---
title: Audience and breakdowns
description: Every way Dozenfold splits sessions, issues, journeys, pages and performance by shopper, device, connection, page and acquisition, and how each is counted.
updated: '2026-10-06'
slug: audience
---

# Audience and breakdowns

Dozenfold records a small set of anonymous context with each session: where it landed and came from,
the device and browser, the country, and on Chromium browsers the connection and device memory. The
same context splits every part of the product, so a question that starts on Overview can follow the
same segment into an issue, a journey step, a page or a Core Web Vital.

## The dimensions

| Dimension | Values | Source |
| --- | --- | --- |
| Landing page | Page group where the session started | First page view |
| Channel | Paid search, organic search, paid social, social, email, affiliate, display, referral, direct | Referrer and UTM parameters |
| Campaign | The `utm_campaign` value | Landing URL |
| Country | Country of the visitor | Edge location, not stored IP |
| Device | Mobile, desktop, tablet | User agent |
| Browser and OS | Browser family and operating system | User agent |
| Viewport | Coarse width band | Browser |
| Connection | slow-2g, 2g, 3g, 4g | Network Information API (Chromium only) |
| Device memory | 1 GB or less, 2, 4, 8 GB or more | Device Memory API (Chromium only) |
| In-app browser | Instagram, Facebook, TikTok, Pinterest, Snapchat or none | User agent |
| Page group and URL | Normalized page family and privacy-safe path | Page view |

Safari and Firefox do not report connection or device memory, so their sessions appear as
**Unknown** in those two breakdowns. Unknown is a real share of traffic, not missing data.

## Where each breakdown appears

| Place | Breakdowns | Counted as |
| --- | --- | --- |
| **Overview, audience** | Landing page, channels, campaigns, countries, devices, browsers, OS | Sessions, each once |
| **Overview, shopper segments** | Country, connection, device memory | Purchase conversion per segment |
| **Issue detail** | Browser, device, OS, viewport, country, connection, device memory, page group, URL, channel, campaign | Occurrences, affected sessions and occurrence share |
| **Issue detail, origin** | In-app browser, offline at failure, browser-extension code | Share of the issue's occurrences |
| **Journey step** | Device, browser, channel | Share among drop-offs vs share at entry |
| **Page detail** | Environment segments and channels | The page group's own value as the baseline |
| **Performance** | Browser, OS, viewport, connection, device memory | p75 of LCP, INP and CLS |

Rows in the Overview audience open the matching sessions. Rows on an issue keep the issue in focus
as you switch dimensions.

## Read a share correctly

- **Occurrence share** describes an issue's composition: 68% of its occurrences on Mobile Safari.
  It is not a failure rate. Mobile Safari may simply be 68% of the store's traffic.
- **Drop-off vs entry share** on a journey step is the comparison that matters: a segment that is
  20% of the shoppers entering the step but 45% of those who left it is over-represented.
- **Page baselines** compare a segment with the same page's overall value, so a product page is
  not judged against checkout.

## Shopper segments

Overview names the country, connection type or device-memory band that converts clearly worse than
the rest of its dimension. It does so only when the segment and the rest each have at least 200
sessions, the gap is at least 25%, and a two-proportion test separates it from noise. Until then
the card says it is still collecting.

## Bots are not in the audience

Search-engine renderers, headless browsers and synthetic monitors are not collected, so they never
appear as a segment, count as a session or inflate an issue. See
[Privacy and data collection](/docs/privacy-and-data#automated-traffic).

## Limits

A large segment is not a cause. Use breakdowns to choose a reproduction environment or a page to
inspect, then confirm with the occurrence, session timeline or replay. Campaign and channel describe
how a session arrived; they do not show that a campaign caused an error.