Advanced analytics: moving beyond simple click counts

Link analytics is the practice of turning "somebody followed this link" into a decision. The raw click count is the least useful part of it: on its own it has no baseline, no composition and no comparison, so it can go up for reasons that mean nothing and down for reasons that matter enormously.

What distorts a click count

Before any breakdown is worth reading, it is worth being precise about what a click is. It is an HTTP request that reached the redirect. Several things produce that request without a person deciding anything.

• Bots and crawlers, which follow links continuously. Detected bot traffic is reported as its own line rather than folded into your total, but nobody catches all of them.
• Link-preview prefetching. Paste a link into a chat app and the app fetches it to build the preview card — one click nobody made, sometimes several.
• Privacy blockers, which remove clicks you genuinely got, in the opposite direction.
• Corporate scanners, which follow every link in inbound email before delivering it.

The breakdowns, and what each is good for

Referrer

Where the click came from. The most decision-relevant breakdown, because it tells you which of your placements is doing the work. Expect a large "direct" bucket: it is not a category, it is the absence of one — links opened from apps, from QR codes, from copy-paste and from anywhere the referrer was stripped.

Device and browser

Useful mainly as a check on the destination. A campaign with 80% mobile clicks and a landing page that is awkward on a phone has a problem the click count will never reveal. This is also where device-targeted destinations earn their place — one link, different pages by device. Campaign tagging is the other half of this picture, and UTM parameters are how the referrer breakdown gets names you recognise.

Time

The shape over the first 48 hours tells you whether a link had a spike or a slope. A spike means one placement carried it; a slope means it is being found repeatedly. Those two want completely different follow-ups, and the totals for both can be identical.

Location

Coarse and labelled as an estimate, because Trizlink does not store visitor IP addresses. It is good enough to notice that a campaign aimed at one market is being read in another, and not good enough for anything finer than that. We would rather say so than present a precise-looking number that is not.

The comparisons worth making

1. The same offer on two surfaces — a bio page item against a scheduled post — which only means anything if both are measured by one system on one timeline.
2. Two variants of one link, split by weight, where the only difference is the destination.
3. This campaign against the last one, at the same number of days since launch rather than at the same calendar date.
4. Clicks against what happened afterwards on your own site. The gap between the two is the real conversion question.

The first of those is the reason the three surfaces share a workspace at all: comparing a bio-page click with a post click across two products means comparing two different definitions of a click, and the answer is not meaningful no matter how carefully you build the spreadsheet.

Splitting a link, and reading the result honestly

A weighted A/B split sends a share of traffic to each of two destinations from one link, and keeps a returning visitor on the variant they first got. It is the cleanest comparison available to a link platform, because the only thing that differs between the two groups is the page they land on.

Two cautions apply, and they are the same two that apply to every split test. Small numbers say nothing: a 60/40 difference on ninety clicks is noise, and calling it early is how teams convince themselves of things that are not true. And stopping the moment one variant is ahead guarantees you stop on noise, because at some point in almost every test one side is ahead.

What day one actually looks like

A link you made an hour ago has no history and nothing to say. This is worth stating because the empty state is where most people first decide whether analytics are working, and an empty chart is the correct answer rather than a failure. Give a link a fortnight before drawing conclusions from its shape.

Analytics history is 30 days on Free, 12 months on Pro and 24 months on Team, and export to CSV is available on every plan including Free — so the history you can read in the product is a separate question from the history you can keep. The limits are listed with the plans.