Your landing page’s weakest score is CTA.
How OneTriangle ranks
62Growth scoreupdated
category scores
category scoresShowHide
Score over time
CTA
52FairCopy, size, verb, cost-feel — does the primary CTA earn the click?
Problems
- missing from mobile fold
- hero slot given to 'About us'
- demo booking feels costly
- nav-sized, small footprint
Trust
52FairProof that the product is real, competent, and safe to try.
| Trust read | |
|---|---|
| OneTriangle | "Backed by Y Combinator" badge |
| "We've worked on Google DeepMind's early scientific initiatives, developed Anthropic's interpretability efforts, and com | |
| 5%+ accuracy gains | 30%+ cost savings | 2x faster dev cycles | |
| Stat bar: "400T+ Monthly Tokens · 10M+ Global Users · 80+ Providers · 500+ Models" | |
| implicit context Fireworks AI may benefit from pre-existing familiarity among its target audience. | |
| "TRUSTED BY" logo strip with Cohere, DeepMind, LG AI Research, VFS.Global, ElevenLabs |
Mobile web
62FairThe phone experience as a first-class citizen.
12- About us CTA — Primary action is a small outlined button labeled "About us →" — an informational link, not a conversion action, and it sits at the very bottom edge of the fold.
- Body paragraph — Six lines of dense KV-cache copy push the CTA to the fold's bottom edge; the value statement should compress to two or three lines on a phone.
Information hierarchy
68FairDoes the design help or inhibit what the page is trying to say?
123- About us hero CTA — The only in-content action points to a bio page, not the conversion; 'Book a demo' lives isolated in the top-right corner.
- Diverging arrows diagram — Arrows fan out toward OpenAI/Anthropic logos with no labels, so the visual explains routing worse than one caption would.
- Nav link row — Five same-weight links, including 'Privacy & Terms', sit ahead of the demo CTA and add exits before the pitch lands.
Value props
78FairTitle, subtitle, hero — do they sell the product's true value?
OneTriangle’s headline sells
Cheaper & Faster Inference
By using frontier research on KV cache transfer between models, cut your inference costs and time by 20%. Prefill runs on a small model, and decode on a large one.
Working
- Names the two outcomes buyers want
- Subhead gives a concrete 20% number
- Hero diagram shows the mechanism
Could improve
- Headline is generic comparative copy
- 20% claim buried in subhead
- No quality/accuracy reassurance