Course contentsShow
← Data Science
Lesson 1518 of 2,14537. A/B TestingFree lesson

The Relationship Between Surrogate and Business Metrics

Explore how surrogate metrics must correlate with business outcomes and the risks of weak or broken relationships.

The Relationship Between Surrogate and Business Metrics

What you'll learn: How surrogate metrics must reliably connect to business outcomes, and what happens when that link weakens or breaks.

The Critical Connection

In A/B testing, you often use surrogate metrics as proxies for business metrics that are hard to measure directly. But here's the crucial part: a surrogate metric is only valuable if it has a strong, stable relationship with the business outcome you actually care about.

Think of it like using body temperature to diagnose illness. Temperature correlates with infection, but it's not perfect—you can have a fever without infection (false positive) or infection without fever (false negative). If that correlation weakened significantly, temperature would become a useless diagnostic tool.

Why the Relationship Matters

When you optimize for a surrogate metric in an A/B test, you're making an implicit bet: improving this surrogate will improve the business metric. This requires:

  1. Strong correlation: Changes in the surrogate should consistently predict changes in the business metric
  2. Stability over time: The relationship shouldn't suddenly shift
  3. Directional alignment: When the surrogate goes up, the business metric should too (not down!)

The Risks of Weak or Broken Links

A weak surrogate-to-business relationship creates serious problems:

  • False positives: You "win" the test on the surrogate but see no revenue impact
  • Optimization traps: You improve the surrogate at the expense of the business metric (e.g., increasing clicks but decreasing purchases)
  • Wasted resources: You run experiments that don't move the needle on what matters

Example: Page views might seem like a good surrogate for engagement, but if users click around frantically because they can't find what they need, high page views could actually indicate poor user experience and lower conversion rates.

Key Takeaway: A surrogate metric is only as good as its correlation with business outcomes—regularly validate this relationship, or risk optimizing in the wrong direction.