CRO vs Growth Experiments: Know Which Mode You Are In

In 2025, Kameleoon surveyed 400+ experimentation teams and found that companies with mature testing programs are 69% more likely to hit their growth targets than companies that test ad-hoc. Yet the same report showed only 33.5% of A/B tests produce a positive result. The gap between those two numbers is explained almost entirely by one thing: teams that know whether they are running a CRO experiment or a growth experiment get far more out of both.

Quick Summary

• CRO (Conversion Rate Optimization) improves an existing funnel, it assumes the destination is right and fixes the road.
• Growth experiments question the destination itself, they test whether a new channel, audience, product change, or business model unlocks faster growth.
• The two modes share tools (A/B testing, analytics, user research) but have different hypotheses, success metrics, and timelines.
• Mixing the two in a single sprint is the most common reason experimentation programs stall.
• Mature programs run both in parallel with separate owners, separate KPIs, and a clear handoff protocol.

What It Actually Is

Think of your business as a car. CRO is a tune-up: tighten the bolts, re-inflate the tyres, clean the fuel injectors. The car goes faster on the road it is already on. Growth experiments ask whether you need a boat instead of a car, or a helicopter, or a different destination entirely.

CRO mode starts from a fixed conversion goal (purchase, sign-up, demo request) and systematically removes friction between the user and that goal. Changes are typically small: headline copy, button colour, form field count, page load speed, checkout step order. The hypothesis is "users want to convert but something is stopping them." Success is measured in conversion rate lift, revenue per visitor, or cost per acquisition.

Growth experiment mode starts from a strategic question: "Is there a meaningfully better way to acquire, activate, retain, or monetise customers?" Changes can be large: a new pricing model, a new traffic channel, a new onboarding sequence, a new market segment. The hypothesis is "there is a better path we have not found yet." Success is measured in user growth rate, retention curves, payback period, or total addressable market reached.

The two modes are not ranked. Neither is senior to the other. A team that only runs CRO optimises itself into a local maximum. A team that only runs growth experiments never converts the traffic it generates.

Why It Matters

The CRO opportunity is real and measurable

The 2025 CRO benchmark from Unbounce and Invesp put the average ecommerce conversion rate at 2.9%. Companies in the top quartile convert at 5.3% or higher. That 2.4 percentage point gap, at scale, is the difference between a profitable business and a break-even one. Invesp data also shows companies with structured CRO programs report an average ROI of 223% and a 3.5x correlation with revenue growth versus companies with no CRO program.

Growth experiments have compounding returns

Booking.com grew its global accommodation market share from 14% to 18% between 2019 and 2024. A significant portion of that growth came from its culture of running thousands of simultaneous experiments, many of them growth experiments testing new verticals (flights, attractions, taxis) rather than CRO experiments on existing hotel booking flows. The two types of experiments ran in separate squads with separate roadmaps.

Confusion between the two is expensive

When a team runs a growth experiment (testing a new pricing model) and measures it with CRO metrics (immediate conversion rate), the experiment almost always "fails", because a new pricing model takes weeks to show up in conversion data and often depresses short-term conversion while training users on a new mental model. The experiment gets killed, and a potentially winning idea disappears. The reverse is also true: running a CRO test on a page that belongs to a growth experiment (where the product itself is still in flux) produces noisy, unreliable results.

How It Works

CRO Playbook

1. Pick one funnel stage. Do not test the whole funnel at once. Choose the step with the largest drop-off: landing page, product page, checkout step 2, or activation email.
2. Diagnose before you hypothesise. Use session recordings, heatmaps, and exit surveys to understand why users drop. A hypothesis without diagnostic data is a guess.
3. Write a falsifiable hypothesis. Format: "If we [change X] for [audience Y], then [metric Z] will improve by [amount] because [reason]."
4. Size the test correctly. Use a sample size calculator. Most CRO tests need 1,000+ conversions per variant to reach 95% confidence. Running for fewer produces false positives.
5. Measure the primary metric and one guardrail metric. Example: primary = checkout conversion rate; guardrail = average order value. A test that lifts conversion by 8% but drops AOV by 15% is a net loss.
6. Ship or kill within the agreed window. Do not extend a test because you do not like the result. Set the end date before you start.

Growth Experiment Playbook

1. Start with a strategic question, not a tactic. Example: "Can we acquire SMB customers at lower CAC through product-led growth instead of sales?" That question generates experiments. Tactics do not.
2. Define the leading indicator. Growth experiments take longer to reach a revenue conclusion. Pick a leading indicator you trust: activation rate, week-2 retention, net revenue retention. That is your interim success metric.
3. Set a kill threshold before you start. Example: "If week-2 retention does not reach 25% by week 6, we kill this track." Pre-commitment prevents sunk-cost extension.
4. Run the smallest version that can answer the question. If the question is "can we grow through a freemium model," do not rebuild the product. Run a fake-door test or a manual concierge pilot first.
5. Document the learning, not just the result. A growth experiment that fails is still valuable if it eliminates a strategic hypothesis. Write it up. Share it. Most teams do not do this and repeat the same failed bet a year later.
6. Hand off winners to CRO. When a growth experiment finds a new channel or model that works, the next step is CRO: optimise the new funnel you just discovered.

How the Two Modes Connect

Real Company Examples

Booking.com: parallel tracks, separate teams

Booking.com runs over 1,000 experiments simultaneously. The company organises these into two distinct tracks. The CRO track owns the existing hotel booking funnel: search result ranking, price display, review snippets, urgency signals. The growth track owns new product lines: flights, attractions, taxis, and long-term stays. Each track has its own OKRs. Results from one track do not count toward the other team's targets. This separation is why Booking.com can ship improvements to its core product weekly while simultaneously testing entirely new business lines without the two efforts contaminating each other's data.

Airbnb: growth experiments at the platform level

Airbnb's Q4 2025 results showed 16% gross booking value growth year-over-year, with 64% of bookings now coming through the app and a 20% year-over-year app growth rate. A meaningful part of this shift was driven by growth experiments at the platform level: testing whether pushing users toward the app (through app-exclusive pricing, app-only early access to new experiences, and app-first notifications) would improve retention and lifetime value. These were growth experiments, not CRO experiments, because the hypothesis was "a different platform mix produces better long-term economics" rather than "the existing web checkout has friction we can remove." Once the app-first hypothesis was validated, CRO teams went to work on onboarding flows, search ranking, and booking confirmation screens within the app.

Common Mistakes

Mistake 1: Measuring a growth experiment with CRO metrics. A new pricing model, a new channel, or a new onboarding sequence will often depress short-term conversion rate. If you kill the experiment based on week-1 CVR, you will kill ideas that compound. Use leading indicators and longer windows for growth experiments.

Mistake 2: Running CRO tests on a product that is still in flux. If the growth team is actively changing the product, CRO tests on that product produce noise, not signal. Freeze the product scope before running a CRO test, or your results will not replicate when you ship them.

Mistake 3: No pre-committed kill threshold. Without a pre-committed "we will kill this if X does not happen by date Y," teams extend failing experiments indefinitely due to sunk-cost bias. This is more common in growth experiments (which feel more strategic and therefore harder to abandon) but happens in CRO too.

Mistake 4: One team owns both modes. CRO and growth experiments have different cadences, different risk tolerances, and different success metrics. One team that owns both will almost always default to CRO (faster feedback, cleaner metrics, lower risk) and underfund growth experiments. Separate ownership, even if informal, produces better outcomes.

Mistake 5: Treating a winning CRO test as a growth strategy. A 12% lift on a checkout page is real money. It is not a growth strategy. CRO compounds incrementally; growth experiments find step-change opportunities. Teams that celebrate CRO wins as "growth" stop asking the bigger strategic questions. Keep the two frames distinct in how you talk about results internally.

Key Takeaways

• CRO fixes the road you are on; growth experiments ask if you are on the right road.
• Companies with mature experimentation programs are 69% more likely to hit growth targets (Kameleoon, 2025).
• Only 33.5% of A/B tests produce positive results, which means the quality of your hypothesis, not the quantity of your tests, determines outcomes.
• CRO tests need a primary metric and a guardrail metric; growth experiments need a leading indicator and a pre-committed kill threshold.
• Booking.com runs 1,000+ experiments simultaneously by separating CRO and growth tracks with different owners and different OKRs.
• When a growth experiment wins, the next step is CRO: optimise the new funnel you just found.