Case Study 04: Platform Integration
Cornerstone to Salesforce
One person, one record, across the learning platform and the CRM
- 01ProvisioningA person created or changed in Salesforce appears in Cornerstone with the right role, without a ticket.
- 02Single sign-onOne login across both platforms. The training is one click from the customer record.
- 03Training recordsCompletions written by Cornerstone are pushed to Salesforce, so a manager sees them where they already work.
- 04ReportingOne report across both systems: who is trained on what, against the accounts they serve.
2 platforms, one record per person. 1 login.
01: ANALYSIS
Two systems, two versions of every employee.
- The organization
- The same regulated consumer-finance firm as the Compliance-First Learning Ecosystem, after the learning platform had launched. Sales and service ran on Salesforce; training ran on Cornerstone.
- What existed
- Two user directories kept by hand. A new hire was created twice, a leaver was removed once, and a manager who wanted to know whether a rep was trained on a product opened a second system to find out.
- Who it affected
- Managers, who needed training status next to the customer record. Compliance, who needed one answer to who was trained on what. Learners, who had two logins for one job.
- The constraint
- Regulated data on both sides. Anything that moved a record had to be tested, validated and signed off before production, and had to fail safely.
The gap, in one picture
Every employee existed twice.
Before: two user directories kept by hand, new hires created twice, leavers removed once, two logins. Target: provisioning, single sign-on, training records and reporting over one validated connection.
Before
- Two user directories, kept by hand
- A new hire created twice
- A leaver removed once
- Two logins for one job
Target
- User provisioning
- Single sign-on
- Training records
- Reporting
One validated, two-way connection, with no manual steps.
02: THE PROBLEM
A training record nobody could find is a record nobody trusts.
- The gap
- The learning platform held accurate completions, and the people who needed them were in a different system. Every lookup was a second login, a second search and a chance to read the wrong row.
- Why a report alone would not fix it
- An export is stale the moment it is sent, and it still leaves two directories to keep in step. The fix had to be the connection itself, running without a person in the loop.
- What success had to look like
- One login. One place a manager looks. One report across both systems. Nothing created or removed by hand that a sync could do.
Stated as a performance gap
Managers could not see training status where they worked, and every learner account was maintained twice. The target was a validated, two-way connection in production, with user provisioning, single sign-on, training records and reporting all running without manual steps.
03: THE DESIGN
Scope the four connections. Direct the build. Validate each one.
The approach
I wrote the scope for each connection as a plain statement of what moves, which way, and when, before any configuration. Then I directed the configuration against that scope and ran the validation myself, one connection at a time, in the order a learner meets them.
01
User provisioning sync
Salesforce became the source of record for people. Create, change and deactivate flow to Cornerstone, carrying the role that decides what training the person is assigned.
02
Single sign-on across both
SSO configured so one identity opens both platforms. A learner reaches a course from the customer record without a second password.
03
Training records pushed to Salesforce
Completions written by Cornerstone appear against the person in Salesforce, so a manager reads training status where they already work.
04
Cross-platform reporting
One report that answers who is trained on what, by team and by product, drawn from both systems at once.
What each connection had to do, stated so it could be tested
- Provisioning
- A person created in Salesforce exists in Cornerstone, with the right role, without anyone raising a ticket.
- Access
- A learner opens assigned training from Salesforce with one login.
- Records
- A completion recorded in Cornerstone is visible on the person’s Salesforce record.
- Reporting
- A manager runs one report across both systems and gets the same answer each system would give alone.
04: OUTCOME
Tested, validated, in production.
What was delivered
All four connections configured against their scope and validated end to end: provisioning sync, single sign-on, training records pushed to Salesforce, and cross-platform reporting. Signed off and put into production.
Attribution, stated plainly
The scope, the configuration and the validation were mine. The platforms’ own connectors did the moving. What the work caused was one record per person and one login; the compliance and adoption results belong to the learning platform launch, told in its own case.
05: EVALUATION
Validated at acceptance; measured once it has run.
| Level | Evidence | Status |
|---|---|---|
| Level 1: Reaction | Managers asked for training status in Salesforce; this is what they asked for. Satisfaction with the report after launch is the next measure. | Planned |
| Level 2: Learning | Not a training deliverable. Administrators were walked through each connection as it was validated. | Built in |
| Level 3: Behavior | Each connection validated against its objective in a test environment before sign-off: a created user, a single login, a pushed completion, a matching report. Whether administrators work from the pushed records rather than the old export is the behavior measure, counted once production has run a quarter. | Validated |
| Level 4: Results | Tickets to create or remove a learner, and time to answer who is trained on what, before and after. Both are counted once production has run a full quarter. | Planned |
06: WHAT I WOULD DO DIFFERENTLY
Scope the integration with the platform.
Sequence
The integration was scoped after the learning platform launched. Scoping it during platform selection would have put the connectors on the evaluation criteria, and the two directories would never have been kept by hand at all.
Measurement
I would count the manual steps before the first connection went in, so the before-and-after was a figure rather than a description.
A record is only evidence where someone can find it.