Data validation

This is for:

Developer
Important

The Event Protocol and Coveo Relay are currently in open beta. If you’re interested in using the Event Protocol and Relay library, reach out to your Customer Success Manager (CSM).

This article covers the importance of validating the event sent to your Coveo organization using the Event Protocol and the tools available to help you validate your events.

Why validate?

Capturing commerce events allows you to trace a user’s journey through your Coveo-powered commerce solution by gathering data on the interactions with various elements. Validation is a critical process that ensures the integrity and effectiveness of the logged data by verifying the events have been implemented correctly.

In the context of commerce events, precise tracking and comprehensive reporting of user interactions prove indispensable. Coveo Machine Learning algorithms perform well when fed with good quality data. Inferior data quality inevitably results in below-par machine learning models, leading to a substandard personalized experience. Thus, data validation becomes vital as flawed or incorrect data can lead to inaccurate predictions and unreliable model performance.

Measuring visitor interactions on your site, particularly within the context of conversions and attribution, relies on accurate data. Therefore, it’s important the events sent to Coveo are accurate, as this data forms the foundation for reporting conversions and revenue for each of your Coveo experiences.

Requirements for validation

These guidelines assume that the person performing the validations is familiar with the following:

Note

If authentication is required, having a valid account with the necessary privileges is vital. Access to all interfaces will allow you to inventory all search hubs and search interfaces (such as main search, recommendation components, listing pages, case deflection, etc.) to ensure that all Coveo-powered components are integrated correctly.

Performing data validation

There are several ways to validate your data, ranging from validating one event at a time to gaining a broader view of your data.

Validating single events

Single-event validation involves ensuring that the event sent to Coveo contains a valid payload.

The recommended approach for validating events is by using Coveo Explorer.

Coveo Explorer is a Chrome extension offering insights into your Coveo implementations by validating events sent to Coveo. Explorer works by validating the payloads of events, ensuring they follow the correct structure, contain the required fields, and have the correct data types.

It’s advisable to use Explorer to make sure that specific user interactions, such as clicking a product or adding a product to a cart, send the correct event payload.

For detailed information on installing Explorer and using the extension, refer to the guide on Validating events with Explorer.

Note

It’s highly recommended to use Explorer to validate your events. If unable to do so, use your browser developer tools to inspect the requests sent to Coveo.

To achieve this, record the network traffic and use the filter box to filter events by events/v1. This will display only requests sent to Coveo.

Inspect the request payload to validate that it contains the required fields by cross-referencing them with those provided in the Event Protocol Reference documentation for the specific event you’re validating.

Data completeness

During and after the implementation of data tracking, you should review the data health dashboard on the Data Health (platform-ca | platform-eu | platform-au) page of the Coveo Administration Console.

Data health refers to the integrity of organizational data, determining its reliability and accuracy for sound analytics and Coveo Machine Learning models. The dashboard explicitly identifies failures detected through data validation rules. You can review this dashboard to pinpoint inconsistencies by analyzing the impact of submitted usage analytics events on the data health.

The data health dashboard provides a comprehensive snapshot of your Coveo organization’s data quality. It outlines the number of events failing validation rules, categorized by severity, and breaks down these occurrences by event type. This dashboard is helpful for finding inconsistencies in recorded data and understanding how each event affects the overall quality of the data.

Additionally, the Data Health page also provides a data health score. This numerical value ranges from 0 to 100, offering a rapid assessment of an organization’s data quality. A corresponding color indicator offers a visual prompt for potential data quality concerns. This score reflects the effect of data quality on commerce dashboards. When encountering a low score, it’s essential to conduct a thorough investigation into potential factors contributing to the issue. This includes analyzing the cause of any validation rules marked as "Critical" severity.

For most usage analytics related issues, front-end difficulties are the probable culprits. Grasping the scope of failed validation criteria and locating them within the Data Health page can help troubleshoot potential data health problems for your organization.

For additional details on utilizing the data health dashboard, see the Data Health troubleshooting tutorial.

Validating conversions

Another layer of validation checks you can perform is validating the event data sent to Coveo UA with your database that records transactions. We recommend you perform this validation approximately one week after implementing data tracking.

The primary objective here is to align orders with their respective revenue figures. This process enables you to ensure the completeness of orders and revenue data within Coveo’s records.

As the Data Health page can’t verify the data health of events not sent to Coveo, it’s vital to validate transaction records from your backend by comparing them with the data available in Coveo.

For a thorough validation, compare a substantial volume of orders from Coveo against your backend. This could encompass a few days' worth of records or even up to a week’s worth, depending on the scale of the organization. Match each individual order using its transaction ID and its corresponding revenue figure. This approach allows you to identify any discrepancies between the two data sources.

We recommend conducting this process regularly, such as quarterly. Repeating these validations over time reinforces the trustworthiness of the aggregated information and upholds the integrity of your data analysis.