Smart routing

One link, several destinations

Normally a link has one destination. With routing rules it can choose between several, decided at the moment of the click from what we can tell about the visitor. Rules belong to a link: open a link from the Links list and choose its Routing tab, beside the Details tab that holds the destination the rules override.

Those three are conditional rules: they ask a question about the visitor and match or they do not. The fourth rule type, an A/B split, works differently: it is a weighted choice rather than a test that passes or fails, it takes no part in the ordering below, and it changes where unmatched clicks end up. A/B splits covers it on its own.

How a destination is chosen

Conditional rules are tried from the top of the list, and the first one that matches decides the destination. Drag to change that order. The dashboard lists them under Match in order, which is the sequence a click walks. Paused rules are skipped entirely; the redirect engine is only ever handed the active ones.

When no conditional rule matches, the click does not go straight to the link's ordinary destination. Any active A/B split on the link is considered next, and the ordinary destination is only reached when the link has no split taking traffic. Keep it pointing somewhere sensible all the same: it is what a link with no split falls back to, and what every link resolves to once you pause or remove its rules.

Writing the rules

• Device: choose Mobile, Desktop, or Tablet.
• Country: search and select a country by name or code (ISO 3166-1 alpha-2) from the list, such as United States (US), United Kingdom (GB), or Germany (DE).
• Time of day: a start and an end time, in UTC. Each field shows an approximate local-time equivalent alongside it.
• A/B split: weighted destinations. Weights are relative, so 50/50 and 1/1 mean the same thing. Each saved split rule shows the share it takes as "% of traffic", and the New rule form previews that share for the rule you are still typing.

A/B splits

A split sends a proportion of visitors to each destination and lets you compare. Splits sit outside the ordering above: they do not compete by priority, they carry no drag handle, and they are listed in their own A/B split group under the conditional rules. A click reaches them only after every conditional rule has been tried and none matched.

1. Change one thing between the variants. If the pages differ in three ways, a difference in the results tells you nothing about which one mattered.
2. Decide up front how long you will run it and do not stop early. Early numbers on small samples are noise, and stopping the moment one variant looks ahead is the most common way to reach a confident wrong answer.
3. Know which of the two behaviours below your workspace is getting before you read the result. Whether a returning visitor stays on the variant they were first shown depends on a workspace setting, and it changes what the numbers mean.

When the workspace has Audience analytics switched on, a split is sticky per visitor: the choice is computed from the visitor's own first-party cookie together with the link, so the same browser keeps the same variant on that link every time it clicks. The cookie lasts a year, so the identity holds for that long, but it is a browser, not a person: the same reader on a second device, in a private window, or after clearing cookies is a new visitor and is drawn again. The link is part of the calculation on purpose, so a visitor who lands on variant A here is not pinned to the same side of every other split you run.

Without that setting there is no identity to be sticky to, and every click is an independent weighted draw, so someone who opens the link twice can land on both variants, and a comparison you read as "people" is really "clicks". The same fallback applies to any click with no visitor: traffic we classify as a bot, and a visitor whose browser refuses the cookie. Stickiness also has a limit worth knowing before you touch a running experiment: increasing a variant's weight leaves the visitors already assigned to it where they are, but reducing a weight, or adding and removing variants, reshuffles some existing visitors onto a different destination.

Testing, pausing, and removing rules

You do not need a real phone to check what a rule will do. Test these rules, beside the rule list, runs the same resolution the redirect engine runs: pick a Device, a Country, and a Time (UTC), then Run simulation. It reports the destination the click resolves to and which rule matched, or tells you no rule matched and the default destination decided. A rule you are still typing into the New rule form joins the simulation as soon as its destination URL is valid, so you can try it before saving it. A split is a fresh random draw on every run, so repeat runs on the same input can differ. The simulator has no visitor of its own, so it does not take the sticky path a real click can.

Each rule row carries a pause control. A paused rule keeps its settings, its destination, and its place in the order, and the redirect engine ignores it completely, so pausing is how you take a rule out of service without losing how it was built. Paused rules are marked Paused, and still count toward the Rules meter. Select several rows and the toolbar that appears offers Resume, Pause, and Remove for all of them at once.

What your plan allows

Smart routing is a Pro feature, and its cap is counted per link rather than per workspace. Every link carries its own allowance, so adding rules to one link does not use up another link's room.

Smart routing rules / link, by plan

Free Not included · Starter Not included · Pro 20 · Business 50

Paused rules count towards that allowance too: it counts every rule on the link, not only the active ones. If a link is at its cap, remove a rule you no longer need, or split the intent across a second link.

Step by step