Four detection layers. One URL. 60 seconds.
Every scan runs four passes in parallel: structured DOM analysis, rule-based threshold checking, Lighthouse performance measurement, and LLM qualitative interpretation. Click each stage to see exactly what it detects and how it contributes to finding confidence.
DOM Fetch
PlayHero fetches the full rendered DOM of the target URL across three viewport widths: 375px (mobile), 768px (tablet), 1440px (desktop). This is not an HTTP request — we render the page in a headless browser to capture the post-JavaScript DOM, exactly as a visitor would see it.
Every field exists for a reason.
A Rallycard is a governed decision artifact — not a notification. Every field is designed to give the human reviewer exactly what they need to make a confident decision. Hover each field to see why it's there.
Once a card is approved, the original finding is locked. Status changes (In Progress, Resolved) are appended as a log — the original card is never overwritten. This is how the Heroboard becomes institutional memory, not just a task list.
Six scored dimensions. Specific thresholds. No guesswork.
The WebUX Score is not a proprietary black-box metric. It is the weighted composite of seven pillar scores, each measured against published or well-established thresholds. Click any pillar to see the exact thresholds used and why they were chosen.
Four specialists. Defined domains. No overlap.
Each agent has a stable identity, a specific detection domain, and a confidence methodology suited to its signal type. Domain isolation is deliberate — a UX finding from UX Sentinel and a CWV finding from Page Vitals Enforcer are detected by fundamentally different methods, and conflating them would lower the credibility of both.
UX Sentinel | Drift Detector | Page Vitals Enforcer | Lumen | |
|---|---|---|---|---|
| Domain | Conversion friction & technical UX | WebUX signal drift & page anomalies | Core Web Vitals & performance | AI discoverability & agent readiness |
| Detection method | DOM analysis + LLM + rule-based | Page signal monitoring + trend analysis | Lighthouse + script enumeration | Structured data + llms.txt + content extraction |
| Confidence threshold | DOM evidence + LLM agreement | Statistical significance vs. 30-day baseline | Objective threshold breach | Deterministic signal presence |
Governance layer. Institutional memory.
The Heroboard is not a task manager. It is the permanent, searchable record of every UX decision your team has made against your monitored pages. Every card that passes through it — approved, dismissed, or deferred — creates a log entry that cannot be deleted.
Card has been surfaced by an agent and is waiting for a human decision. The agent continues monitoring while the card waits — additional evidence may update the finding.
Every decision — approve, dismiss, defer — is logged with actor identity and timestamp. Deletions are not permitted. Over time, the Heroboard becomes the only complete history of every UX decision that touched your pipeline.
Script audit and competitor mode. Both free. No account.
Script audit runs on every scan automatically — no configuration needed. Competitor observation mode activates automatically when you scan a URL you don't own, suppressing pipeline estimates that would be speculative.
Every third-party script, sized and risk-rated.
Included free on every scan. PlayHero enumerates all third-party scripts loaded on the page, identifies them by origin, sizes them by transfer weight, and risk-rates them based on their contribution to LCP and Total Blocking Time.
Scan any URL. Pipeline estimates removed automatically.
When you scan a URL you don't own, PlayHero automatically enters observation mode. The full 7-pillar analysis runs, but pipeline dollar estimates are suppressed — you can't model conversion impact for a site you don't have traffic data for, and fabricating those numbers would undermine the credibility of the findings.
Five things that are non-negotiable.
Every feature we ship is evaluated against these five commitments. They are not marketing copy — they are constraints that have shaped product decisions from day one. The ones that slow us down are the ones that matter most.
Every Rallycard shows which technical method detected the finding. DOM structure analysis is objective — a CTA below fold is measurable. LLM interpretation is probabilistic — a value proposition missing ICP clarity is an assessment. Users see the difference on every card.
Finding Confidence measures the likelihood that the finding itself is valid. It is never shown as a measure of revenue impact certainty. The pipeline impact range is always a benchmark-based estimate, always labeled as such, with methodology disclosed inline — not in a footnote.
No agent moves a card to execution. No change is made to your site. No action is taken on your behalf. Approval is a governance signal — it moves the card to In Progress and notifies your team. Your team executes. Agents track what's running.
The traffic volume, average deal value, and CVR baseline underlying every pipeline impact estimate are editable post-signup. The default assumptions are disclosed on every card. You can calibrate them to your actual site data, making every subsequent impact estimate more accurate.
Every decision — approve, dismiss, defer — is logged with actor identity and timestamp. This record cannot be deleted. Over time, the Heroboard becomes institutional memory: the only complete history of what was found, who decided what, and what changed afterward.
See the system work on your site.
One URL. 60 seconds. Your first Rallycard before you've created an account.