Guide · On
[ 00 / INTRO ]
Built for founders running growth experiments alone

Too small for A/B testing?That's exactly the problem this solves.

NeriLabs writes, checks, and routes your pricing, onboarding, and messaging experiments — so a real winner shows up long before you'd hit statistical significance the old way.

frictionless_signup.exp · live routing
// bandit.assignment · thompson_sampling
/signup
control0 conv
/signup?v=b
email_only0 conv
/signup?v=c
social_proof0 conv
// session.telemetry0 sessions
min. sample gate30/arm
leading arm
Control
33%
Email-only
33%
Social proof
33%
0visits

What a standard A/B test needs before it'll trust a 2% lift. Most seed-stage sites won't see that in a quarter.

Optimizely and VWO were built for teams with an existing million visitors a month, testing button colors and image placement.

You have a few hundred visits a week, and the change that matters isn't a button — it's whether people pay $49 or $99. Small swings need small traffic to prove. Big swings are what you can afford to test.

So instead of a 50/50 split held for a month, NeriLabs shifts traffic toward the winner as evidence comes in.

[ 01 / METHOD ]
FOUR STAGES · HYPOTHESIS TO LIVE TRAFFIC
01 /

Write the hypothesis

You describe the change in plain language — no ticket, no spec.

"Our onboarding is too long — test a version that only asks for an email, and delay the rest until after they see the dashboard."
02 /

Generate & check

Copy, layout, and schema changes get drafted, then every variant runs through the guardrail layer before it's eligible to go live.

brand_safety — no flagged claim patterns
css_breakpoint — no mobile overflow risk
critical_element — checkout button intact
03 /

Route

A bandit shifts traffic toward whichever arm is actually converting as results come in.

control
v_b
v_c
04 /

Stay in sync

Prices, required fields, and onboarding steps stay consistent with your own backend for the whole session.

{ "aid": "opaque_id", "sid": "session_id", "cid": "customer_id" }
[ 02 / GUARDRAILS ]
NOTHING REACHES A VISITOR UNREVIEWED

Every variant is checked before it's eligible for traffic.

Rejected variants never go live, and every decision is logged — not discovered by a founder after the fact.

Brand tone and claim-pattern checks on generated copy
Mobile-breakpoint checks before layout changes ship
Checkout and CTA elements confirmed present in every variant
A signed, tenant-scoped token — never price directly — decides what a visitor sees
// guardrail.pipelinelive
input
pricing_hero_v3
check: brand_safety
✕ rejected — flagged claim: "guaranteed"
input
email_only_signup
check: css_breakpoint, critical_element
✓ approved — live, 1,204 impressions
341 rejected to date0 unauthorized price mismatches
[ 03 / MODES ]
THREE WAYS A TEST GETS STARTED
founder-directed

You bring the hypothesis

Type what you want to test in plain language. NeriLabs expands it, then holds it for your approval.

trigger: dashboard
behavior-triggered

The visitor tells you

Friction signals surface as pre-approved interventions, never invented live in the moment.

trigger: session telemetry
autonomous

The data flags itself

A confounder-adjusted anomaly gets surfaced as a proposal with its reasoning attached.

trigger: scheduled scan
[ 04 / PRICING ]
PRICED FOR THE TRAFFIC YOU ACTUALLY HAVE
Starter
$49/mo
up to 50,000 monthly visitors
UI & copy experiments
Bandit-based routing
Guardrail audit log
Start with Starter
Growth
$99/mo
up to 250,000 monthly visitors
Everything in Starter
Stripe-synced pricing tests
Onboarding funnel branching
Start with Growth
Scale
$199/mo
up to 1,000,000 monthly visitors
Everything in Growth
Autonomous anomaly detection
Priority guardrail review
Start with Scale
[ 05 / BEGIN ]

Stop waiting for traffic you don't have yet.

Connect your site, write one hypothesis, and see your first variant routing within the hour.

Start your first test →
[ EOF · 0xNERI · END OF TRANSMISSION ]