Roadmap

The full loop: read, reason, render, verify, remember

One snippet, <8KB gzipped, no framework. Hypotheses ranked by expected lift × confidence ÷ effort; variants shipped to 45/45/10 split traffic.

Permanent 10% holdout + honesty-gate reporting

Every lift number measured against visitors who never see a variant. Below statistical power she says "not yet" — never "winner."

Day-0 baseline: visitor types, competitor scan, brand fingerprint

The homework from section 02 of the homepage — ICP recognition, market study (public pages only), voice/tone/image corpus, power math.

Five-layer safety pipeline + brand drift caps

Static checks, render check, vision review, stylometric review, soft launch. Drift caps: voice 0.07, image 0.10, layout 0.20, CTA 0.10.

MCP server — Claude as the first surface

Sigrid is a Model Context Protocol server: ask about experiments, approve variants, pull the audit log from Claude or any MCP-capable tool. The chat apps are conveniences; the protocol is the contract.

Slack + email surfaces

Hypotheses, approvals and results in one thread. Email digest if you prefer it quiet.

Read integrations: GA4, Microsoft Clarity, PostHog, Stripe, snippet

Read-only, aggregates not individuals. The snippet fills the gaps the tools cannot see.

The €79 standalone audit

The Day-0 baseline productized: ranked hypotheses plus the honest math on how fast lift can be proven on your traffic. Credited to your first month.

The public verified-lift ledger

Verified lift across customers, published live — including inconclusive tests. Launches with the product.

Shopify + Webflow native publishing

The snippet covers both at launch (L-001: snippet-only at MVP); native publishing ships when this earns its place in the queue.

Plausible

Privacy-first analytics as a read source.

Fathom

Privacy-first analytics as a read source.

Matomo

Self-hosted analytics as a read source.

Mixpanel

Product analytics: funnels and cohorts as hypothesis fuel.

Amplitude

Product analytics: funnels and cohorts as hypothesis fuel.

Heap

Auto-captured events as a read source.

Hotjar

Aggregated heatmaps and funnels — no individual replays.

Segment

One pipe for everything you already collect.

Google Search Console

Query and landing-page data: what visitors wanted before they arrived.

Google Ads

Campaign and cost data — closes the loop from spend to verified lift.

Meta Ads

Campaign and cost data — closes the loop from spend to verified lift.

TikTok Ads

Campaign and cost data — closes the loop from spend to verified lift.

Shopify (orders & analytics)

Orders, refunds, revenue per session — the store-side read. (Publishing is already launch-committed.)

WooCommerce

Orders and revenue as a read source; snippet ships the variants.

BigCommerce

Orders and revenue as a read source; snippet ships the variants.

Paddle

Payments read for SaaS that does not run Stripe.

Lemon Squeezy

Payments read for SaaS that does not run Stripe.

Klaviyo

Email revenue attribution and customer-voice mining.

Mailchimp

Email revenue attribution and customer-voice mining.

HubSpot

Lead quality feedback — which page variants produce leads that close.

Pipedrive

Lead quality feedback — which page variants produce leads that close.

Intercom

Support themes as hypothesis fuel — what visitors ask is what the page failed to say.

Crisp

Support themes as hypothesis fuel.

Zendesk

Support themes as hypothesis fuel.

Discord surface

The full Slack experience, in Discord.

Microsoft Teams surface

The full Slack experience, in Teams.

Telegram surface

The full Slack experience, in Telegram.

Native code-commit mode

Winning variants land as pull requests to your repo instead of living in the snippet — the test becomes the code.

Pay-per-verified-lift pricing option

Pricing tied to holdout-verified results. Only auditable because the holdout exists — which is why nobody else offers it honestly.

Multi-site console

Portfolios and agencies: many sites, one ledger, per-site honesty gates.

"Verified by Sigrid" public badge + report embeds

Co-branded, checkable lift reporting you can show clients or investors.

Audit-log export API

Every change, decision and verdict — yours to pull, machine-readable.

Ad-to-page intent matching

Variants matched to the ad that brought the click — message continuity for paid traffic, within the same safety caps.

A/B history importer

Bring your VWO / Optimizely / GO history as priors, so Sigrid does not re-learn what you already paid to know.

Built-in lightweight heatmaps

Aggregate attention maps from the snippet — no third-party tool needed. Considered: may be redundant next to Clarity/Hotjar reads.

Multi-language variant generation

Variants in the page’s language with per-language voice corpora. Considered: EU markets first.

White-glove onboarding for the first cohort

A human call to wire the baseline. Considered: does not scale, might not need to.

Countdown timers / fake scarcity

Manufactured urgency converts and corrodes. Anti-dark-pattern constraints are part of the objective function, not a setting.

Per-visitor session replay

Aggregates only. Testing requires no PII, so Sigrid does not collect any.

Free tier

Free tiers optimize for signups; Sigrid optimizes for proof. 30-day money-back instead. Locked.

Configurable significance ("call winners early")

The honesty gate is not a setting. Below power she says "not yet" — that is the product.

White-label for agencies

Co-branded "Verified by Sigrid" instead. A ledger that hides its name is not a ledger.