Metrics - Beyond the Engagement
A Balanced Inquiry into What We Measure and Why
Starting from where we are
Most digital platforms measure success the same way: page views, time on site, recirculation, return visits. These metrics have good reasons behind them.
Page views tell you reach. Time on site suggests attention – a scarce resource. Recirculation indicates that content is compelling enough to continue. Return visits are the foundation of loyalty and, for subscription or ad-supported models, revenue.
A newsroom that sees these numbers rise is usually doing something right. A social platform that grows time per user is generally delivering value, or users would leave. These metrics are not arbitrary. They correlate with business survival and user satisfaction – imperfectly, but meaningfully.
So why the discomfort?
Because correlation is not identity. A user can spend eighteen minutes scrolling, click nothing of substance, and close the tab feeling worse than when she opened it. The dashboard calls that success. She does not. This gap – between metric and meaning – is the real problem.
The case for keeping current metrics
Before proposing changes, we should acknowledge what existing metrics do well.
They are objective. Page views and time stamps are hard to fake. They require no user self-reporting, which is often unreliable.
They scale. You can measure millions of sessions without surveys or opt-ins.
They drive accountability. A decline in return visits is a clear signal to a product team. A rise is a clear reward.
They align with most business models. Advertising sells attention. Subscriptions sell ongoing value. Both benefit from sustained use.
For many services – search engines, reference sites, productivity tools – these metrics may be sufficient. The user who finds an answer in thirty seconds and leaves has low time-on-site but high satisfaction. That is fine. The problem appears when the metric becomes the goal, and the goal becomes maximizing the metric regardless of user experience.
Where the gap opens
The gap appears in three specific conditions.
First, when the service competes for unlimited attention. Platforms with infinite feeds, autoplay, and variable rewards can drive time-on-site beyond the point of user benefit. The user stays because the design erodes stopping cues, not because she is getting value.
Second, when the value is cognitive or emotional, not just attentional. A student deep-reading an article and a user doomscrolling both generate time-on-site. The metric cannot distinguish them. For informational services, this is a real loss.
Third, when loyalty becomes lock-in. High return visits are good. But when leaving is difficult – dark patterns, hidden delete buttons, social costs – the metric conflates captivity with choice.
These are not failures of the metrics themselves. They are failures of the contexts in which the metrics are used uncritically.
Adding nuance: five complementary metrics
Rather than replacing current metrics, we can add others that address the gap. These are not universally better. Each has costs. But together, they offer a fuller picture.
| Metric | What it measures | Trade-off |
|---|---|---|
| Intention-aligned completion rate | User signals “done” (solved query, saved content, no repeat) | Requires user action; may not suit passive consumption (e.g., entertainment) |
| Cost-per-meaningful-outcome | (User time + server cost) per successful completion | Hard to define “meaningful” across services; requires consistent goal inference |
| Learner gain score | Optional pre/post knowledge rating | Low opt-in rates; self-report bias; only relevant for informational content |
| Sustainable exchange margin | Revenue per user minus cost-to-serve, disclosed corridor | Commercially sensitive; varies by sector; no single “right” number |
| Exit agency score | Clicks or seconds to delete or cancel | Does not measure psychological barriers (e.g., social pressure to stay) |
These metrics are not replacements. They are supplements. A platform could keep its existing dashboard and publish these alongside – letting investors, users, and researchers see a more complete picture.
Four levers for closing the gap (non-punitive, incremental)
Each lever works within the current system, not against it.
Market – Advertisers already shift spending in response to brand safety concerns. Extending that logic to engagement health is a small step. Contextual advertising, privacy-safe platforms, and subscription models are growing because they work, not because of regulation.
Design – Focus modes, “done” buttons, and session limits are already features on some platforms. Making them more visible or default does not require new laws. It requires product managers to prioritize user control alongside engagement.
Collective action – Advertiser boycotts, parental pacts, and voluntary industry ratings are market mechanisms, not government mandates. They have shifted behavior before and can again.
Education – Digital literacy curricula and public health labels inform choice without removing it. Users who understand variable rewards are better equipped to manage their own behavior.
Limitation: These levers are slow and uneven. They rely on consumer and advertiser attention, which is itself scarce. But they are also reversible and low-risk.
Platform redesigns as experiments, not mandates
The following changes are not prescriptions. They are hypotheses worth testing.
| Category | Concern | Possible experiment |
|---|---|---|
| Short-form video apps | Weak stopping cues | A/B test a “done” button that ends session and shows a summary |
| Visual-first social networks | Social approval loops | A/B test hiding like counts by default for a subset of users |
| Social graph platforms | Outrage-driven engagement | Test community goal boards against standard feed |
| Microblogging services | Toxicity as engagement driver | Test requiring intent tag (agree/disagree/fact-check) on quote tweets |
| Long-form video hosts | Autoplay reducing control | Test focus mode that pauses recommendations after 30 minutes |
Each experiment has a control group. Each measures not just engagement but also user satisfaction and retention over time. The results – not ideology – would determine whether the change spreads.
What this means for advertising: trade-offs, not villains
Behavioral targeting enables small businesses to compete with large incumbents. Free ad-supported services provide access for users who cannot pay. Contextual advertising, while privacy-safe, may be less efficient for niche products.
The question is not which model is evil and which is pure. The question is what mix of models serves the most users over the long term.
Contextual advertising – growing market, proven ROI. Trade-off: less precise for rare or personalized needs.
Subscription / hybrid – recurring revenue, aligns incentives. Trade-off: excludes users with limited means.
Creator-led commerce – direct creator-to-fan payments. Trade-off: works best for established creators.
High-trust niche platforms – premium ad rates, lower harassment. Trade-off: smaller scale.
Advertisers can allocate spend across these models based on their own goals. No single model is correct for all.
Practical steps (as options, not commands)
Platform product teams – Consider publishing the five complementary metrics alongside your standard dashboard. Pilot one focus mode or completion signal as an A/B test. Audit exit agency as a user experience metric.
Advertisers and agencies – Experiment with shifting a small percentage of digital budget to contextual or privacy-safe networks. Compare not just ROI but also brand sentiment.
Investors and analysts – Ask portfolio companies about their margin corridors and completion rates. Treat learner gain as one signal among many.
Citizens and users – Use focus modes when available. Join community pacts if they align with your values. Support platforms whose metrics you trust.
Policymakers – Consider transparency mandates (e.g., requiring disclosure of exit agency scores). Fund independent research on healthy metrics. Explore voluntary labeling programs.
Conclusion: a conversation, not a condemnation
We are not arguing that current metrics are wrong or that platforms are exploitative. We are arguing that a narrow set of metrics has become a default, and defaults benefit from periodic reexamination.
Every proposed metric has limitations. Every redesign involves trade-offs. Every business model has costs and benefits. But asking better questions – what are we actually measuring? what does it leave out? – is not an attack on the current system. It is how the system improves.
This paper is an invitation to that conversation. Not because the current system has failed, but because it can be better.
Paper by Andrew Kingdom 2026 - CC-BY