Quality signals beat activity counts

Activity metrics count what happened: messages sent, sessions started, clicks made. The trouble is they rise when the product works and when it fails. On a dating app, messages sent climbs when matching is great and also when it is so bad that people spray messages into the void. A quality signal asks whether the interaction succeeded instead: did the conversation receive a reply, continue past the first exchange, and avoid ending in a block soon after? Those outcomes require two people getting value, and that is exactly why they are hard to fake.

Why it matters to the business

Steering on activity lets a business feel healthy while customers quietly leave. Bain's research found 80% of companies believed they delivered a superior experience while only 8% of their customers agreed, and inflated activity dashboards are one way that delusion survives. Outcome measures also connect directly to money: research by Peter Kriss published in Harvard Business Review found customers of a transactional business who had the best past experiences went on to spend 140% more than those with the worst. Success, not volume, is what compounds into revenue.

How to use it

• For each headline activity metric, define its success version: replies earned, tasks completed, issues resolved without a repeat contact.
• Add a time-boxed failure condition, such as no block soon after the exchange or no quick repeat ticket.
• Run the gaming test: ask how a team could inflate the metric without a customer winning, then close that route.
• Ban features whose primary purpose is to move a metric rather than serve the customer.
• Validate quality signals against retention and spend at least once a year.

Where teams get it wrong

The most common failure is keeping the old incentives. Teams add quality metrics to the dashboard but still reward volume, so behavior never changes. The second failure is worse: shipping features designed to inflate the quality signal, like nagging prompts that manufacture replies. The moment you build a feature to pump a metric, the metric stops measuring anything.

Ask your team

• Which of our headline metrics would keep rising if the product got worse?
• What is our definition of a successful interaction, and who owns improving it?
• Has anything shipped this year whose main job was to move a number rather than help a customer?

Count outcomes, not actions. A metric that needs a real customer to win is a metric worth steering by.