Adobe CJA: 8 Customer Journey Analytics Examples

In Adobe Customer Journey Analytics (CJA), eight analyses answer most of the questions marketing, ecommerce and service teams ask: cross-channel fallout, flow, multi-entry journeys, cohorts, attribution, segment and audience comparison, anomalies with alerts, and Adobe Journey Optimizer (AJO) journey measurement. You build all of them in Analysis Workspace on top of a data view, and all of them depend on a sound connection and on identity stitching.

If you're still evaluating the tool, start with our guide to what CJA is. This post is the hands-on part: what each analysis answers, how to build it and where teams go wrong. Everything we say about features comes from Adobe's Experience League documentation, reviewed in November 2026.

The 8 analyses at a glance

The minimum setup for these analyses to work

When an analysis returns strange numbers, Workspace is rarely the problem. The cause is usually in one of these four layers.

Connection: datasets and a shared person ID

A connection brings together Adobe Experience Platform (AEP) event, profile, lookup and summary datasets. CJA combines all event datasets into one: fields with the same schema path become a single column, each dataset's person ID merges into one column even if the fields have different names, and rows are ordered by timestamp.

That's why a funnel spanning web, app and call center only works if the sources share an identifier in the same format, such as a customer ID, hashed email or phone number. Adobe also recommends a field in each dataset that records the event's source.

Data view and component settings

The data view defines which dimensions and metrics exist and how they behave. Its settings are retroactive and non-destructive, and one connection can have several data views. What matters most here:

• Persistence. Lets a dimension value carry over to later events. It combines an allocation (Most recent, Original, All, First known or Last known) with an expiration (session by default, person, custom time or a metric), up to 90 days.
• Derived fields. Transform data with a rule builder without touching the schema: fix page names, or build a marketing channel dimension with the Marketing channels template.
• Session. Adobe suggests 30 minutes for online behavior and a long timeout (three months in its example) to connect purchases with later call center contacts.
• Containers. Person, session and event. Most visualizations ask you to pick one, and the answer changes with your choice.

In CJA they're called segments: until April 2025 the interface called them "filters," and Adobe aligned the name with Adobe Analytics.

Identity stitching

Stitching rekeys a dataset's anonymous events to the person ID when it can find one; otherwise it keeps the device's persistent ID.

Limits: it only works on event datasets, it's case-sensitive (Bob and BOB are two people), it works within one dataset rather than merging datasets, and today you enable it from the Connections interface. Adobe asks you to confirm you have the end-user permissions the law requires first; our post on AEP data governance covers identity and consent.

Data view time zone and calendar

Each data view has its own time zone and calendar (Gregorian, 4-5-4 retail or custom). If you pick a zone that observes daylight saving time, CJA adjusts the data: a one-hour gap appears in spring and one hour repeats in fall. Match the time zone to where the business closes its day: a sale at 11:30 p.m. lands on a different day otherwise. If you run Mexico alongside the US or Canada, note that Mexico City has been on UTC-6 year-round since 2022, and create a separate data view per reporting time zone on the same connection.

The 8 analyses, step by step

Each analysis follows the same structure: question, data, how to build it, how to read it and the mistake we see most.

1. Cross-channel fallout (Fallout)

Business question. At which step do buyers drop off, and how many pick it back up by calling the contact center?

Data and components. Web and app checkout events, plus a call dataset with the same person ID. For calls, Adobe suggests a metric based on the call reason field, deduplicated to count once per session.

How to build it in Workspace:

1. Add a Fallout visualization.
2. Drag the steps in order: product view, add to cart, checkout start and purchase. They can be dimensions, dimension items, metrics or segments.
3. Add the calls metric as another step; mixing dimensions and metrics is what Adobe calls inter-dimensional fallout.
4. Pick the container: Person (default, spans sessions) or Session.
5. Under each step, choose Eventual path (gets there at some point) or Next event (on the very next event).
6. Right-click a step and select Break down fallout at this touchpoint to see what people who dropped off did right after.

How to read it. The numbers in gray are the drop-off from the previous step, not the cumulative loss.

Common mistake. Reading drop-off as session abandonment with the Person container: someone who abandoned on Monday and bought on Thursday counts as a conversion.

2. Paths before and after a key moment (Flow)

Business question. What do people do right before calling support, canceling or abandoning the cart?

Data and components. A path dimension. To tie calls to the last page viewed, Adobe suggests a page dimension with last-touch allocation and session persistence.

How to build it in Workspace:

1. Add a Flow visualization.
2. Anchor it with Starts with, Contains or Ends with; for "what happens before a call," use Ends with and the calls metric.
3. If you anchor on a metric, choose the pathing dimension (Page by default).
4. In advanced settings, decide whether to keep repeats such as page reloads (Include repeat instances, off by default). You get up to 5 columns and 10 items per column.
5. Select a point in the path to create a segment of the people who went through it.

How to read it. Flow counts occurrences (path views), not people. Use it to discover paths; to measure a fixed sequence, use Fallout.

Common mistake. Presenting Flow numbers as customers, or trying to anchor on a calculated metric: Starts with and Ends with don't accept them.

3. Multi-entry journeys (Journey canvas)

Business question. Which path converts best when people arrive from email, search or the app, and end up buying online or picking up in store?

Data and components. Events, dimension items, segments and date ranges as nodes. The primary metric can be anything, including calculated metrics, and you can add a secondary one.

How to build it in Workspace:

1. Add a Journey canvas visualization.
2. Choose the primary metric (people) and, optionally, a secondary one (revenue).
3. Leave the AJO journey field empty to start from a blank canvas, and pick the container.
4. Drag components onto the canvas as nodes and connect them with arrows; you can have several starts, several ends and crossing paths.
5. Apply a breakdown to a node, or create an audience from a node, an arrow or the fallout.

How to read it. A caption at the top shows the path with the highest conversion rate. Percentages can be based on everyone in the data view, everyone who started the journey or the previous node.

Common mistake. Reading an arrow as "the immediate next step." Nodes connect as an eventual path: people count even if they did other things in between, within the chosen container.

4. Retention and repeat-purchase cohorts (Cohort table)

Business question. Of the people who bought for the first time in a given month, how many come back in the following months, and does it vary by acquisition channel?

Data and components. An inclusion metric (orders) and a return metric. Each criterion accepts up to 10 segments and 3 metrics.

How to build it in Workspace:

1. Add a Cohort table.
2. Inclusion: orders, with a numeric condition if needed. Return: the metric that defines "came back."
3. Choose the granularity (day through year) and the type: Retention (default) or Churn.
4. Optional: Rolling calculation; Latency table, to see behavior before and after an event, or Custom dimension cohort, to build cohorts by channel, campaign or region. The last two can't be combined.
5. Select cells and use Create Segment From Selection to work with that group.

How to read it. Each row is a cohort and each column is a period after inclusion.

Common mistake. Reporting "retention" without stating the calculation. The standard one counts people who return in a period even if they skipped the previous one; rolling counts only people who returned in every prior period. They aren't comparable.

5. Multi-channel attribution (Attribution panel)

Business question. Which channel deserves credit for a sale, and does the answer change with the model?

Data and components. A success metric and a channel dimension. With Web SDK data, the marketing channel is usually built as a derived field.

How to build it in Workspace:

1. Create an Attribution panel with the success metric and the channel dimension.
2. Pick the models: Last Touch, First Touch, Linear, Participation, Same Touch, U Shaped, J Curve, Inverse J, Time Decay, Custom or Algorithmic.
3. Choose the container and lookback window: the reporting window, 14, 30, 60 or 90 days, or custom. With the session container, the window is locked to the reporting window.
4. Build the panel.

How to read it. You get a bar and table comparison by model, an overlap view of the top three channels, a scatter of up to three models, a trend and a channel flow. If a conversion had a single touchpoint, that channel gets 100% under every model: differences only appear with multiple touchpoints. Algorithmic is based on the Harsanyi dividend, a generalization of the Shapley value.

Common mistake. Setting an allocation in the dimension's persistence and a different model on the metric, expecting them to stack. Adobe documents that a non-default attribution model ignores the dimension's allocation.

6. Segment and audience comparison

Business question. How do app customers differ from web-only customers, or loyalty members from everyone else?

Data and components. CJA segments, or Adobe Real-Time CDP audiences brought in with Audience analysis, which turns audience membership into dimensions.

How to build it in Workspace:

1. Build the segments with a person container.
2. In a Fallout, drag both segments to the top to compare them side by side; remove "All People" to show only those two.
3. In a freeform table, put the segments in columns against conversion or average order value, and add a Venn to see the overlap.
4. For Real-Time CDP audiences, set up Audience analysis with the sandbox and merge policy. The Audience overview template compares up to three audiences in a Venn.
5. In the other direction, publish a CJA segment as an audience to Real-Time CDP and use it in AJO.

How to read it. Compare rates, not volumes. On the platform side, our Real-Time CDP post explains how audiences are built.

Common mistake. Looking for Adobe Analytics' Segment Comparison panel: CJA doesn't have it. And changing the date range in Audience analysis: membership is reprocessed nightly and is only accurate for the previous day.

7. Anomaly detection and alerts

Business question. Conversion dipped on Tuesday: real drop or normal variation? How do you find out in time?

Data and components. A time series with history. At daily granularity, the model compares against up to 35 prior days and the same period one year earlier.

How to build it in Workspace:

1. In a daily freeform table, turn on Show anomalies in the column settings; in a line chart, in the visualization settings (it's off by default).
2. For an alert, select rows, right-click and choose Create alert from selection. Set the granularity, delay, metrics and threshold; alerts arrive by email or SMS with a link to a project.
3. Alerts on anomaly thresholds (Intelligent Alerts) depend on your CJA package.

How to read it. The shaded band is the expected range; a white dot outside it is an anomaly.

Common mistake. Assuming the model knows every sales event. The holidays it adjusts for are Memorial Day, July 4, Thanksgiving, Black Friday, Cyber Monday, December 24-26, December 31 and January 1. Anything else, such as Mexico's El Buen Fin or your own promotions, needs an annotation. Also avoid hourly alerts on batch-loaded data such as call center files. And if you're coming from Adobe Analytics: CJA has no Contribution analysis, so break the metric down by dimensions in a freeform table instead.

8. Adobe Journey Optimizer journey measurement

Business question. At which step do people leave a journey, and what do they do next on the site, in the app or in store?

Data and components. The AJO datasets: Journey Step Events (who went through each node), message feedback (sends, bounces), email tracking (opens, clicks, unsubscribes) and push tracking, and the AJO Entity Dataset, with journey and campaign metadata.

How to build it in Workspace:

1. When you enable CJA as AJO's reporting engine, a connection (AJO Enabled Connection) and a data view are created per sandbox, with components tagged "(AJO)."
2. To connect journeys with purchases, add those datasets to your main connection with the same person ID; the automatic one only holds AJO data.
3. In a Journey canvas, select the journey (Live, Stopped or Finished) and it renders with the same structure as in AJO. You can also open it from AJO.
4. Add a secondary metric such as revenue, and break down nodes by channel or segment.

How to read it. Per Adobe, discrepancies between AJO and CJA are typically under 1-2%, and larger for the last two hours of data.

Common mistake. Comparing today's numbers: use date ranges that exclude today. To design the journeys themselves, see our AJO journey examples.

CJA B2B Edition and Content Analytics

• CJA B2B Edition adds account-based reporting, with account, global account, opportunity and buying group containers in cohorts, flows and attribution, plus lookback windows of up to 13 months. It fits when you sell to companies with long cycles and multiple decision makers.
• Content Analytics measures how content drives your KPIs. An AI service describes images and experiences with attributes (photography style, colors, persuasion strategy) and recognizes the same asset wherever it appears. It uses AEP Edge Network data and paid media data, such as Google and Meta. Adobe notes its AI output should be reviewed.

Example: an omnichannel retailer during El Buen Fin 2026

A hypothetical case, with no figures: a retailer selling in Mexico through ecommerce, an app, physical stores and a call center prepares for El Buen Fin, Mexico's annual nationwide sales event, running November 13-17, 2026. The same playbook applies to Black Friday week in the US or Canada.

Before:

1. A data view in Mexico City time, a saved date range for the five days and an El Buen Fin annotation.
2. A Cohort table with "orders during last year's event" as inclusion shows how many of those buyers came back; a Latency table shows how they bought before and after.

During:

3. A checkout Fallout with app and web segments side by side, in a session container, to see where payment breaks.
4. Daily anomaly detection on orders, knowing El Buen Fin isn't on the model's holiday list, and daily rather than hourly alerts if store data arrives in batches.
5. A Flow with Ends with "calls" to see what people viewed before calling.

After:

6. A Journey canvas for the store pickup path: online order, AJO notification and pickup.
7. An Attribution panel with a person container and a 30-day window, comparing Last Touch, Linear and Algorithmic.
8. Loyalty members versus non-members: average order value and repeat purchases in the following weeks.

Common mistakes when analyzing in CJA

1. Expecting numbers to match Adobe Analytics. CJA counts people by the connection's person ID and defines sessions and persistence in the data view; if those rules differ from your report suite, the numbers won't match. We cover this in our guide to migrating from Adobe Analytics to CJA, and this post covers ecommerce measurement in Adobe Analytics.
2. Not validating stitching. Adobe suggests a Venn with two segments (people with web data and people with call center data) and the People metric: the overlap is what actually stitched.
3. Different ID formats across sources. An uppercase email in the CRM and a lowercase one on the web are two people.
4. One data view for everyone. Marketing, ecommerce and a team in another time zone need different sessions, attribution or time zones.
5. 30-minute sessions for offline data. A short session splits the journey between the purchase and the call.
6. Forcing upstream transformations into Workspace. If a calculation doesn't fit in a derived field, prepare it in AEP; Adobe names Data Prep and Adobe Data Distiller as alternatives.

How we work at WolfSellers

At WolfSellers we start with the business questions and then pick the analysis. Those questions define which datasets go into the connection, which identifier links the sources and how many data views you need, and we validate stitching before the first Workspace project. We do this through our Adobe Analytics and Adobe Experience Platform services, and with data engineering when ERP, CRM or call center data needs preparation before it reaches AEP.

Frequently asked questions

What analyses can you run in Customer Journey Analytics?

The most common are fallout, flow, multi-entry journeys (Journey canvas), cohorts, attribution, segment and audience comparison, anomalies with alerts and AJO journey measurement, all in Analysis Workspace.

What's the difference between Fallout, Flow and Journey canvas?

Fallout measures a defined sequence with a single entry point and compares two segments side by side. Flow explores paths from or to a point and counts occurrences. Journey canvas supports multiple entry points and paths, any metric, and opens AJO journeys.

Does CJA have contribution analysis like Adobe Analytics?

No. Adobe's feature support page lists it as not yet planned for CJA, and the Segment Comparison panel isn't supported either. CJA does have anomaly detection, forecasting and alerts.

Do I need identity stitching for cross-channel analysis?

Only if your datasets don't carry the person ID on every event, which is typical for web and app, where many visitors browse without logging in. If every event already has a common identifier, the connection is enough.

Can I analyze Real-Time CDP audiences in CJA?

Yes, with Audience analysis, which turns audience membership into dimensions; it's reprocessed nightly and accurate for the previous day. You can also publish CJA segments as audiences to Real-Time CDP.

Which time zone should the data view use?

The one where your business closes its day. If teams report in different time zones, such as Mexico City (UTC-6 year-round) and US Eastern, create one data view per time zone on the same connection.

Related services

If this topic is relevant to your business, these services from WolfSellers can help you implement it:

• Adobe Experience Platform
• eCommerce Marketing
• AI & Automation