Skip to content
Platform · Experiments

Experiments worth running.

Visual builder with JS, TypeScript, CSS, or SCSS per variation. Anti-flicker in, events flowing, Bayesian or Frequentist out.

01 — Visual builder

Code blocks, your language.

Each variation gets its own script and style files, with a shared trigger block that decides when the experiment activates. Monaco ships TypeScript types for window.avsb.* and the options toolkit, so autocomplete just works — no @types install, no manual wiring.

Pick your language per block. Script blocks support JAVASCRIPT / TYPESCRIPT. Style blocks support CSS / SCSS. SCSS compiles in-browser — no bundler required.

  • Live preview against any URL.
  • Git-style diff between control and variant.
  • Shared helpers library across variations.
variant.ts
1// variant.ts — runs once this variation activates
2function initVariation(options) {
3  const btn = document.querySelector<HTMLButtonElement>('.checkout-cta')
4  if (!btn) return
5

6  btn.textContent = 'Claim 30% discount'
7  btn.classList.add('urgency')
8

9  // Fire a conversion with revenue attribution
10  options.track.event('purchase', { revenue: 49 })
11}
02 — Lifecycle

DRAFT → SCHEDULED → RUNNING → PAUSED → COMPLETED.

DRAFT
SCHEDULED
RUNNING
PAUSED
COMPLETED

Timezone-aware scheduling. Scheduled start and end dates. Traffic allocation 0–100%. Pause any experiment without losing exposure or event data — resume when you’re ready.

03 — Targeting

URL match + audience rules.

URL match types
  • SIMPLEGlob-style match with wildcards. Great for paths like /checkout/*.
  • EXACTFull URL match. No trailing slash surprises.
  • SUBSTRINGFire when the URL contains a fragment anywhere.
  • REGEXDrop in a regular expression when nothing else will do.
10 audience conditions

Compose 10 condition types into nested AND / OR rules. Reusable segments attach to many experiments and flags.

  • Location
  • Device
  • Browser
  • Platform (OS)
  • Language
  • Query param
  • Cookie
  • New vs returning
  • Custom attribute
  • Custom JavaScript
04 — Traffic and scheduling

Ship safely.

Anti-flicker
Zero flash.

The snippet hides the page until variations apply, then reveals in a single paint.

SPA support
Re-evaluates on route change.

pushState, replaceState, and History API events trigger re-bucketing automatically.

Consent mode
Respects your CMP.

Defer activation until consent is granted. Flush queued events on opt-in.

05 — Exclusion groups

Don’t let experiments trip over each other.

Carve traffic across concurrent experiments so no visitor sees two changes to the same surface at once. Define a group, assign experiments, and the snippet guarantees mutual exclusion at bucketing time.

06 — Statistics

Bayesian or Frequentist. You choose.

BAYESIAN

Bayesian

Low-to-mid traffic experiments. Peek whenever you want — no multiple-comparisons penalty.

  • Probability to beat control
  • 95% credible interval
  • SRM check
FREQUENTIST

Frequentist

High-traffic clients who want the textbook framework their stats team already trusts.

  • p-value
  • 95% confidence interval
  • Sample-size reached
SEQUENTIAL

Sequential

Teams that want to stop the moment significance lands without inflating false positives.

  • Always-valid p-value
  • 95% always-valid confidence sequence
  • Peek anytime, stop when it lands
07 — Results

The readout your analysts live in.

Filters, metrics, intervals, and segment breakdowns in one place — the page consultants and CRO analysts keep open all day. Pick a date range and a segment, read the lift, check the health signals, and export the table when it’s time to write it up.

Probability to beat control

Bayesian probability-to-win, or a frequentist p-value and confidence interval — whichever engine the experiment was registered with.

Observed lift + interval

Relative improvement on the primary metric with a 95% credible (Bayesian) or confidence (frequentist) interval around it.

CUPED variance reduction

Pre-exposure covariate adjustment tightens intervals and reaches significance sooner — applied automatically when it helps.

SRM health check

Sample-ratio-mismatch detection flags a broken split before you trust a number that was never valid.

Secondary metrics

Track guardrails and supporting goals alongside the primary — so a win on one metric never hides a loss on another.

Segment lift (HTE)

Break the result down by audience to find where the effect is strongest — heterogeneous treatment effects, surfaced per segment.

How every number is computed — the methodology

Ready to ship the real answer?

Start free — 10,000 monthly visitors on us.

Start free — no cardBook a demo