Documentation
Guides for installing, running and measuring
Each guide covers one task and says what to do, in order. Start with the script tag if Pertento is not live on the site yet, or search for the thing you are stuck on.
128 guides
01
Pertento Runtime
The script tag, who sees a variant, and what it costs the page.
- URL Targeting – Choosing Which Pages an Experiment Runs OnLearn how to use URL targeting conditions to control exactly which pages an experiment runs on.
- Cookie Targeting – Running Experiments for Specific AudiencesLearn how cookie targeting restricts an experiment to visitors carrying particular cookie values.
- Traffic Allocation and Variant WeightsLearn how Pertento splits traffic between variants, and what happens when a variant has no explicit weight.
- Pertento Runtime Script and Core Web Vitals ImpactSee how the Pertento runtime script preserves all Core Web Vitals using a lightweight, deferred, non-blocking architecture that ensures experiments run without…
- Anti-Flicker Snippet (Optional)A quick guide to the optional anti-flicker snippet that reduces visual flicker during experiment activation.
- Installing the Pertento scriptLearn how to install the lightweight Pertento script.
02
Editor
Help articles relating to Pertento's website editor
- Applying a change to every matching elementThe CSS selector tool styles or adds HTML to everything a selector matches, including elements that load later through infinite scroll or carousels.
- Dynamic content: showing page data in a variantBuild a change once and let every page fill in its own data: product prices, campaign parameters, visitor details, countdowns and cart values.
- Adding components and blocks in the editorUse the Add tab to drop ready-made sections, layouts and media into a variant. Components take on your site's fonts and colours, so they look native.
- HTML, CSS and JavaScript in a variantWhen the visual tools are not enough, edit an element's HTML or write the variant's global CSS and JavaScript, with the same page data as tokens.
- Previewing a variant on another deviceShare a link that shows a variant's saved changes on any phone or browser, and a second link that puts that browser back to normal.
- Data feeds: using your own feeds in experimentsConnect a JSON or Google Merchant feed your website already publishes, choose which item each visitor sees, and use its fields in any variant.
- Visual editor keyboard shortcutsEvery keyboard shortcut in the Pertento visual editor: saving, undo and redo, editing HTML, duplicating, hiding, adding, removing and moving elements.
- The editor inspector: styles, states and attributesThe inspector styles the selected element: typography, spacing on a box model, size and layout, hover and focus states, attributes and any CSS property.
- Using the Pertento editor with CSPLearn how to whitelist Pertento (including wildcard subdomains) in your CSP to prevent blank pages in the editor, or use a local browser plugin as an alternative…
- Setting up the Pertento editor in Google ChromeInstall the Pertento editor extension, sign in from Chrome's side panel, grant access to your site and open the visual editor on any experiment and variant.
- Visual editor: editing experiments without codeHow the Pertento visual editor is laid out, how to select, edit, move and remove elements on your live site, and how changes, undo and saving work.
03
Platform
Help docs relating to the Pertento Platform
- Payment Details, Invoices and CancellingLearn where to update your card, find receipts and VAT details, download a monthly usage report, and cancel a plan.
- Going Over Your Session AllowanceLearn when Pertento warns you about usage, when going over moves you up a plan, and what choosing Keep my plan does instead.
- Starting a Trial and Changing Your PlanLearn how the 12-day free trial works, how to choose a plan at checkout, and when an upgrade or a downgrade takes effect.
- AI Credits and Credit PacksLearn how AI credits work, which credits are spent first, what happens at zero, and how to buy a credit pack.
- Plans, Sessions and Your Monthly AllowanceLearn what a session is, how your plan's monthly allowance is counted across your websites, and how to read the usage panel in Settings.
- Failed Payments and Read-Only AccountsLearn what happens when a payment fails, how long experiments keep serving, what a read-only account can still do, and how to restore it.
- Getting Around the Pertento ConsoleLearn how the sidebar is organised, how to switch account and website, and what each item in the header tells you.
- Billing for Agencies: the Agency Fee and Client PlansLearn how agency billing works: the agency fee, a separate plan for each client account, who is notified, and what a client sees.
- Your Profile Photo and Badge ColourLearn how to add a profile photo or choose a badge colour, so colleagues can tell who did what at a glance.
- Your Data: Retention, Export and Closing an AccountLearn how long Pertento keeps your data, what you can export from the console, and how to have an account and its data deleted.
- Roles and Permissions in PertentoLearn what members, admins and owners can each do, where the rule is enforced, and how the activity log records who changed what.
- Your Account and Organization SettingsLearn what lives in your own account settings versus your organization settings, and how to secure your login.
- Managing Client Accounts as an AgencyLearn how agency accounts manage separate client organisations, each with their own websites and experiments.
- The Pertento Dashboard ExplainedLearn what every tile on the dashboard measures, and which figures follow the selected time period.
- Notifications in PertentoLearn which events Pertento notifies you about, and what each notification is telling you to do.
- Setting Up Passkeys for Account SecurityLearn how to enable passkeys for your Pertento account, add multiple devices and recover access if a passkey is lost.
04
Experiments
Help articles relating to building, running and reading experiments in Pertento
- Checking a variant on your live site with QAPreview a draft experiment on your live site with a link before you start it. Visitors never see it, nothing is tracked, and QA ends by itself after 3 days.
- Asking AI about an experimentLearn how to ask the AI assistant about an experiment's result, set-up, brief and debrief, what it can and cannot see, and what each answer costs.
- Duplicating variants and experimentsYou can copy a variant inside an experiment, or copy a whole experiment, including onto another of your websites.
- Experiment DebriefsWhen a test ends, the numbers tell you what happened. They do not tell you why, what you expected, or what you will do next.
- Experiment Library timelineView every experiment on one timeline: what ran when, what's scheduled, and what's expected next in your pipelines.
- Experiment Library search, filter and resultsSearch past tests by hypothesis, filter by outcome and expected mechanism, and see what each test tried and taught without opening it.
- Experiment Run length plannerBefore you launch, find out how long a test needs to detect the improvement you care about, based on your website's real traffic and conversion data.
- Program Monitor ReportSee how often your tests win, how fast you launch them, whether winners get rolled out, and which kinds of ideas tend to work.
- Pipelines: Running Experiments in StagesLearn how a pipeline runs a set of experiments as ordered stages, how each stage advances, and what the two automation switches do.
- Scheduling Experiment Starts and StopsLearn how scheduled starts and stops work, what each schedule state means, and the conditions an automated stop has to clear first.
- The Implementation QueueLearn how the implementation queue tracks winning variants that still need to be built into your codebase.
- Finding Experiments in the Experiment LibraryLearn how to filter, search and sort the Experiment Library, and what the Draft, Running, Deployed, Ended and Archived views each contain.
- How Pertento Decides What Needs Your AttentionLearn how Pertento ranks experiments by what needs doing, and what each signal on the attention list means.
- Duplicating an ExperimentLearn what Pertento copies when you duplicate an experiment, and what it deliberately leaves behind.
- Prioritising Experiments with PXL ScoresLearn how PXL scoring rates Potential, Exposure and Level of effort to decide which experiment to run next.
- The Experiment Timeline and Activity LogLearn how the timeline and activity log record what changed during an experiment, and why that record matters when a result looks wrong.
- When Traffic Does Not Split as ConfiguredLearn why a traffic split that does not match its weights invalidates an experiment, and how to find the cause.
- Reading the Power Analysis CardLearn how to read statistical power, resolvable effect and time remaining, and tell "no effect" apart from "not enough data".
- The Performance Page – Positive and Negative ExperimentsLearn how the Performance page separates experiments trending positive from those trending negative.
- The Health Page – Monitoring Every AccountLearn how the Health page surfaces problems across every account and website you can reach, in one view.
- Frequentist and Bayesian EvaluationLearn what frequentist and Bayesian evaluation each tell you, and which one answers the question you are actually asking.
- Correcting for Multiple VariantsLearn why testing several variants at once produces false winners, and what each correction method does about it.
- The Variant Combination ReportLearn how to check whether two overlapping experiments interfere with each other, and how to read the interaction odds ratio.
- Choosing Objectives and Currency on the Statistics PageLearn how the objective and currency selectors change every figure on the results page, and why revenue needs one currency.
- Deploying an ExperimentLearn how Pertento deploys a winning variant, stops traffic splitting and creates a new 100% allocation variant for full rollout, including how to edit or end a…
- Understanding Significance in the Statistics PageLearn how to interpret the zScore, pValue, SPRT result and overall significance indicators on the statistics page to understand whether a variant’s performance is…
- Enabling debug mode for experimentsLearn how to activate Pertento’s debug mode using the console or local storage to inspect experiment behavior and troubleshoot activation during development.
- Overriding variants trough local storageLearn how to manually switch experiment variants using local storage for testing and QA. This method lets teams preview and validate variants without influencing…
- Making changes to a running ExperimentLearn how to safely update variant code, targeting or tracking for an active experiment and how to enforce changes to apply them without compromising test…
- Instant Launches – Deploying and Pausing Experiments ImmediatelyThis article explains how to quickly launch and pause A/B tests using Pertento. With Pertento’s platform, you can deploy experiments live on your site instantly –…
05
Learnings
What your finished tests have taught: turning learnings on, reading the evidence, opportunities, Ask, AI credits and privacy.
- Learnings and AI CreditsLearn what Learnings costs in AI credits, how the estimate for past tests works, what happens when credits run out, and the allowance for accounts without credits.
- Reading the Learnings PageLearn what each part of the Learnings page shows: the three views, a pattern, a test's debrief, the device filter, and learnings in the Library and on drafts.
- Opportunities in LearningsLearn what the five kinds of opportunity mean, how Draft experiment turns one into a draft with its hypothesis and targeting filled in, and how to dismiss one.
- How Learnings Weighs the EvidenceLearn how Learnings pools your tests into patterns, what the five evidence levels mean, why ranges are adjusted, and which tests are left out or count for less.
- Getting Started with LearningsLearn what Learnings does, how an Admin or Owner turns it on for a website, how to include past tests, and what happens while the first learnings are written.
- Learnings Privacy and DataLearn what Learnings sends to our AI provider, what it keeps, and how the account and website settings for anonymised data work.
- Ask About Your LearningsLearn how to ask questions about a website's learnings, what the assistant reads, how it quotes figures and cites tests, and its limits and costs.
06
Feature flags
Releasing a change safely: the two kinds of flag, rollouts and plans, guardrails, rings, and reading a flag in the browser or on your server.
- Turning Off, Archiving and Deleting a Feature FlagWhat Turn off, Archive and Delete each do to a flag and its visitors, how fast they take effect, how to schedule a turn off, and how to clean up afterwards.
- Reading a Feature Flag in the BrowserHow your page's own JavaScript asks whether a code flag is on, reads its payload, and sets the attributes rings can match, plus the cases where a flag reads Off.
- Who Gets a Feature Flag: Targeting, Server Rules and RingsHow Pertento decides which side a visitor gets: targeting first, then rings, then the rollout.
- What Is a Feature Flag?A feature flag turns a change on for a share of your visitors and widens that share step by step.
- Rolling Out a Feature Flag: Rollouts, Plans and SchedulesHow a flag's rollout works, how a rollout plan moves it on its own, what a holdback keeps aside, how to schedule a publish or a turn off, and what each status in…
- Reading a Feature Flag on Your ServerHow your backend decides a code flag with the Pertento server SDK or over HTTP, what a decision tells you, how exposures are counted, and how to seed the page so…
- Why Does My Feature Flag Read Off?A checklist for a flag that reads Off when you expected On, a guardrail that stays at Collecting, a plan that is not moving, and a change that has not reached…
- Creating and Publishing a Feature FlagHow to create a feature flag, choose how it runs, name its key and sides, and what happens to visitors the moment you publish it.
- Quick Fixes with Editor FlagsAn editor flag carries changes made in the Pertento visual editor and puts them live without a release.
- Feature Flag GuardrailsA guardrail checks that On does no worse than Off on the flag's primary goal before a rollout plan moves on.
07
Surveys
Help articles relating to on-site surveys in Pertento
- Survey Results by Experiment VariantLearn how to split a survey's results by experiment variant, what each column shows, and why a variant with few respondents is flagged.
- Choosing Who Sees a SurveyLearn how page rules, devices, an experiment link and the share of visitors decide which visitors can be shown a survey.
- Writing Survey Questions, Options and TextsLearn the four question types, how choice options and optional questions work, how to reorder questions, and how to write your own button texts and scale labels.
- Sentiment Analysis on Survey CommentsLearn how to switch on sentiment analysis for a survey, what it costs in AI credits, how to analyse comments collected earlier, and what happens when credits run…
- Survey Appearance: Position, Colour and CornersLearn the four survey positions, how the accent colour and corner radius are applied, how the widget steps through questions and closes, and how it behaves on your…
- How Often a Visitor Sees a SurveyLearn the rules that decide whether a visitor is asked again, what the two frequency settings control, and what happens with several surveys running.
- Building a Survey: Types, Questions and PreviewLearn how the five builder steps work, what the NPS, CSAT, CES and Custom types each start with, and what the live preview shows.
- Starting, Ending and Managing SurveysLearn what each survey status means, why a started survey goes live at the next publish, what you can still change once it has run, and what duplicating, archiving…
- Why Isn't My Survey Showing?Work through the common reasons a running survey does not appear, from an unpublished website and targeting rules to frequency limits in your own browser.
- Reading Survey ResultsLearn how a survey's score, response rate and per-question results are calculated, how to read responses over time, and how the responses feed groups each…
- What Surveys Collect, and What They Cost the PageLearn exactly what a survey sends and stores, how answers are checked before they are kept, what it keeps in the visitor's browser, and how the survey script loads.
- Choosing When a Survey AppearsLearn how the five survey triggers work, page view, delay, exit intent, a dataLayer event and scroll depth, and which one fits what you are asking.
- Creating and Running SurveysLearn what a Pertento survey can ask, who sees it and when, how it goes live, and how answers tie back to the experiment a respondent was in.
08
Tracking
Resource docs related to experiment tracking within Pertento
- Tracking Shopify Purchases with a Custom PixelSend completed Shopify purchases directly to Pertento while preserving the experiment and variant assignments needed for attribution.
- Objectives and saved objectivesEvery experiment is measured on objectives: the events your website sends, like purchase or add_to_cart.
- Experiment Targeting LibrarySave an experiment's targeting under a name and load it onto other experiments on the same website.
- Experiment Activation targetingActivation targeting lets an experiment wait for a signal from your own website before it runs.
- Anonymous Visitor IDVisitor ID gives each browser that visits your website a random, anonymous identifier. Pertento attaches it to the events and experiment enrolments it records for…
- Setting Experiment Goals from Analytics EventsLearn how to choose which events an individual experiment measures, from your connected analytics property.
- Reporting Experiment Exposures via the Tracking APIA reference for the /visitors endpoint, which records that a visitor was exposed to a particular variant.
- Sending Website Sessions via the Tracking APIA reference for the /sessions endpoint, which records site-wide traffic by device type.
- Posting Conversion Events via the Tracking APIA reference for the /events endpoint: the payload shape, how revenue is parsed, and how events are attributed to variants.
- Accessing Pertento Experiment Statistics in GA4Learn how to analyze experiment performance directly in GA4 using automatically generated Pertento variant audiences and events, and how to apply them to reports…
- How Pertento uses local storageLearn how Pertento uses local storage to persist experiment variants, maintain consistent exposure across sessions and support variant overrides for testing,…
- Connecting a Pertento Website to GA4Learn how to connect Pertento to GA4, configure server-side endpoints and understand how Pertento sends variant events.
- Reporting Conversions and EventsLearn how to report purchases and other key events to Pertento using pertentoReportConversion.
- Live Insights – Real-Time Experiment Results and AnalyticsPertento’s Live Insights provide real-time A/B test results. This article shows how to monitor your experiment’s performance live – including visitor counts,…
- Privacy‑Safe Experimentation: How Pertento Ensures Compliance Without Sacrificing SpeedPertento ensures your A/B tests are privacy-safe by design – no cookies, no personal data, and no external trackers.
09
AI assistants and MCP
Connect Claude, ChatGPT or any other MCP client to Pertento: setup, scopes and websites, previews and confirmations, and managing access.
- Using Pertento from an AI assistantConnect Claude, ChatGPT or any other app that supports MCP to Pertento, and let it read your results and build and run experiments for you.
- Connecting ClaudeAdd Pertento as a custom connector in claude.ai, sign in, choose what Claude may do, and start asking about your experiments.
- Connecting Claude CodeConnect Claude Code in your terminal to Pertento with a personal token, or with the same sign-in claude.ai uses.
- Connecting ChatGPTAdd Pertento to ChatGPT as a connector in developer mode, then sign in and choose what ChatGPT may do.
- Connecting LovableAdd Pertento to Lovable as a custom MCP server, so Lovable can read your experiment results while it builds your app.
- Connecting another MCP clientPertento works with any app that supports remote MCP servers, not only Claude and ChatGPT. Here is how to connect one.
- Scopes and websites for an AI assistantWhat each permission lets an assistant do, how to limit it to some of your websites, and how your own role sets the ceiling.
- How an assistant makes live changes safelyAssistants only edit drafts, and anything that changes what visitors see is previewed and confirmed first. Here is what the preview shows.
- Visual changes made by an assistantAn assistant can add, edit and remove visual changes the way the visual editor does. Here is how they appear in the editor, and what it cannot change.
- Managing tokens and connected appsSee every token and connected app in Agent access, revoke or disconnect one, and know when access ends on its own.
- AI assistants and MCP: FAQShort answers about using Pertento from an AI assistant: which apps work, cost, data, running tests and common connection errors.
10
Server-side
Experiments your own code renders: the server SDK, server keys, the HTTP API and server-side conversions.
- Reporting Conversions from Your ServerLearn how to report purchases and other conversions from your backend with the SDK's track() or POST /sdk/events, using the Server-side tracking method.
- Running a Server-side Experiment with PertentoLearn how a Server-side experiment works: your own code shows each variant, and either the Pertento snippet or your backend decides who gets which.
- Installing the Pertento Server SDKLearn how to install @pertentoai/sdk on your backend, decide variants locally, pass the decision to the page and report conversions.
- Why Isn't My Server-side Experiment Counting Visitors?Work through the usual reasons a server-assigned experiment shows a variant but records no visitors or conversions, from a missing seed to a refused key.
- Setting Up an Experiment with Assignment: ServerLearn the steps that let your backend decide a Server-side experiment: the visitor ID, a server key, a flag key, variant keys and payloads, and the code.
- The Server-side HTTP API: Decide, Events and ConfigA reference for the authenticated /sdk endpoints, for a backend without the Pertento SDK: deciding variants, reporting conversions and reading the setup.
- Server Keys: Creating, Rotating and RevokingLearn how to create a server key for your backend, keep it safe, rotate it without downtime and revoke it.
- Showing a Browser-assigned Variant from Your Own CodeLearn how your page code shows the variant the Pertento snippet picked in a Server-side experiment with Assignment: Browser.
- Attribute Rules and Bucketing on a Customer IDLearn how to choose who a server-assigned experiment enrols with attribute rules, and whether it buckets on the visitor ID or your own customer ID.