Session recording vs heatmaps: which to use when

What each one measures

A heatmap is an aggregate over many sessions, drawn onto a page. Click maps show where pointers and taps landed; scroll maps show how far down the page attention survived; a segmented view splits the same clicks by traffic source, device or campaign, which is often where the finding is.

A session recording is one visitor’s session reconstructed from page changes and interaction events. It is not video: it is a replay of the page state, which is why it is small enough to collect at volume and why it can miss anything the collector did not capture.

The important asymmetry: the heatmap has thrown away the story to gain statistical weight, and the recording has kept the story and has no weight at all. Neither loss is recoverable from the other.

Start with neither

Both are page-level tools, and the first question is which page. Answering that with a heatmap means picking pages by intuition and checking them one at a time, which is slow and biased towards the pages you already suspect.

Start with the funnel instead. Measure each step in order, count a session at a step only if it cleared the one before, and find the transition with the largest absolute loss. That names the page in minutes rather than days, and it names the one costing the most rather than the one with the worst percentage.

Then the heatmap has a job (explain this specific transition) and the recordings have a cohort. Without that, both tools are browsing.

When the heatmap is the right tool

When you suspect an element rather than a flow. Clicks concentrated on something that is not clickable (a product image expected to zoom, a spec table expected to expand, a size guide that was unwired when an app was removed) show up in a click map at once and are among the cheapest things to fix on a storefront.

When the question is about order or attention. Scroll maps settle “is anyone seeing the reviews” and “does the add-to-cart button sit below the fold on a small phone” faster than any other method.

And when you need to compare segments. A click map filtered to mobile, set against the same map on desktop, often shows that the two are effectively different pages, and that the aggregate you had been reading described neither.

When the recording is the right tool

When you know what is happening and not why. A funnel says many mobile carts do not reach checkout; a heatmap says the discount field takes disproportionate attention. Five recordings from that exact cohort will show you the mechanism: shoppers open the field, hunt for a code in another tab, and do not come back.

When behaviour is sequential. Hesitation, backtracking, repeated attempts and the order in which someone tried things are invisible in any aggregate and obvious in one replay.

And when something is outright broken. A validation error that fires on a legitimate postcode, a variant picker that silently fails on one product: these are single-session findings that no aggregate will surface, because they are rare and total.

The sample-size trap, in both directions

Watching recordings at random is the most common way to waste a week. Fifty unfiltered sessions produce a strong feeling and no reliable conclusion, because you remember the vivid ones rather than the representative ones. Five from a defined cohort will usually settle a specific hypothesis, and that is the whole discipline.

The heatmap has the opposite failure. It looks statistically respectable at any volume, and a click map built from a few dozen sessions is decoration. Check what it was built from before you believe a pattern in it.

Both share a deeper limitation: neither tells you how often a problem happens across your traffic, and neither tells you what it costs. They are diagnostic instruments, not measurement, and treating them as measurement is how a fortnight of investigation ends in a change nobody can defend.

Privacy differs between them, and it matters

Heatmaps are aggregates and carry comparatively little risk. Recordings reconstruct what a specific person did on your site, which is a different data-protection question and one your data protection officer will treat differently.

The question to put to any vendor is where redaction happens. Masking input values in the browser, before anything is transmitted, is a different guarantee from redaction on the server after the data has already left. Ask which, and ask how consent is checked before collection starts.

Liftable’s pixel asks Shopify’s Customer Privacy API first, sets no cookie and sends nothing until the shopper has consented to analytics, and records the names of form fields, never what was typed.

Where Liftable sits

Liftable has neither recordings nor heatmaps. Its pixel records the steps a shopper took (pages viewed, clicks, scroll depth), which is the raw material for the sequencing above: the funnel names the page, and the count of sessions that hit a problem says how often it happens.

The evidence sits on the opportunity that needed it, and once sessions and orders are connected each opportunity gets an estimated monthly range with its inputs shown. There is no gallery of sessions to explore.

If watching sessions or reading click maps is something you value in itself, keep a tool built for that. Liftable is not a replacement for it.

Questions