Pricing Experimentation Recipe: Survey Research to Validated Price Point

Purpose

This recipe produces a validated, revenue-maximizing price point — backed by Van Westendorp range analysis (acceptable price boundaries), Gabor-Granger demand curve modeling (revenue-optimal point), and optional live A/B experiment data (real conversion validation) — within 4-13 weeks at $0-$5,000. It outputs a price sensitivity report with intersection analysis, a segmented demand curve with revenue projections, and experiment results that feed directly into pricing page updates, sales enablement, and financial modeling. A 1% improvement in pricing yields an 11.1% increase in operating profit — making this the highest-leverage experiment a startup can run. [src2]

Prerequisites

• Product-market fit confirmed — product has users or qualified prospects → Idea Validation Recipe
• Pricing model decided (seat, usage, tiered, flat-rate) → Pricing Model Selection
• Current price exists — even if set by gut feeling or competitor matching; needed as baseline
• Target respondent list — min 200 reachable prospects/customers for surveys (email, LinkedIn, community)
• Survey tool account — Google Forms (free) or Conjointly (free Basic tier)
• Spreadsheet — Google Sheets for cumulative curve plotting and revenue modeling
• Analytics in place — PostHog or GA4 for conversion tracking (needed only if running live experiments)

Constraints

• Never show different prices to the same audience simultaneously — customers who discover price discrimination churn at 2-3x normal rate. Amazon tested variable pricing in 2000 and faced backlash severe enough for a CEO apology. [src7]
• Min samples: Van Westendorp 100+, Gabor-Granger 150+, live A/B 200+ conversions/variant for 95% confidence at 80% power. [src1] [src4]
• Gabor-Granger overstates purchase intent by 10-50% — apply Newton-Miller-Smith discount factors before projecting revenue. [src8]
• Existing paying customers must be excluded from price-increase experiments unless grandfathering plan is documented. [src2]
• Run each experiment for min 14 days (2 full business-week cycles) to capture weekday/weekend variation. [src2]
• Maximum quarterly experiment cadence — more frequent changes confuse sales teams and erode customer trust. [src2]

Tool Selection Decision

Which path?
├── Under 100 prospects AND budget = $0
│   └── PATH A: Qualitative — 5-10 pricing interviews + Google Forms Van Westendorp
├── 100-500 prospects AND budget = $0-$500
│   └── PATH B: DIY Survey — Google Forms + manual VW + Gabor-Granger
├── 100-2,000 prospects AND budget = $500-$2,000
│   └── PATH C: Automated — Conjointly analysis + optional cohort test
└── 2,000+ customers AND budget > $2,000
    └── PATH D: Full Stack — Survey research + live A/B + statistical validation

Execution Flow

Step 1: Document Current Pricing and Formulate Hypotheses

Duration: 2-4 hours · Tool: Google Sheets

Record current price, how it was set, current conversion rate, and ARPU. Formulate 2-3 hypotheses with specific numeric predictions. Define 5-8 test price points at 15-30% increments spanning 0.5x-3x current price. [src3]

Hypothesis template:
"We believe changing from $[current] to $[test] will [increase/decrease]
[metric] by [X]% because [evidence/reasoning]."

Price points (example, current = $29):
$19, $25, $29, $39, $49, $59, $79 (7 points, ~25% increments)

Verify: 2+ hypotheses documented; 5-8 price points defined · If failed: Analyze 5 competitor pricing pages and talk to 3 customers about value perception

Step 2: Run Van Westendorp Price Sensitivity Meter (Range Finding)

Duration: 1-2 weeks · Tool: Google Forms or Conjointly (free tier)

Design survey with clear product description + four Van Westendorp questions: (1) too cheap to trust quality, (2) bargain/great deal, (3) expensive but would consider, (4) too expensive to consider. Distribute to 150-300 prospects. Plot cumulative frequency curves and find four intersection points: PMC, IPP, OPP, PME. The acceptable price range spans PMC to PME. [src1] [src5]

Van Westendorp intersection points:
PMC (Marginal Cheapness) = "Too Cheap" ∩ "Expensive" → price floor
IPP (Indifference)       = "Bargain" ∩ "Expensive"   → median acceptable
OPP (Optimal Price)      = "Too Cheap" ∩ "Too Expensive" → fewest reject
PME (Marginal Expensive) = "Bargain" ∩ "Too Expensive" → price ceiling

Example: PMC=$22 | IPP=$35 | OPP=$31 | PME=$58
Acceptable range: $22-$58

Verify: 100+ responses; all 4 curves plotted; PMC-to-PME range identified · If failed: Add $10 incentive; extend collection 1 week; if curves never cross, sample too small [src5]

Step 3: Run Gabor-Granger Purchase Probability Study (Revenue Optimization)

Duration: 1-2 weeks (can run parallel with VW using separate respondents) · Tool: Google Forms (branching) or Conjointly

Present 5-8 price points sequentially using branching logic: start at middle price, branch up on "yes," branch down on "no." Each respondent sees 3-5 questions. Apply Newton-Miller-Smith discount factors: 75% for “definitely buy,” 25% for “probably buy,” 0% for others. Plot adjusted revenue curve (price × adjusted probability × market size). Peak = revenue-maximizing price. [src6] [src3]

Example (1,000 addressable, NMS-adjusted):
| Price  | Raw WTP | Adjusted | Est. Customers | Revenue   |
|--------|---------|----------|----------------|-----------|
| $19/mo | 82%     | 48%      | 480            | $9,120    |
| $29/mo | 63%     | 36%      | 360            | $10,440 ← peak
| $39/mo | 44%     | 24%      | 240            | $9,360    |
| $49/mo | 28%     | 14%      | 140            | $6,860    |

Without NMS adjustment, peak appears at $39 — overstating by 34%

Verify: 150+ responses; demand curve downward-sloping; revenue curve has clear peak within VW range · If failed: If everyone says yes, extend range upward 50-100%; if steep dropoff, revisit value proposition [src6]

Step 4: Synthesize Survey Findings and Select Test Prices

Duration: 1-2 days · Tool: Google Sheets

Overlay Van Westendorp acceptable range with Gabor-Granger revenue peak. Segment results by customer type (company size, role, industry) to identify tiering opportunities. Compare against competitor pricing. Select 2-3 test prices for live experiment if proceeding to Path C/D. [src4]

Verify: GG peak within VW range; 2-3 test prices selected with rationale · If failed: If VW and GG contradict, prioritize GG (models revenue directly); investigate whether segment mixing is masking two buyer groups [src4]

Step 5: Design and Execute Controlled Live Experiment (Path C/D Only)

Duration: 1 week design + 4-6 weeks run · Tool: Stripe, PostHog, Google Sheets

Choose safe experiment structure: new-customer-only test (safest), geographic split, cohort by week, or feature-gated test. Create separate Stripe Price objects per variant. Track conversion rate, ARPU, revenue per visitor, churn. Min 200 conversions per variant. Chi-square for conversion; t-test for RPV. Target p < 0.05. Do not peek and stop early. [src7] [src2]

Safe experiment structures:
├── New-customer-only test (SAFEST)
│   └── New signups get new price; existing keep old
├── Geographic test
│   └── Different prices in different markets
├── Cohort test
│   └── Alternate by signup week
└── Feature-gated test (SAFEST alternative)
    └── Same base price, different bundles

Verify: 200+ conversions/variant; 4+ weeks; no audience contamination · If failed: If <100 conversions after 4 weeks, traffic too low — rely on survey results [src2]

Step 6: Analyze Results and Implement Winning Price

Duration: 3-5 days analysis + 1 week implementation · Tool: Sheets, billing system, all channels

Use revenue per visitor as primary metric — captures both price and conversion impact. A 20% price increase with 10% conversion drop is still +8% RPV. Update all channels simultaneously. Grandfather existing customers 6-12 months. Brief sales/CS before pricing page goes live. Monitor 30-60 days post-launch. [src7] [src2]

Verify: New price live across all channels; sales team briefed; existing customers grandfathered; 30-day monitoring active · If failed: If churn spikes >5% above baseline, roll back for existing; keep new price for new signups

Output Schema

{
  "output_type": "pricing_experiment_report",
  "format": "document + spreadsheet",
  "columns": [
    {"name": "vw_pmc", "type": "number", "description": "VW Point of Marginal Cheapness — price floor"},
    {"name": "vw_ipp", "type": "number", "description": "VW Indifference Price Point"},
    {"name": "vw_opp", "type": "number", "description": "VW Optimal Price Point"},
    {"name": "vw_pme", "type": "number", "description": "VW Point of Marginal Expensiveness — ceiling"},
    {"name": "gg_revenue_peak", "type": "number", "description": "GG revenue-maximizing price (NMS-adjusted)"},
    {"name": "gg_demand_curve", "type": "array", "description": "Adjusted purchase probability per price"},
    {"name": "recommended_price", "type": "number", "description": "Final recommended price"},
    {"name": "experiment_winner", "type": "string", "description": "Winning A/B variant"},
    {"name": "revenue_per_visitor", "type": "number", "description": "RPV for winning variant"},
    {"name": "segment_prices", "type": "object", "description": "Per-segment prices if tiered"},
    {"name": "confidence_level", "type": "string", "description": "p-value for A/B or sample size for survey"}
  ]
}

Quality Benchmarks

If below minimum: Re-distribute survey with $10-$25 gift card incentives; extend collection 1-2 weeks. If live test lacks volume, rely on survey results with 30-day monitoring plan. Directional data from 100 responses is better than no data. [src5]

Error Handling

Cost Breakdown

Anti-Patterns

Wrong: Showing different prices to the same audience simultaneously

Customers who compare notes feel cheated. Amazon tested variable pricing in 2000 and faced backlash so severe the CEO personally apologized. “It introduces an element of unfairness to buyers” that causes direct, permanent harm to brand reputation. [src7]

Correct: Test on isolated audiences with zero overlap

New-customer-only tests are safest. Geographic splits work if markets are isolated. Feature-gated tests (different bundles at same base price) avoid the fairness issue entirely. [src2]

Wrong: Using Van Westendorp alone to set the final price

VW identifies a range but not where revenue is maximized within it. In one case, VW suggested $5.75-$7.67 while Gabor-Granger found $13.89 as revenue-maximizing for the same product — nearly 2x the upper boundary. [src4]

Correct: Combine VW (range) with Gabor-Granger (revenue peak)

Run both surveys. Companies using multiple pricing research methods achieve ~30% higher revenue growth than those relying on a single method. [src4]

Wrong: Taking stated willingness to pay at face value

Gabor-Granger and direct WTP questions overstate real purchase intent by 10-50%. If 60% of respondents say “probably buy” at $49, actual conversion will be 15-30%. Revenue models on raw data lead to painful misses. [src8]

Correct: Apply Newton-Miller-Smith discount factors, then validate live

Discount: 75% for “definitely,” 25% for “probably,” 0% for others. Then validate with even a small live experiment before full rollout. Real behavior is the only reliable pricing signal. [src8]

When This Matters

Use when a startup or SaaS company needs to actually execute pricing research — design the surveys, collect the data, plot the curves, calculate the revenue-maximizing point, run the experiment, and implement the validated price. Not a strategic overview, but step-by-step execution with tools, sample sizes, and statistical requirements. Requires an existing product with users/prospects. Most urgent before fundraising (investors scrutinize pricing power), when unit economics are failing, or when preparing to launch tiered pricing.