How CommerceGeek measures ecommerce technology
Our data stories and guides use two kinds of figures: what we find on the stores in our index, and how fast those stores are for real Chrome users. Here is where each comes from, how we check it, and what it can’t tell you.
The stores
We keep an index of a few hundred well-known online stores, picked by hand across fashion, beauty, home, food and drink, tech accessories, pets, big retailers, kids, handbags. Most run on Shopify; some run on Salesforce Commerce Cloud, Adobe Commerce, Centra and other platforms, so we can compare.
It isn’t a random sample of the market. Every figure describes the stores we index, and our articles say so (“of the stores we index…”). We never name individual stores in data stories.
What we find on each store
About once a month, our crawler visits each store’s public pages: its robots.txt, sitemap, one listing page, one product page and its returns page when they link to one, at most 7 requests. It obeys robots.txt, never visits the cart, checkout or account pages, and stops at bot protection. How CommerceGeekBot behaves and how to opt out.
- Platform and tools (the reviews app, site search, email tool and so on) come from rules that match the scripts, headers and markup each tool is known to use. A returns platform can also show itself through a link to the store's returns portal. Every match keeps its evidence, such as the script address.
- Page features (a size guide, sticky add to cart, buy now, pay later messaging, free returns or exchanges…) come from the same kind of rules, plus an AI model that reads a summary of the page. The model must quote the text or markup that shows each feature, and we check the quote is really on the page. No quote, no feature.
We read the HTML the store sends, not what JavaScript adds afterwards. Anything that only appears after the page runs its scripts can be missed, so treat tool and feature counts as lower bounds: “at least this many”.
How we check it’s right
Before any figure from our detections is published, we check a sample by hand. On 15 October 2026 we checked 30 stores across every platform against the browser’s developer tools:
| What | Correct | We require |
|---|---|---|
| Platform | 24 of 24 | at least 95% |
| Tools | 138 of 138 | at least 90% |
| Page features | 108 of 108 | at least 80% |
11 items we couldn’t confirm either way were left out, mostly very large stores that run their own custom platforms. We repeat the check every quarter, and if a check fails, we stop publishing those figures until the detections are fixed.
Speed and devices
Core Web Vitals and the share of visits on phones come from Google’s Chrome UX Report: real visits by Chrome users who share usage statistics, for each store’s whole site, over a calendar month. We read its public BigQuery dataset.
- Each metric is the 75th percentile: three quarters of visits were at least that fast.
- A store passes Core Web Vitals when all three are “good” by Google’s thresholds: LCP (loading) 2.5 seconds or less, INP (responsiveness) 200 milliseconds or less, and CLS (layout shift) 0.1 or less. Google’s Web Vitals guide.
- Stores with too little Chrome traffic aren’t in the report, and other browsers aren’t counted.
How we report figures
- We never report a group of fewer than 8 stores, so no figure points at one store.
- Percentages are rounded to whole numbers, and every chart has its numbers as a table too.
- Each article says how many stores are behind its figures.
- Over time: at the end of each month we record what each store was on, from its latest approved scan. Changes such as one reviews app replacing another compare only stores indexed in both months, so stores joining the index never count as adoptions. When our scanner learns to spot a new tool, stores where it appears are not counted as adding it, and its chart starts from the first full month it could be seen.
How articles are written
Drafts are written with the help of AI, from our figures and from quotes taken from official docs: Shopify’s, Google’s, the W3C’s and the tools’ own. Every number in a draft is checked in code against those figures and quotes, and anything we can’t source is rejected. A person reviews, edits and publishes every article. Nothing is published automatically.
CommerceGeek isn’t affiliated with Shopify or with any of the tools we write about.