How to run a simple A/B test with UTM parameters and traffic splitting

A/B testing sounds complicated, but the basic idea is simple.

You show one group of users version A, another group version B, and then compare which version performs better.

The hard part is not only sending users to two different pages. The hard part is making sure the traffic is labelled properly so the results can be analyzed later.

The fictional case: one QR code, two menu pages

Let’s say you run a food delivery business. You want to promote a QR code in a physical location, on a flyer, or inside a delivery bag.

But you are not sure which menu page will sell more.

Version A promotes family combo meals. Version B promotes quick lunch options.

Instead of printing two different QR codes, you create one public campaign URL. When someone scans the QR code, a small JavaScript function randomly sends that user to Menu A or Menu B.

This is not a perfect scientific experiment. It is a practical business experiment. The goal is to learn whether one menu experience creates more orders, more checkout starts, or higher average order value.

What is traffic splitting?

Traffic splitting means dividing visitors between two or more destinations.

In this example, the split is simple:

• 50% of users go to Menu A.
• 50% of users go to Menu B.

The user starts with the same URL, but the system redirects them to one of two destinations behind the scenes.

This is useful when the business wants to compare two paths without asking users to choose manually.

Why use one public URL?

One public URL is easier to manage.

If the URL is printed on a QR code, flyer, poster, receipt, packaging, or campaign material, you do not want to reprint everything every time the test changes.

The public URL can stay the same while the routing logic changes behind it.

For example:

https://menu-test.example.workers.dev/

That URL can send users to Menu A today, split traffic tomorrow, or later send everyone to the winning menu after the test ends.

Why use a 302 temporary redirect?

A 302 redirect means the move is temporary.

That matters because an A/B test is not a permanent migration. You are not saying “this URL has permanently moved.” You are saying “for now, send this visitor to one of these test destinations.”

For SEO and analytics, this distinction matters. A permanent redirect is a different signal. A temporary redirect is more appropriate when the routing is part of a test, campaign, or short-term experiment.

The example JavaScript traffic splitter

This type of logic can run in an edge function, serverless function, or worker. The important idea is that the user visits one URL, and the function decides where to send them.

export default {
  async fetch(request) {
    // Fictional example: one campaign link splits traffic between two menu pages.

    const menuA =
      "https://example-food.co/menu/family-combos";

    const menuB =
      "https://example-food.co/menu/quick-lunch";

    // Randomly assign each visitor to one variant.
    const variant = Math.random() < 0.5 ? "menu_a_family_combos" : "menu_b_quick_lunch";

    // Pick the matching destination.
    const target = variant === "menu_a_family_combos" ? menuA : menuB;

    // Add campaign tracking parameters.
    const url = new URL(target);

    url.searchParams.set("utm_source", "restaurant_qr");
    url.searchParams.set("utm_medium", "qr");
    url.searchParams.set("utm_campaign", "menu_ab_test_june");
    url.searchParams.set("utm_content", variant);

    // 302 = temporary redirect.
    // Useful for an experiment because the test is not a permanent URL move.
    return Response.redirect(url.toString(), 302);
  }
};

The important part is not the exact hosting provider. The important parts are:

• There is one public campaign URL.
• The script randomly chooses a variant.
• The user is redirected to the correct destination.
• UTM parameters are added before the redirect.
• The redirect is temporary, not permanent.

What are UTM parameters?

UTM parameters are tracking labels added to a URL.

They do not usually change the page content. Their job is to help analytics tools understand where traffic came from and which campaign, channel, or variant created the visit.

A URL with UTM parameters can look like this:

https://example-food.co/menu/family-combos?utm_source=restaurant_qr&utm_medium=qr&utm_campaign=menu_ab_test_june&utm_content=menu_a_family_combos

In simple terms, UTM parameters answer:

• Where did the visitor come from?
• What type of campaign was it?
• Which campaign was it?
• Which version of the creative, link, or experience did they see?

The main UTM parameters explained

For this food delivery test, the most important parameter is utm_content, because it stores whether the user saw Menu A or Menu B.

Without that label, you may see campaign traffic, but you will not clearly know which variant created the order.

Why UTM naming consistency matters

UTM parameters are useful only when they are consistent.

For example, these three values may look similar to a human:

utm_medium=qr
utm_medium=QR
utm_medium=qrcode

But analytics tools may treat them as different values. That creates messy reporting.

A simple naming convention helps. For example:

• Use lowercase.
• Use underscores instead of spaces.
• Keep campaign names short but descriptive.
• Use utm_content for variants.
• Do not change names halfway through the test.

What should the business measure?

The test should not stop at page views.

A menu page can get more clicks but fewer orders. Another menu page can get fewer clicks but higher average order value. That is why the business needs to define the success metric before launching the test.

A practical measurement plan

A simple plan could look like this:

1. Create one campaign URL for the QR code.
2. Split traffic 50/50 between Menu A and Menu B.
3. Add UTM parameters to both destinations.
4. Use utm_content to label the variant.
5. Track menu views, checkout starts, orders, and order value.
6. Compare conversion rate and revenue per visitor by variant.
7. Keep the test stable long enough to collect useful data.
8. Send all users to the winning version after the test ends.

Example reporting table

After running the test, the business could compare the variants like this:

In this fictional example, Menu B has a higher conversion rate, but Menu A has higher revenue per visitor because the average order value is stronger.

This is why analytics matters. If the business only looked at conversion rate, it might choose Menu B. If it looked at revenue per visitor, it might choose Menu A.

Common mistakes in this type of test

What makes this valuable for a business?

The value is not only the redirect.

The value is creating a measurable decision.

Before the test, the business may have opinions:

• “Family combos should sell better.”
• “Lunch offers are probably easier to convert.”
• “The QR code traffic probably behaves differently.”

After the test, the business has evidence:

• Which menu got more checkout starts?
• Which menu created more orders?
• Which menu created more revenue per visitor?
• Which menu should the campaign send users to next?

Where technical SEO and analytics meet

This kind of project sits between analytics, technical implementation, and growth.

The code is small, but the thinking matters:

• Use the correct redirect type.
• Keep the public URL stable.
• Label traffic clearly.
• Define the conversion goal before launching.
• Analyze business outcomes, not only clicks.

That is why I see analytics as part of organic growth. Visibility brings users in, but measurement explains what happens next.

How I can help

I help businesses design clearer measurement systems for campaigns, websites, SEO projects, and conversion journeys.

That can include UTM naming conventions, campaign tracking, A/B test setup, redirect logic, analytics audits, conversion event reviews, and data analysis.

The goal is not to track everything. The goal is to track what helps the business make better decisions.

Sources and further reading

• Google Analytics: Collect campaign data with custom URLs
• Google Analytics: About traffic-source dimensions
• MDN: 302 Found temporary redirect
• Cloudflare Workers: Redirect requests