How we’re fixing utility app bounce rates: Leaderboards, visual context, and AI metrics
Hey IndieHackers,
A few months ago, as part of a research project at ScaleDynamics, we launched SpeedPower.run - a 60-second in-browser benchmark designed to test device performance under modern web and heavy client-side AI workloads.
It received solid initial traffic, but we immediately ran into a classic utility tool problem: brutal single-use bounce rates. People would run the test, see a raw number like 341, say "cool," and never open the site again.
To transform a one-and-done benchmark into an engaging product loop, we just shipped a major three-part update:
-
Gamification: Global Top 20 Leaderboard
To drive repeat traffic and organic sharing, we launched a live Top 20 Leaderboard (https://speedpower.run/top20). By introducing a competitive element, we’re encouraging users to tweak their browser setups, test different hardware, or challenge their friends to beat the current highest score. -
UX Context: Dynamic Global Score Spectrum
Abstract scores mean nothing without a reference point. To fix the "What does 341 actually mean?" issue, we added a visual range bar on the results card. Your score is now dynamically plotted on a scale from 0 to the highest score ever recorded globally (currently sitting at 6,811). Users immediately understand where their machine sits in the hardware ecosystem without needing a cheat sheet. -
Strategic Metrics: BDP & AIR Ratings
Because modern web development is splitting into traditional JS execution vs. WebGPU/client-side AI compute, a single overall score wasn't cutting it anymore. We split the result into two dedicated ratings:
BDP (Browser Device Performance): Measures heavy DOM, JSON, and standard JS tasks (e.g., BDP 1 - Entry-level).
AIR (AI Readiness): Measures WebGPU and local AI inference capabilities (e.g., AIR 3 - High-perf AI).
This repositioned SpeedPower from a standard web benchmark to a future-proof tool specifically designed for the Web AI era.
I’d love your feedback on the product strategy:
If you build utility/calculator tools, what features have actually worked to drive repeat traffic for you?
Does breaking scores down into niche metrics (like BDP/AIR) add value for you as a maker, or do you prefer a single overall score?
Give it a spin at https://speedpower.run, see if you can break into the Top 20, and let me know your thoughts on the new UI below!
The interesting shift is moving from a benchmark into a repeat-use product.
What would convince you that the retention problem is caused by missing engagement loops, rather than the underlying use case naturally being a one-time evaluation?
Two specific signals would prove to me it’s an engagement loop problem:
Multi-run session depth: Users tweaking browser flags, disabling extensions, or toggling GPU acceleration and re-testing 3–5 times in a single session to optimize performance.
External trigger returns: Spikes in returning traffic whenever Chrome/Firefox ship major WebGPU, engine, or OS updates.
If we ship these loops and 30-day retention stays near zero, I’ll gladly accept it's an episodic utility. At that point, the play isn't forcing artificial retention, it's shifting 100% of our focus to top-of-funnel reach and viral distribution.
That’s useful context. Happy to continue the conversation privately — what’s the best email to reach you on?