Salesforce

Data 360 Testing: Identity Resolution to Agent Grounding

Rishabh Kumar

Published On

Data 360 Testing
eBook CTA Background
RECOMMENDED RESOURCES

Salesforce testing: Complex Element Selection Best Practice

While some applications are easy to automate, others, like Salesforce, need a little extra love and attention. This is our best practice guide to complex element selection.

Get the eBook →

Ebook image
eBook CTA Background
RECOMMENDED RESOURCES

Salesforce testing: Complex Element Selection Best Practice

While some applications are easy to automate, others, like Salesforce, need a little extra love and attention. This is our best practice guide to complex element selection.

Get the eBook →

Ebook image
eBook CTA Background
RECOMMENDED RESOURCES
Ebook image

Salesforce testing: Complex Element Selection Best Practice

While some applications are easy to automate, others, like Salesforce, need a little extra love and attention. This is our best practice guide to complex element selection.

Get the eBook →

eBook CTA Background
RECOMMENDED RESOURCES

Salesforce testing: Complex Element Selection Best Practice

While some applications are easy to automate, others, like Salesforce, need a little extra love and attention. This is our best practice guide to complex element selection.

Get the eBook →

Ebook image

Table of contents

Lorem ipsum

Lorem ipsum

Lorem ipsum

Lorem ipsum

Lorem ipsum

How Do You Test Salesforce Data 360?

Test Data 360 with a controlled set of known people and accounts whose correct outcome you already know. Check that the right records merge into one profile and the wrong ones don’t, that each profile lands in the right segments, that activations reach their destinations correctly, and that agents grounded on the data give answers that match it.

Data 360, the platform Salesforce previously called Data Cloud, takes in data from Salesforce and other systems, maps it to a common model, resolves records into unified profiles, and makes them available for segments, activation, automation and Agentforce.

Its defects are quiet. A wrong merge doesn’t throw an error. It produces a confident, wrong customer.



Where Data 360 Goes Wrong

Salesforce renamed Data Cloud to Data 360 in October 2025, as part of Agentforce 360. It’s now positioned as the data layer that gives Agentforce its context, as well as the customer data platform behind marketing and analytics.

Stage

What happens

How it goes wrong

Ingestion

Data streams bring in records from Salesforce, warehouses and other systems, or query them in place through zero-copy federation

Streams stall or lag, fields arrive in unexpected formats, deletions aren’t reflected

Harmonisation

Source data is mapped to data model objects

A field maps to the wrong attribute, or a source is left unmapped

Identity resolution

Match and reconciliation rules merge records into unified profiles for people, accounts and households

Different people merge into one profile, or one person stays split across several

Insights

Calculated insights derive measures such as lifetime value

Wrong inputs produce plausible, wrong numbers

Segmentation

Audiences are built from profile attributes and behaviour

People who should be excluded, such as those who opted out, are included

Activation

Segments and changes are pushed to Marketing Cloud, ad platforms, flows or external APIs

The wrong audience is sent, or nothing is sent and nobody notices

Grounding

Agentforce agents use unified profiles and indexed content to answer and act

Agents answer from the wrong profile or outdated content

Identity resolution is the row to worry about most. A confidently wrong merge is worse than no merge at all, because an agent or a marketer then acts on someone else’s data.



Know Your Sandbox Before You Test in It

Data 360 sandboxes behave differently from ordinary Salesforce sandboxes, and most early test failures trace back to one of these.

No Data Comes Across

A Data 360 sandbox copies metadata, such as data streams, data model objects, calculated insights and segments, but none of the data. Test records have to be seeded on purpose.

Unified Profiles Start From Nothing

Identity resolution rules are copied, but profiles aren’t. You have to trigger the first identity resolution run yourself after ingesting data.

Refreshes Break Connections

Refreshing a connected CRM sandbox changes its org ID and permanently breaks its link to Data 360, and connectors to systems such as Marketing Cloud need reauthorising.

Old Sandboxes Test the Past

A sandbox that hasn’t been refreshed in months reflects old identity rules and segments, so results won’t match production.

Sandboxes Cost Credits

Data 360 sandbox use draws on production credits, which is worth planning around before running large tests.

Seeding, masking and refreshing data across environments is a discipline of its own. Get the full picture from our article on test data management.



Testing Data 360, Step by Step

  1. Build a Known-Identity Test Set

Create a small cast of synthetic people and accounts whose correct outcome you decide in advance. Include two records for the same person with different email formats, which should merge, and two different people at the same address, who shouldn’t. Add a household, a customer who has opted out, and a customer whose data arrives from three systems.


  1. Seed it Where it Would Really Come From

Load each test record into the source systems and streams it would naturally arrive through, so the test exercises real ingestion and mapping.


  1. Run Identity Resolution, Then Check the Profiles

For each expected profile, check which source records it contains, and which it mustn’t. A wrong merge and a missed merge are both failures.


  1. Check Segments and Insights

Each test person should appear in exactly the segments they belong in, and excluded people never should. Check calculated insights against values you worked out by hand.


  1. Follow Activation to Its Destination

Where a segment or change triggers an activation or a flow, check the result in the destination system. The right people should be added to the marketing audience, and the right record updated, once.


  1. Test Agent Grounding With Known Facts

Ask an Agentforce agent about a test customer whose unified profile you control, and check that its answer and any actions match that profile, not a neighbouring one.


  1. Retest on Change

Rerun the set whenever match rules, mappings, segments or sources change, and in each Salesforce release.


Keep the known-identity set small enough to reason about and rich enough to break things. Ten to twenty carefully designed identities catch more than ten thousand random records.

Keeping that cast fictional but realistic is exactly what synthetic test data is for. Grounding checks also belong in the same baseline as the rest of an agent's behaviour, which is where Agentforce regression testing comes in.

Testing Salesforce Data 360



Where Virtuoso QA Helps

Virtuoso QA tests Data 360 the way the business experiences it, from the source system where data starts to the profile, segment, activation or agent answer where it ends up.

The known-identity set lives in data tables, and AI test data generation keeps every value synthetic. Browser and API steps create the test records in Salesforce and connected systems, so ingestion and mapping are genuinely exercised, then journeys check unified profiles, segment membership, activation destinations and agent answers, carrying identifiers from one stage to the next.

Touchstone can turn uploaded match rules, segment definitions and data model documents into requirements that cite their source, including the merges that should never happen, for a named person to approve.

Virtuoso QA's Salesforce test automation seeds known identities through your real sources, then checks profiles, segments, activations and agent answers in one journey, so a wrong merge is caught before anyone acts on it.



Related Reads

Frequently Asked Questions

Is Data 360 the Same as Data Cloud?

How Do You Test Identity Resolution?

Do Data 360 Sandboxes Include Data?

How Do You Test Agentforce Grounding on Data 360?

Written by

Rishabh Kumar

AI Testing & Quality Engineering Contributor

Rishabh leads business development and marketing at Virtuoso QA. He writes about AI-powered test automation, end-to-end testing, self-healing tests, and quality engineering, exploring how these capabilities can improve test coverage, reduce test maintenance, and support reliable software releases.

See what your next release looks like with Virtuoso

Book a walkthrough on your applications and your workflows. Bring a requirement, a user journey, or a brittle Selenium script, and watch the loop run on something you recognise.

See what your next release looks like with Virtuoso

Book a walkthrough on your applications and your workflows. Bring a requirement, a user journey, or a brittle Selenium script, and watch the loop run on something you recognise.

See what your next release looks like with Virtuoso

Book a walkthrough on your applications and your workflows. Bring a requirement, a user journey, or a brittle Selenium script, and watch the loop run on something you recognise.

Virtuoso QA is establishing the standard of proof for software releases. Its governed QA loop turns business requirements into tests for any browser-based application: AI proposes, a deterministic engine executes, a person approves what matters, and every decision leaves evidence.

Trust Center

AICPA

SOC

WAVE STRONG PERFORMER

@ Copyright 2026 SpotQA, Creators of Virtuoso QA

Virtuoso QA is establishing the standard of proof for software releases. Its governed QA loop turns business requirements into tests for any browser-based application: AI proposes, a deterministic engine executes, a person approves what matters, and every decision leaves evidence.

Trust Center

AICPA

SOC

WAVE STRONG PERFORMER

@ Copyright 2026 SpotQA, Creators of Virtuoso QA

Virtuoso QA is establishing the standard of proof for software releases. Its governed QA loop turns business requirements into tests for any browser-based application: AI proposes, a deterministic engine executes, a person approves what matters, and every decision leaves evidence.

Trust Center

AICPA

SOC

WAVE STRONG PERFORMER

@ Copyright 2026 SpotQA, Creators of Virtuoso QA

Platform

Solutions

Resources

Platform

Solutions

Resources

Platform

Solutions

Resources