dozenfold Docs

Read the Overview

Overview is the starting point for deciding which storefront problem deserves attention. Revenue priorities lead the investigation, with purchase progression, error activity, page performance and audience context helping explain where to look next.

Choose a window and an interval

Every page opens on the last 24 hours. The window picker offers the last 24 hours, 7, 14 and 30 days, and the choice follows you as you move between pages. The selected period changes the sessions and measurements used in the page. It is not a forecast horizon. Revenue impact always compares at least seven days of sessions, whatever the window; see Revenue impact.

Charts over time share one interval control: Auto, 15 min, Hourly and Daily. Auto names what it picked, for example “Auto · hourly”. An interval a chart cannot draw stays visible but disabled, with its reason: 15-minute points cover up to 24 hours, hourly points up to 14 days. Comparisons with a previous period appear when the corresponding evidence is available; an unavailable comparison is different from no change.

The summary separates four ideas:

  • Estimated revenue impact belongs to the priority issue and is shown per day, not a sum of all issues. It is labelled measured or directional depending on the certainty of the comparison. Open the linked issue to inspect the calculation. Overlapping issue estimates must not be added together.
  • Purchase conversion describes observed store-visit-to-purchase completion. Its change is shown in percentage points when a comparable previous period is available.
  • Sessions with errors counts each affected session once in the selected period, even when it encountered multiple errors.
  • Issues to investigate describes the issues in the current view. The number with a supported revenue estimate is separate from the issue count.

An estimate can rise or fall as a wider window changes the compared sessions, purchase gap and reference orders. It is not a cumulative transaction ledger. Read Revenue impact for how the comparison, its 95% interval and the amount work.

Follow the journey

The funnel shows progression from store visit through product, cart, checkout and purchase. Follow linked issues and sessions to inspect the supporting journey rather than treating a drop-off as proof of a technical failure.

Estimated funnel reach and directly observed sampled sessions are shown separately. They describe different views of the same traffic and should not be added together.

Know the audience

The audience section ranks what the selected window’s sessions had in common, each counted once per session:

TileShows
Landing pageThe page group where sessions started
ChannelsPaid and organic search, paid and organic social, email, affiliate, display, referral and direct
CampaignsThe utm_campaign the session arrived with
CountriesThe visitor’s country, with conversion where enough sessions exist
DevicesMobile, desktop and tablet
BrowsersBrowser family
OSOperating system

Every row opens the matching sessions, so “TikTok sessions on iOS” is one click from their timelines. The same dimensions break down issues, journeys, pages and performance; see Audience and breakdowns.

Shopper segments

Experience signals include a shopper-segment comparison. Dozenfold compares completed checkouts by country, connection type and device-memory band, and names the segment that converts clearly worse than the rest of its dimension, with both rates.

A segment is named only when it and the rest of its dimension 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 collecting. Connection type and device memory come from Chromium browsers only.

Move from a priority to its evidence

Needs attention puts issues with the strongest evidence first: measured losses by missed orders per day, then measured next-stage losses, then directional signals by affected sessions, then the rest. Rising activity is marked red, falling green and new issues amber. A recurring failure can remain worth investigating while financial evidence is still developing. Experience signals add performance and interaction findings without forcing a revenue amount onto each observation.

Use the linked issue, page or performance view for detail. Acquisition, browser and device distributions help choose a reproduction environment; the largest group alone does not prove a cause.

Search documentation

Start typing to search all documentation.