Corporate LMS Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does an LMS build turn into a clone of the platform you are leaving?

Because the specification gets written from the incumbent's admin console rather than from the pain. Somebody exports the feature list, circulates it, and every stakeholder adds the report they once asked for. Social learning goes in, then a content marketplace, then a skills taxonomy nobody has populated. By the time the document reaches a vendor it describes a platform that took a funded product company a decade to build, and it is being priced against a training budget.

This happens harder in learning and development than in most functions, for a specific reason. The LMS is the one system the whole company touches, so the whole company has an opinion, and none of those opinions carry a cost. Finance wants the per-seat line to stop growing. Compliance wants an export in the regulator's format. The head of L and D wants a heatmap. A regional manager wants a leaderboard. Only the first two are why you are spending money.

The fix is uncomfortable but simple. Write down, in one page, the two or three conditions that made a custom build defensible: the renewal quote after a headcount bump, a certification rule the platform cannot express, a partner training portal you cannot get walled off. Build those. Keep the incumbent tenancy running for what it does adequately, on the smallest seat count you can negotiate, and switch that off later. A first release that replaces one costly use case and proves itself in three to five months earns the budget for the next phase. A specification that promises everything gets cut in month nine by someone who was not in the room when it was written.

What goes wrong when you migrate completion history and course packages?

Two things, and the second one is the one that hurts in an audit.

The first is course state. A SCORM package does not just store whether someone finished. It stores suspend data, the bookmark that puts a learner back on slide 34 of a 90 minute compliance module, the per attempt score, and the interaction records behind that score. Export a package and its completion flag, and you have preserved the outcome while destroying everyone's place in it. A workforce forced to restart mandatory training will tell you about it, and the completion rate you report that quarter will be wrong for reasons unrelated to learning.

The second is evidence. Your completion records are what an auditor reads. If the export gives you a name, a course title and a date, but not the version of the content in force when the person completed it, then the question every auditor eventually asks, which is which version of the policy this employee acknowledged, has no answer.

Treat migration as its own phase with an acceptance test rather than a data load at the end. The test is concrete: take a sample of learners spanning roles, locations and years, pull their transcript from both systems, and compare line by line. Anything present in one and absent from the other is a defect, not a variance. Keep the original record exactly as it was exported alongside any normalised form, freeze content versions before you migrate, and store the version identifier on every completion from that point forward.

Why do single sign on and HRIS sync break after launch?

Because both are usually built and tested against the joiner case, and the joiner case is the easy one.

Single sign on is tested with one account in a sandbox tenant. It works. Then it meets reality: a contractor with no directory entry, a person in two business units, a country where the identity provider is a different tenant entirely, and an executive whose account name changed after a marriage. None of these are exotic, and none of them were in the demo.

The human resources information system sync is worse, because most builds implement it as a nightly full file import that handles create and update. Movers break it: somebody changes role on the twelfth, their curriculum changes with the role, and an importer that only reconciles attributes leaves them on the old path with the new job title. Leavers are the liability version: the employment relationship ends, the person keeps a valid session, and your access review finds it eight months later.

Model the sync as events rather than as a file. Joiner, mover and leaver are three distinct cases and each deserves its own test with its own expected outcome, including what happens to in progress training when a role changes. Deprovision on the leaver event rather than on the next scheduled run. Log every provisioning action with a timestamp and a source, so that when somebody asks why a person lost access on a Tuesday, the answer is in the system rather than in a memory.

What happens when certification expiry and audit evidence are not covered?

A certification gets built as a date field with an email reminder attached, and everything expensive about certifications lives outside that model.

A forklift certification valid for 24 months is not a date. It is a lifecycle with states: issued, approaching expiry, expired, revoked, renewed. Each transition needs an owner, an action and a consequence. Approaching expiry should open a renewal path with the re test attached rather than send a message. Expired should have a consequence, and in a safety context that consequence is usually operational: the person is not scheduled on that equipment. If the LMS knows a certification lapsed and nothing downstream changes, the LMS has recorded a risk rather than managed one.

Escalation is the part almost every build omits. A reminder that goes to the learner and nobody else fails quietly at scale, because the people most likely to let a certification lapse are the people least likely to read training email. The reminder needs a named manager on it, and a second escalation when the date passes.

On evidence, build the export as a feature rather than as a report somebody assembles. A stored, versioned query producing a timestamped record with learner identity, course version, completion time, score where applicable and the attestation captured, in the format your auditor asked for, is the difference between a two hour request and a two week one. Ask your compliance lead for a real example of what an auditor sent last time and make that document an acceptance criterion.

Should you build custom or configure what you already own?

Configure, if you are under roughly 2,000 to 3,000 active learners and your compliance needs fit standard settings. Docebo and Cornerstone are genuinely capable at that scale, the speed to launch is real value, and the vendor absorbs the maintenance and security work you would otherwise own. We would tell a head of L and D exactly that, and have.

Configure also if the problem you are actually trying to solve is that nobody finishes anything, because low completion is a content and manager accountability problem in most organisations and a new platform reliably fails to fix it.

There is a middle path worth naming. A custom portal and reporting layer sitting on top of an open source LMS core caps licence spend while keeping control of the learner experience and the reporting, and it costs a fraction of a full replacement. For a lot of organisations whose only real complaint is per seat pricing and a reporting gap, that is the right answer.

Build when per seat cost has become the largest uncontrolled line in the L and D budget, when you need multi tenant partner or dealer training the platform will not cleanly wall off, or when your certification and compliance logic genuinely cannot be expressed in the product and you have the rejected change requests to prove it.

How do hidden costs get into the quote?

By being real work that nobody names. The recurring ones in this category:

• Content re authoring. If any of your library was built in a tool whose output the new platform handles differently, somebody is opening those source files. Audit the library before you scope, not after.
• Migration. Frequently priced as a data load and delivered as a project. See the acceptance test above.
• Integration waiting time. A build blocked on identity or HR sandbox access still burns budget. This is the single most common schedule overrun in LMS work and it is almost always internal, not vendor caused.
• Video hosting and bandwidth. Cheap until a 4,000 person mandatory rollout on the same Monday morning.
• Accessibility conformance. If your organisation has a standard, meeting it is design and test effort across every learner screen, and retrofitting costs far more.
• The overlap. You pay the incumbent renewal while the custom platform is built and piloted. Budget those months honestly.
• Maintenance. Roughly 15 to 20 percent of build cost annually for patching, dependency upgrades and the integration upkeep that stops syncs breaking when a vendor changes an interface.

Ask any bidder to price these as named lines. A proposal that omits all seven is not cheaper, it is less complete.

What separates a build that works from one that fails here?

A single accountable owner inside L and D who can make a decision without convening a committee. Builds that stall in this category almost always stall on decisions, not engineering, and the decision that stalls longest is which of the incumbent's features you are prepared not to have.

Integration access arranged before kickoff, in writing, with named people on the identity and HR side who know it is coming.

A first release small enough to ship in a quarter and real enough that somebody stops using the old system for something. Reject any proposal that asks for the full budget before you see working software running against your own data.

Ownership of the code, the data and the infrastructure accounts, agreed before the first commit rather than at handover. A training record is an employment record and it will outlive your relationship with any vendor.

And a partner who argues with your scope. A bidder who says the first 1,500 learners should stay on the platform you already have, or that you need the mid range build rather than the enterprise one, is optimising for your outcome rather than their invoice. That is the strongest signal available before a contract is signed.