Strategy & Planning

How to run a pricing analysis with AI

Strategy & Planning4 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

A pricing analysis is the structured process of figuring out whether your prices are right — for your costs, your market, and your customers' willingness to pay. Most operators run one when they're preparing to raise prices, entering a new segment, responding to a competitor's move, or building a board deck that needs a defensible pricing rationale. It's not a one-time exercise; it recurs every time market conditions shift or your unit economics drift.

Pricing feels like an AI-native problem because so much of the work is pattern recognition and synthesis: reading competitor positioning across a dozen websites, mapping price points to feature tiers, translating raw transaction data into a coherent story about who's paying what and why. That's exactly the kind of exhausting, context-heavy synthesis that large language models are genuinely good at — which is why so many operators now reach for ChatGPT or Claude before opening a spreadsheet.

General-purpose AI tools — ChatGPT, Claude, Gemini — can meaningfully accelerate the analytical and narrative parts of a pricing analysis. They're useful for drafting competitive comp tables from information you paste in, suggesting segmentation frameworks, interpreting price elasticity concepts, and structuring a pricing memo. They are not useful for the parts that require your actual data: real revenue by plan, churn rates by tier, customer acquisition costs by channel. That gap matters.

Skip the manual work → See the Starch version

Strategy & Planning4 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ChatGPTClaudeGeminiPerplexity

Step-by-step

1 Use Perplexity to survey the competitive landscape: search for your three to five closest competitors and ask it to summarize their publicly visible pricing tiers, price points, and feature differentiation. Export the results as a starting table.

2 Open Claude and paste that competitive table in. Ask it to identify the implied pricing logic — what each competitor is optimizing for (volume, enterprise ACV, usage, seats) — and where your current pricing sits relative to that spectrum.

3 Export your revenue-by-plan data from Stripe and your cost data from QuickBooks or your bank, then copy the relevant rows into ChatGPT. Ask it to calculate gross margin by tier and flag which plans are below your target margin threshold.

4 Prompt ChatGPT or Claude to build a willingness-to-pay framework for your customer segments using whatever customer interview notes, NPS data, or support tickets you can paste in. Ask it to map segments to price sensitivity and identify candidates for price increases.

5 Run a scenario prompt: give Claude your current plan distribution, average revenue per user by tier, and churn rates. Ask it to model what happens to MRR if you raise the mid-tier price by 20% and 10% of that tier churns.

6 Ask ChatGPT to draft the pricing recommendation memo: a one-page summary of current state, competitive position, recommended changes, and the financial rationale. Paste in all prior outputs as context before running this step.

7 Review the output carefully. LLMs will hallucinate specific numbers if you don't supply them — check every figure in the model against your source export before sharing anything externally.

Prompts you can copy

Here is a table of competitor pricing for project management SaaS tools [paste table]. Identify the pricing logic each competitor is using and where gaps or white space exist for a new entrant priced between $15 and $40 per seat.

I'm attaching our Stripe revenue export for Q1. Calculate average revenue per user by plan tier, gross margin assuming 18% COGS, and flag any tier where margin falls below 70%.

Here are notes from 12 customer interviews [paste notes]. Identify which customer segments show the highest price sensitivity and which show the lowest, based on language about budget, approval process, and switching cost.

Model the MRR impact of raising our Pro tier from $49 to $59/month given 340 current Pro subscribers, assuming 8% churn from the price change and 5% new subscriber uplift from the repositioning.

Write a two-page pricing strategy memo recommending we move from per-seat to usage-based pricing. Use this competitive analysis, this margin data, and this customer segment summary as inputs [paste all three].

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

No live connection to your Stripe or QuickBooks data — every run starts with a manual export, copy-paste, and cleanup of whatever CSV format your tool spits out that day.

Context window limits force you to truncate. If you have more than a few hundred transactions or a long interview corpus, you're choosing what to leave out before the analysis even starts.

The LLM has no memory of last quarter's analysis. Next time you revisit pricing, you rebuild the prompt chain, re-paste the context, and re-explain your business from scratch.

Outputs drift between sessions. The structured competitive table you carefully prompted in February looks different when you try to reproduce it in May — different format, different framing, different conclusions.

Scenario modeling breaks down quickly. Asking an LLM to hold five pricing scenarios in context while you iterate on assumptions is possible but brittle; one session timeout and you're starting over.

There's no audit trail. When your board asks how you got to the pricing recommendation, the answer is 'a conversation in ChatGPT' — which is hard to version, share, or defend.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — an agent builds and runs the apps, dashboards, and automations your work depends on, connected to your live business data. For pricing analysis, that means a persistent app that pulls real revenue and transaction data on a schedule, so the analysis reflects what's actually happening in your business, not last month's export.

Connect Stripe and Plaid once — Starch syncs your actual charges, subscriptions, and bank transactions on a schedule. Every time you revisit your pricing model, it starts from live numbers, not a stale CSV you downloaded Tuesday.

The Scenario Analysis app connects directly to your Stripe and Plaid data and lets you test pricing changes — raise the mid-tier by $10, assume 8% churn — and immediately see the impact on runway and burn rate without rebuilding a spreadsheet.

The Transaction Insights app pulls every Plaid transaction and flags cost anomalies automatically, giving you a clean view of your actual unit economics before you set prices — no manual categorization required.

Connect QuickBooks or Xero from Starch's integration catalog and ask the agent to build a margin-by-plan dashboard in plain English: 'Show me gross margin broken down by Stripe subscription plan, pulling COGS from QuickBooks.' The agent builds it and keeps it current.

Describe the competitive monitoring workflow you want — 'every Monday, pull pricing pages from these five competitor URLs and summarize any changes' — and Starch automates it through browser automation, no API needed, delivered to your inbox or Slack.

When you're ready to present the analysis, the Presentation Agent (coming soon) will build a polished deck from your Starch data and analysis summary — so the board deck is generated from the same live data the analysis ran on, not a separate export.

Get closed-beta access →

Toolkit