Conviva DPI Flow Feed

Flows represent critical steps in the user journey. By setting the start and end events of interest, Flow metrics become instantly available. These auto generated metrics represent all the characteristics of the Flow - such as the duration, success rate and frequency - and provide the levers to enable the Flow, and therefore the section of the user the Flow represents, to be optimized. Flows are self-defined, and are not limited in their sophistication or number. You can define custom Flows aligned with business outcomes, such as conversions or error-influenced checkouts, and use the out-of-the-box dashboards to monitor KPIs like success rate, inits, duration, and errors.

Connect Options

There are two methods to programatically interface with Flow data:

Interface	Description	Use Cases
API	RESTful API to query the Flows that have been created, then fetch aggregate metrics for a given Flow. Queries include time information and dimensions to slice the data to the required level. To know more, see DPI Flow API.	KPI Dashboards: Populate management dashboards with real-time App usage levels and the health of key KPIs or conversion rates. Service Monitoring and SLA Reporting: Use real-time data at the service performance layer, for example, API response times, to alert customers about degradation to their service.
Feed	A regular file delivery containing Session, Flow or Event-level data, with each row describing one App interaction. To know more, see DPI Flows in Pulse.	Behavioural Analysis: Map user experience to conversion propensity to understand what drives your customers to transact. Fraud Detection: unpick bad actors looking to circumnavigate your App integrity. Inventory optimization: Uncover what your customers are searching for, what they are adding to their basket and what they are purchasing across any cohort to forecast what to stock next.

Interface

Description

Use Cases

API

RESTful API to query the Flows that have been created, then fetch aggregate metrics for a given Flow. Queries include time information and dimensions to slice the data to the required level.

To know more, see DPI Flow API.

KPI Dashboards: Populate management dashboards with real-time App usage levels and the health of key KPIs or conversion rates.

Service Monitoring and SLA Reporting: Use real-time data at the service performance layer, for example, API response times, to alert customers about degradation to their service.

Feed

A regular file delivery containing Session, Flow or Event-level data, with each row describing one App interaction.

To know more, see DPI Flows in Pulse.

Behavioural Analysis: Map user experience to conversion propensity to understand what drives your customers to transact.

Fraud Detection: unpick bad actors looking to circumnavigate your App integrity.

Inventory optimization: Uncover what your customers are searching for, what they are adding to their basket and what they are purchasing across any cohort to forecast what to stock next.

DPI Feeds

DPI Feeds provide a regular file delivery of data to a chosen location, ideal to gain insight into user behavior, experience quality, and actionable diagnostics for business, technology and operational teams. Feeds can be set to minute or hourly frequencies, in Parquet format.

DPI Flow is a type of Feed where each line provides a description of a user completing a Flow. The exact Flows exported in the Feeds are controlled via the Account Team. Since Sessions are a class of Flow, this Feed also details App Sessions.

Some common use cases of DPI Flow are churn analysis, pricing, user cohorting, trending & forecasting, causal analysis, fraud detection & user journey analysis.

Flow Feed Structures

This section covers the structure of a Flow Feed.

Flow Events

Any Flow is defined by two events:

Initial Event: An event that marks the beginning of the user journey, for example, successful login to a retail platform.

Complete Event: An event that marks the end of the user journey, for example, a successful checkout.

Between these two events, multiple system and user-level activities can occur (e.g. password submission, errors).

Flow Metrics

Each Flow comes with the following metrics:

Flow Init: The number of first initial events within the Flow completion window, maximum 1 hour.
Flow Complete: The number of first initial events followed by a complete event within the Flow completion window, maximum 1 hour.
Flow Avg Duration: The average time from the first initial event to the complete event within the Flow completion window, maximum 1 hour.
Flow Success: The number of complete event pairs (first initial event followed by a complete event within the configured Flow success window)
Flow Error (Optional): The number of errors encountered after the first initial event within the Flow completion window, maximum 1 hour.
Flow Network Request Avg Duration (Optional): The average network request response time after the first initial event within the Flow completion window, maximum 1 hour.
Flow Network Request Failure (Optional): The number of network request failures on or after the first initial event within the Flow completion window, maximum 1 hour.
Flow Network Request Count (Optional): The number of network requests on or after the first initial event within the Flow completion window, maximum 1 hour.

Flow Dimensions

Conviva automatically captures dimensions that can be used to filter to the exact cohort of users of interest. Flow dimensions include details of the user’s device, network, geographic location, and tags, and also considers dimensions associated to initial and last events.

Dimensions are defined during the Activation process of a Flow. To create dimensions, incoming metadata from an event needs to be mapped in the Activation process.This process ensures only the incoming metadata you require is reported and provides an opportunity to normalize disparate metadata names across various devices into a common vernacular the business can use.

Metadata adds context to the Flow, but is typically not relevant to other Flows. For example, the payment type is not relevant to the login Flow, and as such, the volume and types of metadata varies across Flows.

Note: A Flow retains the dimensions associated with both the initial and complete events. However, if a particular dimension exists in both events, Flow captures the most recent dimension value, typically from the complete event. Also, metadata of a Flow comes only from the complete event.

Session Metrics

A Session describes each user’s engagement with the App. Depending on the user’s activity, a Session contains one or more Flows as they interact with the App. These Flows can be sequential (Flow A follows Flow B follows Flow C), but they can also overlap each other. For example, an overarching Flow could be Login to Checkout, with separate lower-level Flows describing each part of the journey, for example, login, discovery, etc. A Session is closed if the user has no interaction with the App for longer than 90s.

Session metrics describe the overall interaction. They include the user’s total enagement time, the number of errors incurred.

Each Flow is mapped to it’s corresponding Session by using the clientID and respective timestamps, enabling the complete end-to-end user journey through the App.

Note: Conviva class a Session as a specific type of Flow. Session data is therefore made available in the Flow Feed and Flow API.

If a Flow itself represents a Session, the following metrics are displayed:

app_startup_count: The number of event App launches in a session.
app_startup_time_ms: Total duration in milliseconds of app startups in a Session. If there are two app startups in a Session, with startup times of 0.1s and 0.2s, the metrics show the sum of the startup times, that is 0.3s.
app_startup_time_ms_max: Maximum duration in milliseconds for an App to start.
user_event_count: Count of user event occurrences.
bad_event_count:Count of bad events. Bad Event is a custom metric. Users can define a specific mapped event as a 'Bad Event' that they do not want to happen too many times. A higher count of a 'Bad Event' indicates lower user experience (QoE).
user_active_time_ms: Duration in milliseconds that a user remains active.
page_load_count: Number of document loading activities that occurred in a Session.
page_load_time_ms: Total duration in milliseconds for the page loading process in a Session.
page_load_time_ms_max: Maximum duration in milliseconds for the page loading process after an App launc.
screen_load_count: Number of conviva_screen_view events with screen load time greater than 0 seconds and less than or equal to 60 seconds in a Session.
screen_load_time_ms: Total duration in milliseconds for the screen loading activities in a Session.
max_screen_load_time_ms: The maximum duration in milliseconds for the screen loading activities.
network_request_success_time_ms: Total duration in milliseconds for a network request to receive a success response code (between 100 – 399) in a Session.
network_request_failure_time_ms: Total duration in milliseconds a network request takes to receive a failure response code in a Session, with codes (4xx, 5xx), and also 0 (zero) or NULL.
network_request_success_count: Count of network requests with a success response code (between 100 and 399), in a Session.
network_request_failure_count: The count of network requests for which a response code is either NOT NULL or NOT between 100 and 399 in a Session, and with a duration of more than 0s and less than 90s.
event_count: Count of both Conviva-mapped and custom-mapped events in a Session.
app_crash_count: Total count of events - app crashes (for Mobile) or WEB errors (for WEB) in a Session.
has_first_video_attempt: If a video_attempt event occurred in the App Session. 1 for TRUE, 0 for FALSE. For App Bounce Rate, count of App Session with 0 has first video attempt/total App Sessions.
time_to_first_video_attempt_ms: Duration in milliseconds from App Session starts to the first video play_attempt event.

Custom Metrics

Custom metrics are associated with specified initial/follow-up events to analyze data for complex user actions and App performance scenarios, such as the time spent between viewer login and the start of video playtime, and the success rate of cart uploads to completed purchase.

DPI enables different types of individual metrics. Start mapping a Conviva predefined event, custom event, or a combination of both as mapping rules to track the desired user behavior comprehensively.

Duration Metric: Measures the time interval between two events, such as login screen load event and the login success event to measure login durations and quickly respond to impacting login issues.

Conversion Metric: Measures the count or conversion rates of events from one stage to another, such as the count of Sessions with successful login after opening the login screen to measure the successful login conversion and find highlights to improve user experience.
Event Aggregation Metric: Measures the number of events, or calculates the value of event attributes, such as the average duration of network requests to measure the network request performance and do optimization accordingly.

So, event aggregation is used to derive summary metrics from raw events. The metrics are derived based on:
- Count of events
- Average value of a tag key
- Total (sum) value of a tag key
Each aggregation is based on a specific event type (for example, Application Error, Purchase). This event type is identified by its Flow. Aggregated metrics are derived from and attached back to this source event.

Each event with aggregation automatically carry a default event_count metric, in addition to any custom aggregations.

DPI Flow Feed Schema

Each line within a Flow Feed contains the following detail:

Fields	Description
Flow	Flow identifiers, including the Flow UID, Flow Name, Sensor Version, Start Time, and End Time. If a Flow is named as Session, it represents an App Session. If a Flow is named as Mapped Events name, it represents a Single Event Metric. If a Flow is named as Flow name, it represents the Flow itself.
Viewer	Viewer details, including their UserID and ClientID which pertains to a unique device.
Location	Location the user is based, approximate to the city level.
Device	Device details, including the OS and browser information.
App	The name and version of the App.
Network	The user’s ISP and access method.
Metrics	A list of metrics relevant to the Flow. Metrics are split into two columns: MetricsInt If it were Mapped Events, it is for Event Aggregation Metric, for example, Count of events. If it were Flow, it is for Flow Metrics: Success/Error/Duration, etc. MetricsFloat (for percentages, averages, etc.)
Tags	Mapping metadata to a Flow is done during the Activation phase, which you can change anytime.

Getting started

To get started with DPI Data Feeds, contact your Account Team.

Also, learn more about managing Flows in Conviva Pulse.

DPI DPI DPI Flow Feed Flow Feed Flow Feed Data Feed Data Feed Data Feed