Best customer journey analytics software 2026
Need to connect shopping and support events? Compare customer journey analytics software, identity controls and Adobe pricing considerations.
- Published
- Last updated
Amplitude is the best customer journey analytics software for retail teams investigating how shopping behaviour leads to purchases and support contacts. Its Journeys reports show the routes customers take, while funnels measure completion of a chosen sequence. You can separate repeated attempts from unique customers and examine the events around a failed payment.
A small team can start with event analytics and expand its reporting as its tracking improves. Retailers with established cross-channel data pipelines can justify a more involved implementation. The tools below differ most in their identity handling, reporting controls and operating cost.
Top customer journey analytics tools for retail
1. Amplitude
Amplitude ranks first for teams that need to explore unexpected shopping paths alongside defined purchase funnels. Its Journeys chart has Pathfinder and Journey Map views. Analysts can start at a catalogue search, find branches around payment and compare completed paths with drop-offs. Amplitude’s Journeys documentation distinguishes unique users from total conversions and explains repeated-event controls.
Collect browser and app events through software development kits (SDKs), then send permitted order and support events through a server ingestion route. Amplitude combines device IDs, user IDs and its own identifier. A stable authenticated user ID lets activity on recognised devices join the customer’s history. A support record needs that same identity convention before it can explain a shopping path.
Amplitude’s funnel reports measure selected steps, while its data tables apply attribution models to event properties. First-touch and last-touch models can credit different acquisition channels for the same order. The attribution documentation also explains why summing unique-user rows can double-count people. Its persisted-property documentation lists Customer Journeys as outside that feature’s supported charts, so an attribution setting does not automatically carry into every report.
The practical limit is the tracking work behind the charts. An event named Payment submitted cannot establish that an order succeeded. Budget for server-confirmed purchases and support-event imports before interpreting drop-offs.
Free includes two million events per month. Advanced behavioural exploration appears on Growth, while Enterprise adds data access controls. Buy the reporting and permission scope your team needs, rather than assuming the free allowance includes every analysis.
2. Mixpanel
Mixpanel ranks second and suits a smaller team starting with purchase funnels and frequent event paths. Its Flows report explores events before or after a selected action, with property breakdowns for details such as payment method. Flows is available on free and paid plans, according to Mixpanel’s report documentation.
Use browser or mobile SDKs for shopping interactions and server-side events for orders and support activity. In projects using Simplified ID Merge, a device ID identifies anonymous activity. An event carrying both device and authenticated user IDs creates their mapping. The same stable user ID can connect authenticated activity across devices. Reset tracking identity after logout on a shared device, and keep support imports consistent with the project’s identity model.
Funnels answer whether customers complete a defined sequence. Attribution models use touchpoint properties, with first-touch, last-touch and linear options among the documented choices. Campaign capture needs care because persisted campaign properties can make later events retain earlier acquisition values. Mixpanel’s traffic attribution guidance explains the setting that disables that persistence. This makes Mixpanel a practical choice when the team can own a clear event specification and campaign rules.
Mixpanel’s Free plan includes one million events per month. Growth raises the published ceiling to 20 million and Enterprise uses a quote. Event ingestion still needs engineering ownership, and custom event-data retention belongs to Enterprise.
Mixpanel places below Amplitude here because Amplitude’s dedicated Journey Map adds another way to compare complete paths. A team focused on a small number of checkout questions can reasonably prefer Mixpanel’s Flows workflow.
3. Adobe Customer Journey Analytics
Adobe Customer Journey Analytics ranks third for general retail adoption and is the strongest fit here for an established Adobe Experience Platform team. It can combine online activity with call-centre and point-of-sale data in Analysis Workspace. Its connection and data-view layers let analysts define what those fields mean before building reports.
Customer Journey Analytics uses datasets ingested into Adobe Experience Platform. Connections bring selected datasets into analysis using a common person ID. Data views expose dimensions and metrics, then Workspace provides flow, fallout and attribution analysis. A retailer can inspect an order-status contact after a purchase if both datasets carry a compatible identity. Stitching can extend identity coverage where the licensed package and source records support it.
The cost is a larger implementation commitment. Data modelling, ingestion and reporting permissions need owners, and package entitlements constrain stitching and reporting capacity. Adobe supplies tailored pricing based on licensed volumes and scope. Choose it when existing Platform investment and analyst capacity justify that architecture. Adobe Analytics and Journey Optimizer are separate products, so their wider features do not establish what a Customer Journey Analytics licence includes.
From collected events to a customer path
A customer path is an ordered sequence of collected events assigned to an identifier within a reporting window. Collection supplies the actions. Reporting decides which events belong together and which sequences to count.
For each event, retain an event ID, occurrence timestamp and ingestion timestamp. Add the channel, product ID or order reference where relevant. A browser records Product viewed, while the order system supplies Purchase confirmed. A support system can send Status enquiry with its occurrence time even if the import arrives later.
The reporting pipeline validates those fields, removes duplicate records and applies the permitted identity rules. It orders the remaining events and builds the path. Missing product properties prevent a variant-level breakdown. A duplicated purchase inflates event totals, while a timestamp recorded in the wrong time zone can place a support contact before checkout.
A channel change only becomes a cross-channel path when the records share a permitted identity. Anonymous visits stay separate when that link is absent. Delayed events can alter historical paths after ingestion, so save both the reporting window and the extraction time with an analysis.
An observed path describes recorded behaviour. A manually designed journey map describes stages or an intended experience. Use customer journey mapping software when your team needs a collaborative diagram. For deciding which retail problems to fix, the ecommerce customer experience guide develops the diagnosis beyond the reporting tool.
Identity, consent and access controls
Identity controls decide whether two events describe the same customer. An anonymous browser identifier describes a browser or device. An authenticated identifier describes the account the retailer has recognised. Cross-device reporting requires evidence connecting each device to that account under the retailer’s permitted tracking rules.
For example, two people using one tablet can produce a false combined path if the first account remains attached after logout. A shared email address also fails as proof of one person. Use a stable internal account identifier and document logout behaviour. Hashing an email address does not make its tracking permission automatic.
Compare the controls you will operate
| Control | Amplitude | Mixpanel | Adobe Customer Journey Analytics |
|---|---|---|---|
| Consent | Control SDK collection and storage through the consent implementation | Client-library opt-out methods stop subsequent tracking | Platform consent reporting and optional ingest filtering use configured policy membership |
| Individual deletion | User Privacy API removes identified user data; upstream copies also need treatment | End-user export and deletion requests use distinct IDs | Platform Privacy Service requests apply to original and rekeyed datasets |
| Retention | Time to Live deletes events outside an organisation or project period | Default event-data retention is two years; custom policy is Enterprise | Connection retention and licensed historical capacity need separate configuration |
| Data location | US and EU data-centre options | US, EU and India residency routes | Provisioned Platform region also holds ingested analytics data; Sydney is listed |
| Access and export | Enterprise data access controls; plan project permissions alongside raw export access | Organisation and project roles; export/deletion requests have restricted permissions | Product profiles restrict data views; export permissions need explicit assignment |
Amplitude’s privacy implementation guide distinguishes deletion from preventing future collection. Mixpanel makes the same distinction in its privacy documentation. Adobe’s data-governance guidance explains sensitive-field labels and export warnings.
Treat an export as another copy of customer data. Assign its owner, destination and deletion process before distributing it. Test a read-only analyst account against a restricted support dataset and retain the result. Vendor controls help you enforce your policy, but their presence alone does not establish legal compliance.
Path analysis and attribution answer different questions
Path analysis shows recorded sequences, a funnel measures completion and attribution allocates credit to earlier interactions. Use the same population and order records when comparing their answers.
Consider an illustrative cohort of 100 permitted customer identifiers that perform Catalogue search during a seven-day intake period. Follow each for seven days after its first search. Sixty view a product, 30 add to basket and 20 complete a purchase in that order. Intervening events are allowed, and each customer counts once at each funnel step.
Read the same purchase sequence three ways
| Report | Question | Denominator and repeat rule | Illustrative result |
|---|---|---|---|
| Path | Which routes follow search? | 100 starting identifiers; retain repeated payment attempts as distinct steps | One branch contains delivery-policy lookup followed by basket and payment retry |
| Funnel | How many complete the chosen sequence? | 100 first-search entrants; one completion per identifier within seven days | 20 purchases produce 20% search-to-purchase conversion |
| Attribution | Which touchpoints receive purchase credit? | 20 unique confirmed order IDs; seven-day pre-purchase lookback | First-touch and last-touch assign different credit to the same orders |
Within that cohort, suppose one customer arrives through paid search and later returns through an email campaign before buying. First-touch credits paid search. Last-touch credits email. Linear attribution splits credit between those two eligible touches. Repeat payment attempts do not create extra orders because attribution uses the confirmed order ID as its conversion record.
Adobe’s attribution overview explains report-time model comparison. Amplitude and Mixpanel also document multiple attribution models, but their report settings and units need matching before figures are compared.
Attribution does not prove that the credited interaction caused the purchase. Customers who already intend to buy can choose the same channel. A controlled experiment answers a different question about incremental effect.
Identity coverage can change the apparent winner too. If the earlier paid-search visit remains anonymous, the joined purchase path may contain only email. Shortening the lookback can remove paid search even when identity is complete. Record the included population and excluded identities beside each result.
Adobe Customer Journey Analytics evaluation and pricing
Adobe Customer Journey Analytics fits retailers that can operate an Experience Platform data pipeline and maintain shared reporting definitions. Its implementation proceeds through ingestion, connections, data views and Workspace.
- Ingest storefront, order and support datasets into Platform using suitable schemas. Preserve event time and the identity fields needed for analysis.
- Create a connection that selects those datasets and identifies the person ID. Align the namespace and value format across sources.
- Create a data view that defines reportable dimensions, metrics and session behaviour. Specify purchase counting so duplicate records do not become extra orders.
- Build Workspace projects for flow, fallout and attribution. Give analysts access to the appropriate data views and export destinations.
The Customer Journey Analytics documentation separates these layers. This architecture is practical when a retailer already has Platform engineers and analysts. A small team still deciding what Purchase means has schema work to finish before this reporting layer pays back.
Field-based and graph-based stitching
Field-based stitching uses a persistent identifier and a person identifier observed within the same event dataset. It can reassign earlier anonymous activity when the person becomes known within the configured replay window. It does not require the Platform identity graph. Adobe’s field-based procedure explains the live and replay passes.
Graph-based stitching queries Platform Identity Service for the selected person-ID namespace. It needs a populated identity graph connecting the persistent identifier to that person. An empty graph cannot establish the relationship. Adobe’s graph-based procedure also requires graph treatment during privacy deletion.
Both methods retain the persistent identifier when they cannot resolve the person. Authentication outside the selected replay window can leave earlier activity unjoined. Choose the method from your available source identities and licensed entitlement.
Compare Adobe package scope
| Entitlement | Foundation | Select | Prime | Ultimate |
|---|---|---|---|---|
| Field-based stitching | Not included | Five datasets; 13-month initial backfill | 15 datasets; 13-month initial backfill | 50 datasets; 25-month initial backfill |
| Graph-based stitching | Not included | Not included | Available within stitching allowance | Available within stitching allowance |
| Concurrent report requests per connection | Five | Six | Eight | Ten |
| Monthly report requests per organisation | 500,000 | 750,000 | 1.5 million | Five million |
| Historical capacity included beyond core period | None listed | None listed | None listed | 12 months |
Adobe’s product description licenses packages per million rows of data. Core data covers timestamps less than 13 months old. Additional historical capacity is distinct from stitching backfill. Add-ons cover ingestion, storage and reporting capacity. A larger backfill allowance alone does not buy longer retention.
Foundation also excludes omnichannel event datasets, which materially limits this retail trial.
Build a comparable Adobe quote request
Adobe Customer Journey Analytics pricing uses a tailored quote, with no public subscription figure on that page. The sales order must specify currency, taxes and the contract term. Request separate lines for licence capacity, add-ons and implementation services.
Use an illustrative retailer with ten million shopping events and 500,000 order/support events per month, plus 24 months of history. The quote request needs these inputs:
- Monthly and annual event counts, average row size and expected growth, separated by source.
- Event, profile and lookup dataset counts, with the proposed person-ID namespace.
- The historical backfill and ongoing retention requirement, including what ages out each month.
- Required stitched datasets, authentication delay and field-based or graph-based treatment.
- Analyst and report-consumer numbers, peak concurrent requests and monthly reporting demand.
- Source connectors, export destinations, hosting region, service support and implementation responsibility.
Require the proposed package to explain how it covers that exact scenario. A quote for core rows alone is incomplete when the retailer also needs older history and additional ingestion capacity.
Trial catalogue, order and support events together
Trial a complete purchase and its later support contacts in one isolated dataset. The fixture below is illustrative and specifies expected evidence for a trial.
Each row needs a unique event ID and occurrence/ingestion timestamps. Keep the analytical account token in a restricted identity mapping. Product IDs and order references remain separate fields. Exclude names, contact details and payment credentials.
Illustrative retail event fixture
| Elapsed time | Event | Product or order reference | Identity and expected evidence |
|---|---|---|---|
| Start | Catalogue search | Category: jackets | Anonymous browser A; paid-search channel |
| Two minutes | Product viewed | Product P100 | Same browser; product branch is visible |
| Four minutes | Variants compared | Variants V101 and V102 | Variant properties remain distinguishable |
| Five minutes | Delivery policy opened | Policy: delivery | Same path retains the policy lookup |
| Seven minutes | Added to basket | Variant V102, basket B10 | Basket quantity is recorded once |
| Nine minutes | Payment failed | Basket B10, attempt PA1 | Failure code excludes sensitive payment data |
| Eleven minutes | Purchase confirmed | Order O500, attempt PA2 | Permitted account U10 mapping; one order |
| Two days | Status enquiry | Order O500 | Account U10; support channel |
| Three days | Status enquiry | Order O500 | Second contact remains a separate event |
| Eight days | Return requested | Order O500, variant V102 | Purchase-to-return path requires a longer window |
Run another browser through search and product view without authentication. Its permitted anonymous events must remain a separate path. Run a consent-refusal case with collection disabled under your policy. It must not create a tracked purchase journey or be counted as observed abandonment.
Compare what survives a fault
| Fault or outcome | Amplitude trial evidence | Mixpanel trial evidence | Adobe trial evidence |
|---|---|---|---|
| Completed purchase | User event history and matching funnel completion | Identity cluster and matching funnel completion | Connected person ID and Workspace purchase metric |
| Failed payment followed by success | Journeys branch retains PA1 and PA2 | Flows branch retains failure then confirmation | Flow preserves failure before the confirmed order |
| Repeat status enquiry | Two events for O500; one purchasing customer | Two events for O500; one purchasing customer | Two support events under the connected identity |
| Duplicate confirmation | Replayed event ID leaves one confirmed order | Replayed event ID leaves one confirmed order | Ingestion/report definition leaves one confirmed order |
| Delayed support import | Occurrence time restores the intended position | Occurrence time restores the intended position | Source timestamp and processing restore the intended position |
| Anonymous path without permitted link | Browser history stays separate | Device-only identity stays separate | Persistent ID remains when person resolution fails |
These are acceptance criteria, not measured vendor results. Inject the same purchase twice with its original event ID, then separately send a new event ID for the same order. The first tests transport deduplication. The second tests your business rule for one order, which cannot rely on event-ID deduplication alone.
Delay the second support contact’s ingestion by two days while preserving its occurrence timestamp. Compare the report before and after import using a ten-day follow-up window. Both enquiries and the return now fit inside that window. The earlier seven-day conversion report still measures purchase completion on its own terms.
Save event-level evidence alongside the chart settings. A missing path can mean consent refusal, an absent identity link or a collection fault. Investigate the cause before calling it abandonment. For the customer’s explanation of a repeated contact, use voice of customer evidence alongside the event sequence.
Budget for collection and ongoing analysis
Budget for licensed usage and the people who maintain the event pipeline. Monthly events, retained rows and report requests measure different things, so a price comparison needs the same source volumes and history requirement.
Compare charging models
| Tool | Charging unit and included limits | Currency and contract basis | Practical operating model |
|---|---|---|---|
| Amplitude | Free: two million events/month; Plus scales to 70 million; Growth and Enterprise quoted | Plus terms specify US dollars and applicable taxes; checkout sets selected billing terms | Retail product team with an engineer maintaining event and identity rules |
| Mixpanel | Free: one million events/month; Growth up to 20 million; Enterprise quoted | US dollars under standard terms; monthly/annual options and volume affect price | Smaller analytics team with a defined checkout and support event set |
| Adobe Customer Journey Analytics | Quote per million rows; scope ingestion, storage, history and reporting additions separately | Currency, tax and term set in sales order | Established Platform team with data engineering and Workspace analysts |
Use the vendors’ Amplitude pricing and Mixpanel pricing calculators for the selected volume. A displayed starting price does not price your complete retail workload. Keep paid-plan capacity and any optional products separate from the analytics licence.
Estimate instrumentation for each source, schema maintenance after catalogue or checkout changes and recurring data-quality checks. Add analyst time to investigate apparent drop-offs and reconcile confirmed purchases with order records. Include export storage and the work needed to honour deletion requests across copies.
For the illustrative 10.5-million-event monthly workload, both vendors’ free allowances are too small. Request a paid event-volume estimate and compare it with Adobe’s row-based quote using the same history and identity requirements. Choose Amplitude for exploratory journey analysis, Mixpanel for a focused Flows implementation or Adobe when existing Platform capacity makes its additional reporting control practical.
Give your store its own shopping agent
Customers ask in their own words. Your agent finds, compares and checks out from your catalogue, right on your site.