These are the terms that come up when choosing an analytics tool, defined in one sentence each and then explained. They are here because most of the confusion in this category is vocabulary: "cookieless", "unique visitor" and "sampling" all mean something specific, and two tools using the same word differently is the usual reason their numbers disagree.
Cookieless analytics
Cookieless analytics is web analytics that identifies returning visitors without storing a cookie on their device. Because the ePrivacy Directive attaches its consent requirement to storing or reading information on a device, a tool that stores nothing generally needs no consent banner in front of it.
Consent banner
A consent banner is the dialogue that asks a visitor to accept cookies before a site may store them. It is required by the ePrivacy Directive rather than by the GDPR, and it is required because of the storage rather than because of the analytics. That is why a tool which sets no cookies does not need one.
ePrivacy Directive Article 5(3)
Article 5(3) of the ePrivacy Directive is the rule that requires consent before storing information on, or reading information from, a user's device. Every cookie banner in the EU, and in the UK through retained law, traces back to this one clause. Two exemptions exist. Storage strictly necessary to deliver a service the user asked for, and storage necessary to transmit a communication.
Session
A session is one continuous period of activity by one visitor. Almost every analytics tool ends one after thirty minutes of inactivity, and again at midnight. Where they part company is on how a session starts and who it belongs to, which is the main reason two tools on the same page never report the same number.
Server-side session resolution
Server-side session resolution means the analytics server decides which requests belong to the same visitor, rather than the browser tracking it in a cookie or a tab-scoped value. Its practical effect is that one person with three tabs open counts as one visitor instead of three.
Unique visitor
A unique visitor is one person counted once in a reporting period, however many pages or sessions they produced. No number in analytics travels worse between tools. Two of them will disagree on it by a wide margin without either being wrong, because they decide who counts as a person differently.
Sampling
Sampling is when an analytics tool computes a report from part of the data and scales the result up. Large reports get fast, small numbers get unreliable, and the same period can hand you two different figures inside one tool. DataSpice does not sample.
Thresholding
Thresholding is the suppression of rows below a minimum volume, usually presented as a privacy measure. On a small site the effect is perverse: the segments most worth investigating are exactly the ones it hides.
Tracked event
A tracked event is one recorded action, most often a pageview, and it is the unit DataSpice pricing is based on. Pricing on events rather than on seats means the bill follows the size of your audience rather than the size of your team.
Referrer
A referrer is the address a visitor arrived from, sent by their browser. Every acquisition report is built on it, and it goes missing constantly. AI assistant apps strip it, privacy settings strip it, and so does any redirect that was not written to pass it along.
Direct traffic
Direct traffic is every visit that arrived with no referrer. Calling it a source is generous, since what it really records is the absence of one. Traffic from AI assistants, email clients, messaging apps and bookmarks all end up in there together.
AI-assistant referral
An AI-assistant referral is a visit sent by ChatGPT, Perplexity, Claude or a similar assistant, identified by the hostname in the referrer. Where the assistant strips the referrer, which is common in desktop and mobile app clients, the visit is genuinely unattributable by any analytics tool and lands in direct.
Revenue attribution
Revenue attribution is joining a paying customer back to the acquisition source that produced them. It matters because ranking sources by visits and ranking them by revenue usually produce different orders, and only the second one should decide where effort goes.
Last-click attribution
Last-click attribution credits the whole of a sale to the last source a customer touched before paying. It is the default in most tools, and it systematically over-credits whatever sits closest to the purchase, usually a search on your own brand name. Whatever made the person look for you in the first place gets nothing.
First-party analytics
First-party analytics is measurement collected by the site owner on their own domain, rather than by a third-party network that also sees other sites. It has no cross-site profile to build and is not subject to third-party cookie blocking.
GEO (generative engine optimisation)
Generative engine optimisation is the practice of making a page accurate and extractable for AI assistants rather than only for search rankings. In practice it means server-rendered HTML, headings phrased as questions, definitional sentences, real tables, and stating limitations. Assistants weight sources that say where they stop.