Product Performance, Growth and Continuous Learning

Is the product creating durable customer and enterprise value, and what should change next?

Product performance is not traffic, feature output or a dashboard of convenient numbers. It is the observable relationship between customer outcomes, behaviour, economics, operational quality and risk. Growth is durable only when acquisition connects to realised value and retention; otherwise it is purchased activity.

A strong measurement system is deliberately small, causal enough to guide decisions and segmented enough to expose unequal outcomes. It combines quantitative signals with customer and operational evidence.

Build a performance architecture

Start with the product's intended outcome, then create a chain:

enterprise objective → customer outcome → behaviour or service signal → measure → decision rule

Use four layers:

• Outcome: the material progress achieved by the customer and organisation.
• Experience: task success, comprehension, trust and accessibility.
• Economics: revenue quality, contribution, cost to acquire and serve, retention and cash flow.
• Health and responsibility: reliability, security, complaints, harmful outcomes, concentration and environmental burden.

The Google HEART framework—happiness, engagement, adoption, retention and task success—is valuable because it asks teams to move from goals to signals to metrics.[1] It is a menu, not a requirement to measure everything.

Define value moments and cohorts

Identify the earliest behaviour that reliably indicates the customer has experienced meaningful value. Measure time-to-value, completion and return by acquisition cohort. Cohorts reveal whether apparent aggregate growth comes from a better product or simply more recent acquisition.

Retention should be measured at a frequency consistent with the need. A quarterly product cannot be judged by daily activity. Segment by market, customer type, channel, device and relevant access needs, while protecting privacy and avoiding misleading small samples.

A disciplined growth loop

Sustainable growth follows a linked sequence:

1. attract a well-defined customer with a truthful proposition;
2. help them reach value with minimal avoidable friction;
3. deliver the promised outcome reliably;
4. create a legitimate reason to return, renew or expand;
5. earn advocacy or ecosystem distribution;
6. reinvest learning into the product and proposition.

Optimising an isolated conversion step can damage later outcomes. Every growth experiment therefore needs downstream guardrails covering retention, margin, support, complaints and inclusion.

Continuous learning

The UK Government Service Standard calls for teams to iterate and improve frequently using research, performance data and changing needs.[2] Establish a monthly learning review that examines outcomes, surprising segments, experiment results, service incidents, customer evidence and assumptions approaching expiry. Decide what to scale, repair, investigate or stop.

The review must lead to resource choices. A dashboard without a decision cadence becomes reporting overhead. Preserve negative results and retired metrics so the organisation does not repeat failed reasoning.

Asia-Pacific and Hong Kong implications

Regional aggregation can conceal product-market differences. Compare acquisition, value, retention, support and economics at appropriate market levels. In Hong Kong, bilingual journeys and cross-border customers may require separate cohorts. Avoid league tables without adjusting for product maturity, channel mix or customer composition.

APEC's work on what follows SME digital transformation is especially relevant: adoption counts alone do not establish effectiveness.[3] For products serving businesses, measure operational or commercial improvement and sustained capability, not simply account creation.

AI-native product measures

AI products require outcome and system-quality measures together. Track task success, groundedness or factual quality where relevant, override, escalation, harmful error severity, latency, unit cost and user reliance. Monitor distributions rather than averages because rare severe failures matter. Model or policy changes need evaluation before release and observation afterward.

Do not use an automated score as the sole measure of itself. Combine benchmark evaluation, expert review, user outcomes and live monitoring. Microsoft’s responsible AI reporting demonstrates the organisational value of publishing governance progress and limitations rather than treating responsibility as a one-off claim.[4]

Sources

1. Google Research, “Measuring the User Experience on a Large Scale: User-Centered Metrics for Web Applications.” https://research.google/pubs/measuring-the-user-experience-on-a-large-scale-user-centered-metrics-for-web-applications/
2. UK Government Service Manual, “Iterate and improve frequently.” https://www.gov.uk/service-manual/service-standard/point-8-iterate-and-improve-frequently
3. APEC, What Comes After SME Digital Transformation? (2023). https://www.apec.org/publications/2023/04/what-comes-after-sme-digital-transformation-measuring-effectiveness-of-public-policy-and-identifying-trends-for-the-post-digital-era-apec
4. Microsoft, Responsible AI Transparency Report. https://www.microsoft.com/en-us/corporate-responsibility/topics/responsible-ai/reports/transparency-report/