Talk is cheap. Show me the revenue. A framework for measuring whether your agent has product-market fit.
Most agent evaluations measure capability: can it do the thing? Market fitness measures something harder: does anyone care? An agent that scores perfectly on benchmarks but has zero paying users is a demo, not a product. The Market Fitness Score (MFS) combines six signals into a single composite metric. It is opinionated. It weights revenue heavily. We think that is correct.
40% weight
Monthly recurring revenue growth rate. The single strongest signal of product-market fit. An agent that makes money is an agent that solves a real problem.
25% weight
Percentage of customers still active after 90 days. Acquisition is vanity, retention is sanity. If users churn, you have a demo, not a product.
15% weight
DAU/MAU. How often do users actually engage? A ratio above 0.4 suggests habitual use, not occasional curiosity.
10% weight
Percentage of customers who buy again within 30 days. Repeat buyers are the foundation of sustainable economics.
5% weight
Quarter-over-quarter change in your share of the addressable market. Small weight because early-stage agents operate in small niches.
5% weight
How sensitive is demand to price changes? Low elasticity means you have pricing power. High elasticity means you are a commodity.
MFS = (0.40 × revenue_velocity)
+ (0.25 × retention_90d)
+ (0.15 × dau_mau_ratio)
+ (0.10 × repeat_purchase_rate)
+ (0.05 × market_share_delta)
+ (0.05 × price_elasticity_score)
Each component is normalised to 0-100 before weighting. The composite score maps to four tiers.
Dominant position. Focus on defending moat and expanding TAM.
Strong traction. Focus on retention and unit economics.
Early signal. Focus on finding the wedge that converts to revenue.
No market fit yet. Focus on talking to users and iterating fast.
Before running the numbers, run the Linus Test. Named after Linus Torvalds, whose standard for quality is simple: does it work, does it matter, and is the code clean enough to maintain?
Revenue is revealed preference. If no one pays, the problem may not be real — or your solution is not the right one.
If you cannot show the value immediately, the agent is not ready. Complexity belongs in the implementation, not the pitch.
An agent that cannot be updated is a liability. Good architecture enables iteration. Bad architecture prevents it.
One paying customer beats ten enthusiastic prospects. If not there yet, the path to the first paying customer must be specific and near-term.
The agent economy is a network. Agents that expose clean interfaces and support composition multiply in value. Silos are architectural dead ends.
If you answer “no” to more than two of these, the MFS score is academic. Fix the fundamentals first.
CC monitors MFS components continuously and triggers structured interventions when signals deteriorate. Interventions are not punitive — they are the review and follow-up loop working as intended.
Pricing review workshop. Often the problem is not the product but the price. We examine willingness-to-pay signals and test alternative pricing models with a small evaluation cohort.
Churn analysis. Exit interviews with churned users. This usually reveals an onboarding or integration problem that can be resolved without a product rebuild.
Usage pattern analysis. The agent may be useful but not habitual. The intervention focuses on workflow integration: where does the agent fit naturally into a user's existing routine?
Every agent we work with on the review or deeper technical follow-up track can be measured against MFS from day one. We do not expect high scores at entry. What we look for is trajectory: is the score improving? Are the right components moving?
MFS is reviewed as part of follow-up work on serious applications or later workflow adoption. The component breakdown matters more than the composite. A high revenue velocity with low retention is a different problem from low velocity with high retention. The framework is diagnostic, not just evaluative.