Lift every cart.
Grow your store.
Paste your store URL — Cartlift audits it, drafts conversion variants, and runs the A/B tests for you. Winners ship automatically. More buyers, higher AOV, repeat customers — no agency.
Wake up with
glass skin.
Run any of these on every page that sells.
Paste a product page, a collection, or your checkout. Cartlift fetches the page, reads it, and ships back an annotated growth report in ~30 seconds. Same engine — four lenses for ecommerce.
Conversion audit
Find the parts of your PDP, cart and checkout that are quietly losing buyers — and get drafted page variants ready to A/B test.
- pdp headline + add-to-cart contrast
- social proof above the fold
- checkout + payment-method friction
- mobile cart cutoff
SEO audit
The boring-but-essential checks: title, meta, Product schema, heading hierarchy, speed, and link health — on the pages that actually rank.
- title + meta description
- Product + Offer + Review JSON-LD
- H1 / heading topical match
- render-blocking resources
Trust + policy audit
The checks shoppers (and Stripe, PayPal, Shop Pay) look for. Stops abandoned carts and silent suspensions before they happen.
- privacy + terms + returns alignment
- contact info + business identity
- payment-method visibility
- checkout transparency
Google Merchant audit
Suspension-grade audit for Shopping merchants. The same checks GMC reviewers run, written as actions you can ship this week.
- misrepresentation + prohibited content
- shipping + return policy alignment
- checkout sign-in walls
- policy contradictions across pages
Your store, improving itself — quietly.
Three steps. Cartlift does all of them. You approve, you keep selling, you read the changelog over coffee.
Cartlift reads the page.
Paste a URL. Cartlift fetches the live HTML, parses the visible copy + structure, and runs a Claude (or OpenAI) audit against it — conversion, SEO, trust, or Shopping feed. Findings come back annotated with predicted lift in ~30 seconds.
Cartlift drafts the variants.
Each conversion finding becomes a draft Experiment with 2–3 candidate rewrites — PDP headlines, add-to-cart copy, trust rows, upsell hooks, pricing micro-copy. Every variant carries its rationale. Nothing ships without your one-click approval.
Cartlift runs the trial.
Approved variants ship through a 3KB JS snippet — drop it in Shopify, WooCommerce, BigCommerce, or any headless front-end. A Thompson-sampling allocator re-weights every 30 min as orders come in. When the leader hits ≥95% posterior confidence on ≥500 samples with positive uplift, it pins to 100% traffic — losers trimmed to 0.
All your store experiments, one terminal.
A live look at what shipped, what won, and what is still in trial. AOV, conversion, returning-customer rate — all in one view. The numbers update themselves.
See what cartlift finds on yours.
Annotated feedback on your actual product page. ~30 seconds. No install. No card.
You wanted to ship one A/B test.
Three weeks later, you have eleven Slack threads, two stand-ups, a Figma comment storm, a Linear ticket no one owns, and zero variants live on your store. Cartlift replaces all of it: it audits, drafts, ships, and auto-pins the winner from one repo.
Blocked by · DESIGN
- ~30spaste the url → cartlift returns ranked conversion findings with predicted lift
- 1 clickpick a finding → claude (or openai) drafts 3 candidate rewrites with rationale
- 1 approvesnippet picks them up on the next page load — control + variants, sticky per shopper
- autothompson allocator re-weights every cycle · pins leader at ≥95% confidence with ≥500 samples
You can read every line.
No black-box pricing, no locked-in dashboards. The audit prompts, the bandit math, the snippet — all in one repo, MIT-licensed.
Growth tools should be code, not contracts.
Closed-source CRO and personalization platforms charge $1k–$5k/mo and lock your shopper data + experiment history inside their dashboard. Cartlift ships the same engine — audits, variant generator, Thompson-sampled allocator, A/B snippet — under MIT license. Clone the repo. Run it on a $5 droplet. Bring your own Claude or OpenAI key. The orders and conversion data stay in your Postgres.
Docker compose. Postgres + Django + Next.js. Runs on your laptop, your VPS, or your cluster.
Paste a Claude or OpenAI key in /dashboard/settings. Pick the model. Per-user — keys never touch any other server.
Edit the audit prompts, swap the bandit algorithm, fork the dashboard. It is all in one repo.
The questions store owners ask us most.
How does cartlift actually change my store?
A small JS snippet — about 3kb. Drop it in your Shopify theme, WooCommerce header, BigCommerce script manager, or any headless front-end. It loads asynchronously and only swaps the elements that match a variant's CSS selector. The original page is the control; shoppers who get the variant are routed by a Thompson-sampling allocator that re-weights as orders come in.
Will this break my store or my checkout?
The snippet wraps every DOM swap in a try/catch — if anything throws, the original page renders untouched. Every experiment has approve / pause / kill controls in the dashboard. The whole snippet is one template in api/snippet/views.py — read it before you paste it. It never touches your checkout flow unless you point a variant at it.
How do you decide what to test on my store?
You paste a URL — a PDP, a collection, your homepage, your cart. Cartlift fetches the live HTML and runs a CRO audit against it (Claude or OpenAI, your key). Each finding — vague headline, low-contrast add-to-cart, missing trust row, hidden return policy — comes back with a predicted lift, and you click generate variants to turn the findings you like into draft experiments. No telemetry, no session recordings — just the page.
How long until I see lift on AOV or conversion?
Depends on traffic. The allocator auto-pins a winner only after a variant clears ≥500 samples with ≥95% posterior confidence and positive uplift vs control. At ~10k weekly sessions, that is typically 5–7 days. Lower traffic stores see longer trials. The allocator can run on cron, but you can also trigger it manually from the dashboard or manage.py allocate_bandits.
What if I do not like a variant cartlift shipped?
Open the experiment and hit kill. Status flips to killed, the snippet stops serving the variant on its next refresh, and the original control is what every shopper sees. There is no "30-day suppression" magic — kill it, archive it, move on.
Is my store data shared with other customers?
No. Audits, variants, experiments, and exposure / conversion samples all live in your own Postgres (or our hosted Postgres scoped to your workspace). No shared training data. The only network call leaving your stack is the one to Anthropic or OpenAI with the key you set in settings.
Audit → Draft → Ship → Win. On repeat.
Run cartlift. Keep the data.
Run a free annotated audit on the hosted plan, or clone the repo and self-host the whole stack. CRO · SEO · Trust · Shopping feed — same engine either way.