ~zero marginal cost; for high-COGS or AI features, maximize P × accept × margin% instead (see AI pricing below).

Bias watch: respondents over-accept hypothetical prices — discount stated acceptance, and validate the top candidate with a real fake-door / checkout test before locking. Gabor-Granger gives revenue-max and elasticity; it does not test packaging or feature trade-offs — that's conjoint.

Worked example · steps 1–3 reconciled

How three methods triangulate to $199.

B2B analytics product, new "Pro" tier, target = mid-market data teams. Every number illustrative. illustrative

EVC · the ceiling

next-best tool $200/mo + differentiation $300/mo (saves an analyst ~5 hrs/mo × $60/hr loaded) − switching ~$50/mo = EVC ceiling ≈ $450/mo

Capturing 25% of the $300 differentiation ≈ $75 premium over the alternative → value-based intent floor ~$275; ceiling $450. Headroom exists; WTP, not value, will bind.

PSM · the band (n≈150)

PMC $120 · OPP $180 · IPP $210 · PME $320 → acceptable band $120–$320, optimal ~$180

The acceptable band brackets everything the two point-methods propose.

Gabor-Granger · the point

$150/$200/$250/$300 → accept 70% / 55% / 38% / 22% → revenue index 105 / 110 / 95 / 66 → revenue-max ≈ $200/mo

Reconcile → defended list price

EVC ceiling $450 (lots of room — not over-charging on value), PSM optimal $180 / ceiling $320, G-G revenue-max $200. The three agree on a ~$180–$220 zone.

Defended list price: $199/mo — at the G-G revenue peak, just above PSM OPP (resistance-minimized), far under the EVC ceiling (expansion headroom + a credible value story). Annual: $2,000/yr (round, ~16% off).

Defense in one line: "Revenue-max on the demand curve, inside the acceptable band, with 2× EVC headroom for upsell."

Steps 4 & 6–7 · capture it

Compose the tiers, then fence them.

PSM and Gabor-Granger price the whole product; conjoint prices the parts so you can build Good / Better / Best — then fences let segments self-sort.

Green & Rao (conjoint) · Louviere (MaxDiff) · Ramanujam

4 · Conjoint / MaxDiff — what goes in which tier

Method	What it does	Use it to
Choice-based conjoint (CBC)	Show bundles (features × price), force trade-off choices, estimate part-worth utility per feature. WTP for attribute = utility ÷ price-coefficient	Decide tier-movers vs table stakes; simulate share/revenue under a proposed line-up
MaxDiff (best–worst)	Cheaper, no price; ranks features by relative importance	Triage the feature list when you only need "what do they value most?"

Use MaxDiff to triage the feature list, conjoint to price the bundle. Map outputs to Ramanujam's three buckets:

Leaders

High WTP + tier-mover — lead the tier, charge for it.

Fillers

Low WTP — round out the tier, don't build the story on them.

Killers

Negative part-worth or "expected free" — never bundle a Killer with a Leader.

Ramanujam · Campbell (value metric)

6 · Tier architecture — Good / Better / Best

Inter-tier ratio ~1 : 2.5–3 : 5 (directional). With Better = $199 → Good ≈ $79 (the "1"), Best ≈ $399–$499 ($79 × 5 ≈ $395, rounded to an anchor).

Good $79/mo the "1"

Entry — charm-priced for self-serve.

7 · Price fences — let segments self-sort without arbitrage

A fence is a rule that keeps a low-WTP buyer from taking the high-WTP price. Three families:

Fence	Mechanism	Examples	Watch-out
Feature / product	capability gating	advanced analytics, SSO/SAML, API limits, SLA, audit logs	gate SSO behind premium only when moving up-market — "SSO is table stakes" is survivorship bias
Usage / metric	quantity of the value metric	seats, events, rows, storage, calls, environments	the metric must be one the buyer equates with "more value," and hard to game / under-report
Segment / identity	who or how they buy	student / nonprofit / region / PPP, channel, annual vs monthly, volume	needs verification + must be defensible; geographic fences invite grey-market arbitrage

Rule: a fence is legitimate only if the lower-WTP segment can't easily jump it and the gating tracks real value — not artificial crippling (which becomes a Killer).

Step 8 · outcome & AI pricing

Cost-per-call envelope + the margin-inversion trap.

As agents replace seats, per-seat logic breaks → price on outcomes (per resolved ticket) or consumption (per run/token). But inference is variable COGS — model P × accept × margin%, not P × accept.

The cost-per-call envelope

cost per successful outcome = (tokens/attempt × $/token + infra/attempt) × attempts per success

~86%

gross margin

margin floor ≥70%

$0.04/attempt × 1.8 attempts = $0.072 COGS per resolved ticket · price $0.50 → margin ~86%

Set a margin floor (e.g. ≥70%) as a guardrail metric — anything under it is a pricing bug, not a growth cost.

The margin-inversion trap

Pure outcome pricing inverts margin the moment model cost > price captured — a hard task that needs 6 retries, or a model-price / usage spike, can make a "win" lose money per call.

Mitigations: a per-outcome cost ceiling / circuit-breaker, a retry cap, a hybrid floor (small platform fee + usage), and re-pricing rights on model-cost change. Hold the envelope as a live guardrail, not a launch-day assumption.

Metric stack: quality (eval pass-rate) → experience (resolved-without-handoff) → unit economics (cost per successful task, margin envelope). Delight that loses money per call doesn't ship.

Step 5 & the lock · confirm it's incremental

Is the tier real revenue — and did you check your blind spots?

9 · Cannibalization / cohort-migration model

A new tier that just relocates revenue you already had is a loss dressed as a win. Decompose monthly:

Net-new revenue = (net-new customers × tier price) + (upsell × Δprice) − (downgrades × Δprice) − (cannibalized from higher tiers × Δprice) − churn delta

Cohort split: tag every buyer of the new tier as net-new / upsell / lateral / downgrade. Only net-new + upsell are expansion; lateral and downgrade are migration. Track ≥2–3 cohorts before declaring success.

Decision rule: ship / keep the tier if expansion > migration AND it doesn't degrade the guardrails.

Post-launch instrumentation

NRR, tier mix / attach rate, win rate by tier, and a sales-cycle-length guardrail — a new tier that adds a decision and slows deals can be net-negative even if mix looks good. Re-test pricing ≥ quarterly and on every packaging or model-cost change.

Pre-lock · pricing is Type-1 / high-stakes

Run the FULL blind-spot check before you lock.

1
Which method am I trusting, and what does the other say? Name the disagreement — PSM band vs G-G point vs EVC ceiling.
2
Is my survey segment the real buyer, or a convenience sample?
3
Are my benchmark ratios outlier-winner survivorship?
4
Is "outcome pricing is inevitable" a recency artifact of cheap capital / current model prices?
5
Contrarian: would a single simple price out-convert this whole architecture by removing choice friction? Name a concrete failure mode to watch.

Keep reading