Why this is hard to get right
A Finance Director's Friday Afternoon Problem
Mariana leads finance for a mid-market B2B SaaS company. Her CEO asked her on a Tuesday to have a pricing recommendation ready for the board by Friday. The company had been at $1,000/month for 18 months. Sales reps were discounting 15–20% to close deals. Churn was creeping up. A competitor just cut prices by $100.
Mariana had the data. She had 14 months of sales exports by price tier, a competitor pricing spreadsheet her sales team maintained, and a Typeform survey where 340 prospects indicated their willingness to pay. What she didn't have was time to build a multi-scenario model from scratch in Excel — or the confidence that a manual model would hold up to a CFO asking hard questions.
She tried the obvious first move: pasting her data into an AI assistant and asking it to "analyze our prices and tell me what we should charge." The output was a generic overview of pricing strategy concepts — Van Westendorp, conjoint analysis, price anchoring. Useful for a textbook, useless for a Thursday board prep session.
The problem wasn't the AI. It was the prompt.
Without knowing her business's unit economics, her actual price tiers, or what success looked like, the model had no choice but to stay generic. It couldn't distinguish between her SaaS business and a retail shoe company. It didn't know whether she cared more about margin or volume. It didn't know churn was her biggest risk factor.
When Mariana rebuilt her prompt with structure — defining her role as pricing analyst, specifying ACV, CAC, churn rate, gross margin, and sales cycle, listing historical unit sales by tier, naming the $800–$1,200 scenario range she needed modeled, and asking for a board-ready executive summary with a clear recommendation — the output changed completely.
The AI produced a structured table of scenarios with revenue, margin, LTV, and CAC payback at each price point. It flagged discounting risk. It called out where her assumptions were weakest. It gave her a recommendation with a rationale she could defend in the room.
She presented on Friday. The board approved a price increase to $1,100 with a 90-day ramp. Churn held. Revenue per customer climbed.
The difference between a generic AI response and a board-ready analysis was not better data or a smarter model. It was a prompt that gave the AI exactly what it needed to reason like a pricing analyst — not a search engine.
Common mistakes to avoid
Omitting Unit Economics From the Prompt
Without ACV, CAC, gross margin, and churn, the AI can only produce generic pricing commentary. It cannot model LTV or CAC payback without these inputs. Always include your actual numbers — even rough estimates produce far more useful scenario outputs than leaving the fields blank.
Asking for a Single Price Recommendation
Prompting for "the right price" forces the AI into a single-answer mode when pricing decisions are inherently comparative. Scenario-based outputs — modeling $800 through $1,200, for example — give you the trade-offs stakeholders need to make a defensible decision. Ask for a range, not a verdict.
Skipping Competitor and WTP Data
Price elasticity analysis without market anchors produces internally consistent but externally blind outputs. Reference competitor price ranges and survey willingness-to-pay bins explicitly in your prompt. The AI uses these as calibration inputs — omitting them produces recommendations your sales team will immediately challenge.
Forgetting to Define the Output Format
An unstructured prompt produces an unstructured response — paragraphs of analysis with no tables, no scenarios, and nothing a stakeholder can skim. Specify your deliverables explicitly: executive summary, scenario table, chart descriptions, and a recommendation section. This single addition cuts post-processing time dramatically.
Ignoring Risk Flags and Assumptions
Pricing analyses presented without stated assumptions get picked apart in meetings. Ask the AI to surface its assumptions and flag risks — discounting behavior, volume drop thresholds, churn sensitivity — as a named section. This transforms the output from a model into a defensible recommendation.
Not Specifying Audience and Tone
"Board-ready" and "analyst working document" require completely different depth, vocabulary, and visual emphasis. Name your audience and desired tone explicitly. A prompt that says "concise, board-ready" produces tighter prose and cleaner tables than one that says nothing about the reader.
The transformation
Analyze our prices and tell me what we should charge.
You are a pricing analyst. Create a pricing sensitivity report for our B2B SaaS analytics tool. 1) Context: ACV $12k, current plan $1,000/mo; churn 3.2%/mo; CAC $2,400; gross margin 82%; sales cycle 45 days; North America SMBs. 2) Data: Last 12 months unit sales by price tiers ($700, $900, $1,100), competitor prices ($850–$1,050), survey WTP bins. 3) Tasks: - Estimate price elasticity and revenue/margin at $800, $900, $1,000, $1,100, $1,200. - Model churn, LTV, CAC payback under each price. - Flag risks (discounting, volume drop) and assumptions. 4) Output: Executive summary, table of scenarios, 2 charts, and a recommendation with rationale. Tone: concise, board-ready.
Why this works
Role Assignment Anchors Reasoning
The After Prompt opens with "You are a pricing analyst" — this single line shifts the AI from general assistant mode into domain-specific reasoning. It activates pricing-relevant heuristics: elasticity, margin sensitivity, LTV modeling. Without a defined role, the AI defaults to surface-level commentary rather than structured analysis.
Concrete Inputs Enable Real Modeling
The After Prompt supplies ACV $12k, churn 3.2%, CAC $2,400, gross margin 82%, and historical unit sales by price tier. These aren't examples — they're the actual variables a pricing model requires. Concrete inputs prevent the AI from fabricating plausible-sounding but fictional numbers, which is the most dangerous failure mode in financial analysis.
Scenario Range Drives Comparative Output
Specifying "$800, $900, $1,000, $1,100, $1,200" as the scenario range forces the AI to produce a comparative table rather than a single-point recommendation. Comparative outputs surface trade-offs, which is exactly what decision-makers need. The range also bounds the analysis — preventing the AI from exploring irrelevant price points.
Explicit Deliverables Eliminate Ambiguity
The After Prompt names four specific outputs: executive summary, scenario table, two charts, and a recommendation with rationale. This functions as a project brief for the AI. Each deliverable has a distinct purpose — the summary for executives, the table for finance, the charts for visual communication, and the recommendation for action.
Risk and Assumption Flags Build Defensibility
The instruction to "flag risks (discounting, volume drop) and assumptions" produces outputs that hold up in stakeholder reviews. Most pricing analyses get challenged on their hidden assumptions. Surfacing them explicitly turns a model into a recommendation you can defend — a distinction that matters enormously in board-level conversations.
The framework behind the prompt
The Theory Behind Pricing Sensitivity Analysis
Pricing sensitivity analysis sits at the intersection of microeconomics, behavioral psychology, and financial modeling. Understanding the underlying frameworks helps you build prompts that produce outputs worth acting on.
Price elasticity of demand is the foundational concept: how much does quantity demanded change for a given percentage change in price? Elasticity below -1 means demand is elastic — a price increase shrinks revenue. Above -1 means demand is inelastic — you can raise prices without proportional volume loss. Most B2B SaaS products operate in the inelastic range for established customers, but the new-customer funnel is often more elastic than finance teams assume.
The Van Westendorp Price Sensitivity Meter is a survey-based framework that identifies four price thresholds: too cheap, cheap, expensive, and too expensive. It's frequently cited in pricing research because it captures perceived value boundaries — not just purchase intent. When your prompt references "survey WTP bins," you're effectively encoding a simplified Van Westendorp input.
Conjoint analysis is the gold standard for isolating attribute-level value — how much willingness to pay increases when you add a specific feature. Full conjoint requires controlled survey design and statistical tooling, but AI can approximate the logic if you provide attribute-level preference data from customer interviews.
For SaaS specifically, LTV/CAC ratio has become the dominant pricing health metric. A ratio below 3x signals that acquisition costs are too high relative to customer value — and a price increase is often the fastest lever to fix it. The After Prompt on this page asks the AI to model LTV and CAC payback under each scenario precisely because this ratio is the decision criterion most leadership teams use.
Behavioral pricing research — anchoring, decoy pricing, and price-quality signaling — reminds us that price is not just a financial variable. It communicates quality, filters prospects, and anchors negotiation. A well-structured pricing analysis prompt should ask the AI to flag these second-order effects, not just model revenue curves.
Prompt variations
You are a pricing analyst specializing in consumer e-commerce.
Analyze pricing sensitivity for our direct-to-consumer skincare line, currently priced at $48 per unit.
Business context: Average order value $72; repeat purchase rate 34% within 90 days; CAC $22 via paid social; gross margin 68%; primary channel: Shopify.
Data inputs:
- Last 9 months of unit sales at $42, $45, $48, and $52 price points
- Three competitors priced at $38, $44, and $55
- Post-purchase survey: 41% of buyers said price was "about right," 22% said "slightly expensive"
Tasks:
- Estimate demand elasticity and project revenue and margin at $44, $48, $52, and $56.
- Model repeat-purchase impact on LTV at each price point.
- Flag promotional pricing risks (margin erosion, anchor price damage).
- Identify the price point that maximizes 12-month contribution margin.
Output: One-page executive summary, scenario comparison table, and a recommended price with a 60-word rationale. Tone: clear and direct, suitable for a founder review.
You are a senior pricing strategist with expertise in enterprise SaaS packaging.
Conduct a pricing sensitivity analysis across three product tiers for our project management platform.
Current packaging:
- Starter: $25/seat/month (1–10 seats)
- Professional: $45/seat/month (11–50 seats)
- Enterprise: $65/seat/month (50+ seats, custom contract)
Business metrics: Median contract $38,400 ACV; net revenue retention 108%; CAC $8,200; sales cycle 62 days; gross margin 79%.
Data available:
- 18 months win/loss data segmented by deal size and price tier
- Competitor pricing: Tool A ($35/$55/$80), Tool B ($20/$40/custom)
- Churned accounts by tier and seat count
Tasks:
- Model revenue and NRR impact of a 10%, 15%, and 20% increase on the Professional tier only.
- Assess whether the Starter-to-Professional gap creates upgrade friction or healthy segmentation.
- Benchmark against competitor tiers and identify positioning gaps.
- Recommend a packaging adjustment with estimated net ARR impact and implementation risk.
Output: Tiered scenario table, competitive positioning matrix, and a written recommendation. Format for a quarterly business review. Tone: data-driven, executive-ready.
You are a pricing analyst advising an early-stage B2B startup.
I'm trying to set initial pricing for a legal contract automation tool targeting small law firms (2–10 attorneys). We have no historical sales data yet but have completed 12 customer discovery interviews and a 50-response pricing survey.
What we know:
- Survey WTP: 30% comfortable at $199/mo, 52% at $149/mo, 18% at $99/mo
- Closest competitors charge $129/mo and $189/mo
- Our estimated gross margin target: 75%+
- Goal: reach $10k MRR within 9 months
Tasks:
- Use survey WTP data to estimate demand curves at $99, $129, $149, $179, and $199.
- Model monthly signups needed and implied CAC budget at each price to hit $10k MRR in 9 months.
- Assess competitive positioning at each price point.
- Recommend a launch price with a 90-day test-and-adjust plan.
Output: Scenario table with demand estimates, MRR projections, and implied CAC headroom. Include a launch recommendation and one alternative if the primary scenario underperforms. Tone: practical, founder-friendly, flag assumptions clearly.
When to use this prompt
Marketing Managers
Quantify how a price change could affect volume, CAC payback, and campaign ROI before a new offer rollout.
Product Managers
Test pricing for a new tier or add-on, compare scenarios, and align cross-functional stakeholders on trade-offs.
Sales Leaders
Assess discounting impact on margin and churn, then set guardrails backed by scenario analysis.
Finance Directors
Estimate revenue, gross margin, and LTV sensitivity to price moves for quarterly planning and board updates.
Founders and CEOs
Decide on a price increase with clear risk flags, assumptions, and an executive-ready recommendation.
Pro tips
- 1
Specify segments to compare elasticity by cohort (e.g., SMB vs mid-market) to reveal different sensitivities.
- 2
Include time horizon and data windows (e.g., last 12 months) to keep assumptions consistent.
- 3
Define success metrics upfront (e.g., maximize LTV/CAC while keeping churn under 4%).
- 4
List known constraints (billing model, discount policy, price thresholds) to avoid unrealistic recommendations.
A single elasticity estimate hides enormous variation across customer segments. SMBs typically show higher price sensitivity than mid-market or enterprise buyers — and that gap matters for packaging decisions.
To model segment-specific elasticity, structure your prompt with separate data blocks per cohort:
- Cohort A (SMB, 1–25 employees): Historical unit sales by price tier, average churn, average expansion revenue
- Cohort B (Mid-market, 26–200 employees): Same fields
- Cohort C (Enterprise, 200+): Same fields plus contract length and discount frequency
Ask the AI to estimate elasticity coefficients separately for each cohort, then model revenue impact of a price change assuming cohort mix stays constant. This reveals whether a price increase that works for mid-market would cannibalize your SMB pipeline.
A follow-up instruction worth adding: "Identify which cohort is most sensitive to a 10% price increase and recommend whether differentiated pricing by segment would improve total revenue." This turns a standard sensitivity analysis into a packaging strategy brief — significantly more valuable for a leadership review.
AI-generated pricing analysis can be structurally sound but numerically inconsistent if your inputs were ambiguous. Before presenting, run through this verification checklist:
Data accuracy:
- Do the revenue figures in the scenario table match your actual unit sales baseline?
- Does the AI's elasticity estimate feel directionally consistent with what your sales team observes?
- Are the competitor prices current, or did you use figures from a spreadsheet that hasn't been updated?
Model integrity:
- Does LTV increase monotonically with price, or does the model account for churn sensitivity at higher prices? (It should.)
- Is CAC payback calculated correctly as CAC divided by gross margin per month — not just ACV?
- Are the "two charts" described in the output actually producible from the table data?
Presentation fit:
- Does the executive summary lead with the recommendation, not the methodology?
- Are all stated assumptions visible — not buried in footnotes?
- Is the tone consistent with your board's communication style?
If any of these checks fail, feed the specific inconsistency back to the AI with a correction instruction. Targeted follow-up prompts fix numerical errors faster than asking for a full regeneration.
The core structure of this prompt — role, context, data inputs, scenario range, risk flags, and output format — transfers cleanly to non-SaaS pricing problems. Here's how to adapt the key variables:
Professional services (agencies, consultancies): Replace ACV and churn with average project value, rebooking rate, and utilization. Model scenarios as hourly rate changes ($150, $175, $200/hr) with projected billable hour impact. Ask for win rate sensitivity — not volume elasticity.
Physical products / CPG: Switch to contribution margin per unit, average order value, and repeat purchase frequency. Use retailer margin requirements as a constraint. Ask the AI to model MAP (minimum advertised price) compliance risk at each price point.
Marketplaces and platforms: Replace gross margin with take rate. Model supply-side and demand-side responses separately — a price increase affects both buyer conversion and seller participation. Ask for a two-sided elasticity assessment.
Healthcare SaaS / regulated industries: Add reimbursement thresholds and procurement cycle length as constraints. Flag regulatory pricing caps if they apply. Ask the AI to note where assumptions may not hold in a procurement committee context.
When not to use this prompt
When This Prompt Is Not the Right Tool
Don't use this prompt when your pricing problem is primarily qualitative. If you're deciding whether to shift from transactional to subscription pricing, or whether to introduce a freemium tier, you need a strategic framing analysis — not a sensitivity model. Sensitivity analysis assumes a pricing structure exists and tests it at the margin.
Avoid this prompt when your data is too thin to support modeling. Fewer than 3 months of sales data across fewer than 2 price points produces elasticity estimates that look precise but are statistically meaningless. In this case, a qualitative analysis of customer interview themes or a structured WTP survey design prompt would serve you better.
This prompt is not appropriate for regulated pricing environments — healthcare reimbursement, government contracts, or utility rate-setting — where price is set by external authority rather than market demand. The output may look reasonable but will miss the actual constraints that govern your pricing decisions.
Don't expect this prompt to replace a full financial model for a fundraising round or M&A transaction. Board-level decisions with legal or fiduciary weight require auditable models built in Excel or purpose-built software — not AI narrative outputs. Use this prompt to build intuition and frame the conversation, then validate with a proper model.
Troubleshooting
The AI produces a narrative essay instead of a structured scenario table
Add an explicit formatting instruction at the end of your prompt: "Format the scenario analysis as a markdown table with columns: Price, Projected Units, Monthly Revenue, Gross Margin, LTV, CAC Payback (months)." Naming the columns removes all ambiguity. If the AI still defaults to prose, prepend the instruction with "Do not explain the table — just produce it."
Elasticity estimates feel implausibly high or low compared to sales team intuition
This usually means the AI is extrapolating from too few data points or treating all price tiers as equally reliable. Add a data quality note: "Note that the $700 tier had only 12 observations — weight this tier accordingly." Then ask the AI to state its confidence level for each elasticity estimate. This surfaces where the model is guessing versus computing.
The recommendation section hedges too much to be actionable
Add a direct instruction: "Give a single recommended price with a clear rationale in 75 words or fewer. Do not hedge with multiple options unless one is clearly a contingency." Over-hedging is the AI's default when decision criteria are ambiguous — so also check that you've specified your optimization target (maximize LTV/CAC, hold churn under 4%, etc.).
The output ignores competitor pricing data you included
Move competitor pricing out of the general context block and into a dedicated numbered section with a specific instruction: "Use competitor prices ($850–$1,050) as market anchors when assessing positioning risk at each scenario price point." When data is buried in a long context block, AI models sometimes underweight it. Making it a named task forces engagement.
LTV calculations don't account for expansion revenue or upsells
Add net revenue retention as an explicit variable: "Assume net revenue retention of 112% — incorporate expansion revenue into LTV calculations at each price tier." Without this instruction, the AI defaults to a simple churn-adjusted LTV model. Expansion-adjusted LTV can change the recommended price point significantly for products with strong upsell motion.
How to measure success
How to Evaluate the Quality of the AI Output
A strong pricing sensitivity analysis output should pass these checks before you present it:
Structural completeness:
- Does it include an executive summary, scenario table, and a clear recommendation?
- Does the table cover all requested price points with consistent metrics across rows?
- Are risk flags and assumptions called out explicitly — not implied?
Numerical coherence:
- Do revenue figures derive logically from the unit volume and price inputs you provided?
- Does LTV scale consistently with price and inversely with churn?
- Does CAC payback shrink as price increases — or does the model account for volume-dependent CAC changes?
Decision utility:
- Is the recommendation specific enough to act on — not "consider raising prices"?
- Does the output name the primary risk associated with the recommended price?
- Could a stakeholder who didn't run the analysis understand the trade-offs from the summary alone?
If the output fails more than two of these checks, identify the specific gap and issue a targeted follow-up instruction rather than regenerating from scratch.
Now try it on something of your own
Reading about the framework is one thing. Watching it sharpen your own prompt is another — takes 90 seconds, no signup.
Turn your pricing data into a board-ready scenario analysis with a defensible recommendation.
Try one of these
Frequently asked questions
At minimum, you need: current price and at least one alternative price point you've tested, basic unit economics (CAC, gross margin, churn or retention), and a target metric (maximize revenue, margin, or LTV). Competitor pricing and WTP survey data improve accuracy significantly but aren't required. The AI will flag where it's making assumptions when inputs are thin.
Yes — substitute survey willingness-to-pay data or win/loss data. Be explicit: tell the AI you're working from survey data rather than observed demand. Ask it to note the assumption risk explicitly in the output. Early-stage pricing analysis on survey data is legitimate; the AI just needs to know what kind of evidence it's working from.
Replace the flat-price scenario range with usage tiers or consumption thresholds. For example: model revenue at $0.008, $0.01, and $0.012 per API call across low, medium, and high usage cohorts. Specify your median and 90th-percentile usage volumes so the AI can project realistic revenue distributions rather than flat-rate equivalents.
Your prompt is missing specific inputs. The AI defaults to education when it lacks data to model. Add at least three of these: current price, historical sales by tier, CAC, gross margin, and a defined scenario range (e.g., $800–$1,200). With concrete numbers, the output shifts from conceptual to computational.
Specify your deliverables explicitly in the prompt: "executive summary (150 words max), scenario table with columns for price, unit volume, revenue, gross margin, and LTV, and a recommendation section with rationale." Adding word counts and column names gives the AI a precise template to follow — reducing the editing you need to do afterward.
Yes, but you need to reframe the inputs. Replace churn rate with trial conversion rate and add free-tier cost-to-serve as a variable. Ask the AI to model net revenue impact of conversion rate changes at different paid price points. This is a variant of elasticity analysis — the mechanics are the same, the inputs differ.
Trigger a new analysis when any of these change: a competitor moves price by more than 10%, your churn rate shifts by more than 0.5 points, you enter a new customer segment, or you're planning a feature expansion that changes perceived value. Annual re-runs are a reasonable baseline for stable markets.