When to Replace Something

Replacing a module is the usual response to a persistent problem and is frequently the wrong one, because the problem is at a boundary.

The diagnostic question

Is the problem inside the module, or at one of its edges?

Inside: the product genuinely cannot express a rule you need, or its interface is unusable by the people who use it daily.

At the edge: figures do not reconcile, data arrives late or wrong, or the same rule is calculated twice.

Replacing a module fixes the first and moves the second, because the new product inherits the same identity mapping and the same reference data.

Signals that it really is the product

A rule you need that cannot be configured, confirmed with the vendor rather than assumed.

Daily users working around the interface, consistently, across sites.

A vendor no longer investing: no meaningful releases, support degrading, a roadmap that has not moved.

Per-seat cost rising faster than value with no negotiating position.

Structural limits — no external identity, no per-site or per-country configuration — that cannot be worked around.

Signals that it is the join

The problem appears at period end rather than during use.

It is described as "the systems don't talk".

Each module's own reports are fine.

And the same problem existed with the previous product, which is the strongest signal of all and the one nobody checks.

Before deciding

Run the four reconciliations and see which fails.

Check the rule register for the calculation in dispute.

Ask the daily users what they actually work around.

A fortnight, and it usually answers the question — frequently by showing that the module everyone blames is downstream of the one causing it.

If replacement is right

Fix identity and reference data first, because the migration will otherwise carry the same problems into the new product.

Build the payroll join early and run parallel.

And use the change to reduce what you hold, which is the cheapest opportunity you will get.

The honest position

Most workforce estates would improve more from a rule register and a monthly reconciliation than from any replacement, and both cost a fraction of one.

Which is worth establishing before a procurement starts, because afterwards the momentum makes it unsayable.

Reproduce the workflow

For another way to make this requirement testable, consult the policy example. Reproduce the case with real roles, codes, failures and recovery steps.

Independent reference

For a thematic point of reference, see Microsoft Learn. Its current material provides useful context beyond product documentation.