Measuring live shopping ROI beyond concurrent viewers

Why concurrent viewers misleads

It is a peak, not a total. Peak concurrency tells you how many people were watching at the busiest moment. It says nothing about how many different people saw the stream or how long they stayed. Two streams with the same peak can have very different unique audiences.

It rewards the wrong tactics. Anything that spikes attention (a giveaway, a celebrity cameo, paid promotion to a cold audience) lifts the peak. Whether any of that attention becomes orders is a separate question, and the one that pays for the stream.

It is not comparable across formats. A 20-minute product drop and a 90-minute variety stream will produce different peaks for reasons unrelated to how well they sold. McKinsey's guidance on live commerce recommends tracking views alongside conversion rates and best-selling products, and that combination is the minimum.

The metric stack that decides the next stream

Tracking it without a data team

One UTM campaign per stream. Every link shown or pinned in the stream should carry utm_source for the platform, utm_medium set to something consistent such as live, and utm_campaign unique to that stream (live-2026-10-03-autumn-edit, for example). Google Analytics lists utm_content as the parameter for differentiating creatives, which makes it a good place for the product or segment. Google's URL builder documentation covers the full parameter set.

Shopify's marketing reports read those tags. The sales attributed to marketing and sessions attributed to marketing reports in Shopify group orders and sessions by UTM campaign, so a consistently tagged stream shows up as its own line. Check which attribution model you are reading. Shopify's marketing reports can show last click, first click or last non-direct click, and last click undercounts anyone who watched, left, and came back later through email or search.

A stream-only discount code as a backstop. Codes catch buyers who type the URL or come back through search. They overlap with UTM attribution, so count each order once. If the stream sells through a platform's native checkout (TikTok Shop, for instance), take orders from that platform's report and do not add them twice.

Replays need their own tags. When the recording or clips from it go onto product pages, measure them as a separate source so replay revenue is not confused with the live window. A post-level attribution setup, or distinct UTMs on each clip's links, does the job.

A worked example

The numbers below are a hypothetical 45-minute stream for a homeware brand, invented to show the arithmetic. They are not a benchmark.

Three things fall out of the example. The replay added almost half as much again as the live window, so a same-day report would have understated the stream. Returns took out more than a tenth of live revenue, which is why the 30-day pass matters. And the peak of 310 does not appear in any calculation that decides whether the stream paid for itself.

Separate the lift from sales you would have made anyway

Stream-attributed orders are not all incremental. Loyal customers who would have bought the restock anyway will happily buy it during the stream. The simplest check is a baseline: sales of the featured products over the same days and hours in a comparable previous period, compared with the stream period. Better still, hold back promotion to a random part of your email list and compare purchase rates. The method is the same one used for holdout testing UGC.

The three-pass report

Keep the decision rule written down before the stream: for example, repeat the format if contribution is positive at 30 days and live conversion is at or above the previous stream. The broader framework for looking back on any campaign is in the post-campaign teardown template, and the case for building replays into the plan is in live shopping vs on-demand shoppable video.