--- title: "Price Testing — Pricing Growth Tactic" description: "Runs structured price tests on the pricing page, reads the results against how demand across the portfolio responds to price, and rolls the winning price into the billing catalog." canonical: "https://churnkey.co/growth/library/price-testing" category: "growth" related: - /growth/library/index.html.md - /growth/library/methodology.html.md - /tools/average-revenue-calculator.html.md last_updated: "2026-06-09" --- # Price Testing Runs structured price tests on the pricing page, reads the results against how demand across the portfolio responds to price, and rolls the winning price into the billing catalog. - **Stage:** Pricing - **Evidence grade:** strong (large, consistent correlation) - **Evidence:** 82 percent of company pricing sits suboptimally against similar-vertical peers - **Trigger type:** Scheduled - **Channels:** in-app, webhook ## What is it? Most teams set a price once and never test it again, and the data says it shows: roughly 82 percent of company pricing sits suboptimally against similar-vertical peers. This tactic makes price testing a running discipline rather than a one-off project. It proposes a small set of price variants for a plan, splits new pricing-page traffic across them, and reads the results against the portfolio’s measured price response until one variant clears the decision bar. How demand responds to price is what makes the experiment worth running. Across the portfolio, the measured price elasticity is −0.82: for every 10 percent a price moves up, demand falls by roughly 8.2 percent. Demand falls less than proportionally to price, which means a well-sized increase usually nets more revenue than the volume it costs. Most companies are underpriced and have never run the experiment that would tell them so. ## When it fires Experiments run in scheduled cycles. A cycle opens when the previous one concludes, either because a variant reaches the traffic floor and clears the decision bar or because the cycle times out without a winner. The engine then proposes the next test from the plans with the largest gap between current price and the peer benchmark. It does not fire on impulse or on a single bad revenue week. Price experiments need enough traffic to read honestly, so the engine paces them: one experiment per plan at a time, each cycle sized to the traffic the pricing page gets. ## What the evidence shows The case for testing is the gap: 82 percent of company pricing sits suboptimally against similar-vertical peers, measured in a pricing analysis spanning 628,000 subscribers. The case for testing upward is the price response: at −0.82 portfolio-wide, demand falls slower than price rises, so increases lose less in volume than they gain in revenue, within reasonable bounds. The supporting structure findings shape where variants should not land: the $50–$100 price band carries 88 percent churn across the same analysis, so the engine treats that band as a hazard rather than a destination. The evidence is graded strong: a large, consistent portfolio pattern, though each company’s own price response is what the experiment measures. ## How it runs In production, the engine renders the assigned variant on the pricing page for new visitors only. Existing subscribers never see a test price and are never repriced by an experiment. Each variant writes through to the billing provider so that a signup at a test price is a real subscription at that price, not a discount dressed up as one. When a cycle concludes, the winning price rolls into the catalog for new subscribers, with existing subscribers grandfathered at their current price. Guardrails hold the variant range within operator-set bounds, keep variants out of the dead zone, and stop any experiment the operator pauses, immediately and cleanly. ## Run it with the Churnkey MCP This tactic runs via the Churnkey MCP. Connect it to your favorite AI agent and it reads your own usage and billing data, then recommends and runs the plays most likely to move your LTV — starting with whether this one fits. **Trigger:** A scheduled experiment cycle opens once the previous cycle has concluded (reached its traffic floor and decision bar, or timed out) and a plan shows a material gap between its current price and the peer benchmark. **Surface:** In-app on the pricing page, where new visitors are assigned a price variant. Winning prices write to the billing provider via the integration, with the experiment ledger available over webhook. **Guardrails:** - Subscriber protection: existing subscribers never see a test price and are never repriced by an experiment. Winning prices apply to new subscribers only, with everyone else grandfathered. - Traffic floor: no variant is declared a winner before the cycle reaches its minimum traffic. Cycles without enough traffic end without a decision rather than with a guess. - Variant bounds: every variant stays within operator-set price bounds and out of the $50–$100 dead zone. One experiment per plan at a time. - Operator control: the operator can pause or kill any experiment immediately, and every cycle is recorded (variants, traffic, outcome) in an auditable ledger. **Evidence:** Roughly 82 percent of company pricing sits suboptimally against similar-vertical peers, measured across a pricing analysis of 628,000 subscribers. The portfolio-wide price response is −0.82: a 10 percent price increase costs roughly 8.2 percent of demand, so well-sized increases usually net positive revenue. Graded strong: a large, consistent portfolio pattern, with each company’s own price response confirmed by the experiment itself. **How to run it:** 1. Rank plans by the gap between current price and the similar-vertical peer benchmark. Pick the largest-gap plan as the next experiment candidate. 2. Define the variant set within operator-set bounds, sized using the portfolio price response as the starting estimate and excluding the $50–$100 dead zone. 3. Split new pricing-page traffic across variants, writing each test price through the billing provider so signups are real subscriptions at the assigned price. 4. Run the cycle until a variant reaches the traffic floor and clears the decision bar on revenue per visitor, or the cycle times out without a winner. 5. Roll the winning price into the catalog for new subscribers, grandfather existing subscribers, and record the full cycle in the experiment ledger. 6. Track post-rollout conversion, revenue per visitor, and early-tenure churn at the new price. Feed the company’s measured price response back into the next cycle’s variant sizing. Install the MCP server: ```bash npm install -g @churnkey/mcp ``` The full reference is in the [Churnkey MCP docs](https://docs.churnkey.co/data-integrations/mcp). ## Related - [Growth Tactics Library](/growth/library/index.html.md) — The full register of graded growth plays. - [How Tactics Are Graded](/growth/library/methodology.html.md) — The four-grade evidence ladder behind this grade. - [Average revenue calculator](/tools/average-revenue-calculator.html.md) — Related calculator.