A/B Testing on 1C-Bitrix: Server-Side and Client-Side Methods

Imagine: you rewrote the product card, changed the order form, and conversion dropped by 20%. Without an A/B test, you won't know which element caused it. Bitrix developers often face the fact that the built-in abtest module does not consider caching and integrates poorly with analytics. From our practice: an electronics e-commerce store with 5,000 products wanted to test a new dynamic discount system. We implemented a server-side A/B test with cookie-based split, passing the group to dataLayer and tracking conversion in GA4. Baseline conversion was 2.5%, after 3 weeks of testing variant B reached 3.1% — a 24% increase with a p-value of 0.03. Let's evaluate your project and suggest the optimal solution. Contact us for a consultation.

Problems We Solve

• Caching: if a component caches its result, both variants get the same HTML. Solution — add the cookie BITRIX_SM_ABTEST_{ID} to the cache key or disable caching for the tested block.
• Analytics integration: the built-in module does not directly send data to Yandex.Metrica or GA4. We configure dataLayer and custom parameters to build reports broken down by groups.
• Server-side logic: testing prices, discounts, or algorithms requires custom PHP code. The built-in module is not suitable for this.

How We Do It: Stack and Case Study

We use PHP 8.1+, infoblocks v2.0, and tag-based composite caching. For the electronics e-commerce store (from our practice), we implemented a server-side A/B test with custom PHP code. The group detection code:

function getABGroup(string $testName, int $percentB = 50): string
{
    $cookieName = 'ab_' . md5($testName);
    if (isset($_COOKIE[$cookieName])) {
        return $_COOKIE[$cookieName];
    }
    $group = (mt_rand(1, 100) <= $percentB) ? 'B' : 'A';
    setcookie($cookieName, $group, time() + 86400 * 30, '/');
    return $group;
}

// Usage
if (getABGroup('discount_algorithm') === 'B') {
    // New discount algorithm
} else {
    // Current algorithm
}

How the Built-in abtest Module Works

The abtest module is available in Business and Enterprise editions. You create a test specifying the percentage of traffic for variant B and choose the type (template, component, include area, PHP code). Bitrix assigns groups via cookie. API creation:

\Bitrix\ABTest\ABTestManager::addTest([
    'NAME' => 'Buy button: red vs green',
    'SITE_ID' => 's1',
    'DURATION' => 14,
    'PORTION' => 50,
    'TEST_DATA' => [
        'type' => 'template',
        'original' => '/local/templates/main/',
        'modified' => '/local/templates/main_test/',
    ],
]);

How to Choose the Approach: Server-Side or Client-Side?

Client-side tests (VWO, Optimizely) require no server code changes but suffer from FOUC delay and cannot test server logic. The server-side approach, on the other hand, allows testing prices, discounts, algorithms — everything that runs on PHP. If you need to test a template or include area change, the built-in module is sufficient. For complex scenarios (different prices for groups, personalization) — custom server-side code. We help you choose the method for your tasks.

Why Caching Is the Main Enemy of A/B Tests

A typical mistake: a component caches HTML, and both variants show identical content. The solution — add the test identifier to the cache key or disable caching for the tested block. For example, via $arParams['CACHE_TIME'] = 0; or by using \Bitrix\Main\Data\Cache::setCacheTag(). Without this, test results will be incorrect.

How to Achieve Statistical Significance

A typical mistake — stopping the test after 2 days seeing a 0.5% difference. For reliability, you need a sufficient sample size: with a baseline conversion of 2% and a desired effect of 20%, you need approximately 20,000 visits per variant. On a site with 1,000 visits per day — 40 days. Do not stop the test until the p-value drops below 0.05. We use a statistical significance calculator for precise calculations.

Comparison of Approaches

What Is Included in Our Work

• Audit of current architecture and traffic
• Formulation of hypotheses and metric definition
• Approach selection and testing mechanism setup
• Caching problem resolution and analytics integration (dataLayer, GA4, Yandex.Metrica)
• Required sample size and duration calculation
• Monitoring, data collection, and statistical processing
• Report with conclusions and recommendations

Work Process

Estimated Timeline

From 2 days to 2 weeks depending on complexity. Pricing is determined individually after an audit. Get a consultation on setting up A/B testing.

Typical Mistakes

• Stopping the test prematurely: even if you see a visible difference, wait for statistical significance.
• Ignoring caching: remember to disable caching for tested components or separate it by groups.
• Incorrect segmentation: if the test runs on all users, results can be diluted. Account for seasonality and audience.

We have 10+ years of Bitrix development experience and certification. We guarantee correct setup and interpretation of results. Contact us for a project audit.

Duplicate Products on Page 3: A Bug That Goes to Production

A real case: an online store with pagination through bitrix:catalog.section duplicates products on page three after every second visit. Cache clearing helps for a day, then the duplicates return. Root cause: a custom sort handler collides with PAGEN_1, and under a specific filter combination CIBlockElement::GetList returns identical IDs. Code review missed it; only testing caught it. We build QA for 1C-Bitrix projects that catches such bugs before they hit production: manual functional, automated E2E, load, and acceptance testing. With over a decade of Bitrix experience, we have a library of typical pitfalls and test scenarios that prevent these issues from the start.

How Does a Standard Bitrix Project Break Without Dedicated Testing?

1C-Bitrix is not a landing page. Behind the frontend lie dozens of modules, external integrations, and non‑obvious dependencies. A discount change in sale.discount breaks a promo code in sale.basket.discount — the discount module is one of the most fragile in the platform. Interchange with 1C via catalog.import.1c or REST fails when property mapping is off, resulting in products without price or stock. Core updates — bitrix:main updated, a custom component uses the deprecated CModule::IncludeModule. Without regression testing, every deployment is Russian roulette. Cross‑browser: sale.order.ajax renders differently in Safari and Chrome; the “Place Order” button can move off‑screen on an iPhone. These are not edge cases — they are daily realities for Bitrix teams.

What Does Functional Testing Cover?

We check every business scenario — not just “works or doesn’t work”, but all boundary cases.

Catalog (catalog.section, catalog.element)

• Smart filter catalog.smart.filter: all property combinations, reset, result counting. Filters by SKUs break most often.
• Sorting + pagination — the duplicate bug described above.
• Comparison via catalog.compare.list — add, remove, display differences.
• Quick view — modal window, cart from modal.

Cart and Order (sale.basket.basket, sale.order.ajax)

• Adding from catalog, product page, quick order.
• Discounts: by quantity, by amount, by coupon, cumulative. Discount intersection — at least eight test combinations.
• Delivery calculation: handlers sale.delivery.services, cost, time, pickup points on map.
• Payment: sale.paysystem — processing, handling declines, refunds.
• Order placement: email via main.mail.event, CRM recording, transmission to 1C via sale.export.1c.

Personal Account (sale.personal.section)

• Registration, authorization, password recovery — including Cyrillic email edge cases.
• Order history, repeat order.
• Subscriptions, bonus program.

Forms and Search

• form.result.new / iblock.element.add.form — submission, validation, file fields.
• search.page — relevance, morphology, typo handling via search.title.

Why Is Regression Testing Critical for Bitrix?

After every deployment we verify that nothing previously working is broken.

• Smoke tests — main page loads, catalog shows products, order completes. 5 minutes, run after every deploy. If smoke fails — roll back immediately.
• Regression suite — 40–80 test cases covering main scenarios before every release.
• Visual testing — screenshot comparison (Percy or Playwright). A button shifted 20px, font changed after update — test shows the diff.
• Module checklists — structured lists for sale, catalog, iblock, search. Each module has its own checklist.

What Happens During Load Testing?

The question is not “will the site handle it” but at how many concurrent users catalog.section starts returning 500 errors.

Tools:

• k6 — JavaScript scripting.
• Apache JMeter — classic, for complex scenarios with cookie authorization.
• Yandex.Tank — real‑time visualization, integration with Overload.

Output: peak RPS, response times by percentiles p50/p95/p99, bottlenecks (CPU, RAM, MySQL slow queries on b_iblock_element, file cache). Recommendations: which index to add, which query to rewrite with D7 ORM, where to enable composite cache.

What Deliverables Do You Receive After Testing?

• Test plan with scope, priorities, and quality criteria.
• Test case suite — functional, regression, load.
• Defect report in a tracker (Jira/YouTrack) with severity classification.
• Auto tests (Playwright/Cypress) — basic smoke suite for CI/CD.
• Load testing protocol with graphs and recommendations.
• Acceptance certificate after UAT — confirming readiness for launch.

After delivery, we provide free consultation for a month — answering questions on test improvements and process adaptation. Contact us to receive a full package of documents and auto tests.

Cross‑Browser Testing

We test where buyers actually are. Statistics from your Metrica are more important than general market data.

Minimum set:

• Chrome (last 2 versions) — main traffic.
• Safari on iOS — critical for mobile checkout, sale.order.ajax often behaves unpredictably.
• Yandex.Browser — significant share in Russia, Chromium‑based but with extension quirks.
• Samsung Internet — mobile Android, often forgotten.

Devices:

• Desktop: 1920×1080, 1366×768.
• iPhone: 375×812, 390×844 — checkout must be verified.
• Android: 360×800, 412×915.

Tools: BrowserStack for real devices, Playwright for automation on Chromium/Firefox/WebKit.

Automation

Playwright — primary choice for E2E on Bitrix:

• Cross‑browser: Chromium, Firefox, WebKit.
• Parallel execution, automatic waits.
• Works well with dynamic forms sale.order.ajax.
• Supports mobile viewports and geolocation.

Cypress:

• Runs in browser — more stable for SPA‑like interfaces.
• Excellent visual runner for debugging.
• Limitation: only Chromium‑based browsers.

PHPUnit for custom code:

• Unit tests for custom Bitrix components and modules.
• Tests business logic without frontend dependency.
• Integration with CI/CD — GitLab CI, GitHub Actions.

UAT – Acceptance Testing

Final check with the client on a staging environment with real data:

• Jointly compile a list of critical scenarios — 15–20 key customer paths, not 200 test cases.
• Staging with a copy of the production database (anonymized personal data).
• Quick bug tracking — Jira/YouTrack, prioritization by severity.
• Acceptance protocol — document with results, signatures, and launch readiness.

Order UAT support and we guarantee a release without surprises.

QA Process – Integrated, Not Tacked On

1. Requirements analysis — QA participates in task discussions, catches ambiguities. “Does the discount apply to the product or the order?” — such a question upfront saves two days of debugging.
2. Test cases before development — scenarios ready before the first line of code.
3. Code review — checks for typical Bitrix mistakes: uncleared component cache, direct SQL queries instead of ORM, missing $USER‑>IsAuthorized() check.
4. Functional → regression → deploy.
5. Post‑release monitoring — errors in bitrix/error.log, metrics in Metrica, alerts for 500 errors.

We have been working with Bitrix for over 10 years and have tested more than 300 projects of various scales — from small online stores to corporate portals with 1C and Bitrix24 integration.

Timelines

Testing cost is calculated individually for your project. Get a free consultation — we will assess the scope within one business day and provide a preliminary estimate and test plan. Contact us to discuss your project and schedule a call.