A redesign saving measured against Microsoft's own compliance determination and the same saving measured like-for-like against today's actual grant are different numbers, and on one client the difference was 69 percent of the headline. Name the baseline on every figure or the number means nothing.
Two baselines, two very different answers
On one client, the saving measured against Microsoft's compliance determination was 562,032 dollars, or 29.8 percent. The same redesign measured like-for-like against a model of the user's actual grant today was 171,552 dollars, or 11.45 percent. Same redesign, same users, same prices. Roughly 69 percent of the headline saving was not the redesign doing anything. It was the two baselines disagreeing about what users need right now.
If you quote one of those numbers without saying which baseline it uses, you have not quoted a saving. You have quoted a number.
What each baseline actually measures
The two baselines are not two price scales. Both put every user on the same base-plus-attach pricing. They are two different determinations of which modules a user needs today.
The compliance baseline is Microsoft's determination for the client's current roles, the figure that falls out of a compliance recompute. The like-for-like baseline is a privilege-derived model of what each user actually uses today, assembled from usage evidence and reduced to a minimum-cost module set.
When these two disagree, they disagree about the present, not about the redesign. The compliance view may say a user needs a module that the usage model says that user never touches. That disagreement shows up as saving if you measure against compliance, and it vanishes if you measure like-for-like, because the like-for-like baseline already agrees with the redesigned state about what the user needs.
Why the distinction decides whether a number is defensible
A saving measured against compliance includes two things rolled together: the effect of the redesign, and the model's disagreement with Microsoft about today's requirement. Only the first is something the redesign delivers. The second is a claim that Microsoft's current determination is too generous, which may well be true, but it is a different argument and it has to be made on its own.
The like-for-like number isolates the redesign's own effect, because it holds the view of today's need constant on both sides. That is the number that answers the question a client actually asks, which is how much the redesign itself changes the bill.
Reporting only the compliance number flatters the redesign. Reporting only the like-for-like number can understate the total opportunity. The honest answer is both, each labeled, with the gap between them explained as a disagreement about the present.
The row has to carry its own baseline
The practical fix is to stop letting any saving float free of its basis. Every saving figure should sit next to the baseline it was measured against, the basis of that baseline, the modeled current grant, and the like-for-like effect as a separately named value, so that baseline minus license equals saving on the face of the row and anyone can check it.
One guard matters here. The like-for-like column has to be left empty when the current cost is a real invoice rather than a model, because subtracting a modeled design from a real invoice is not a like-for-like comparison and the result is not meaningful.
Where the two baselines diverge materially, that divergence is a signal to investigate before quoting either number, not a detail to smooth over. The way to settle it is a fresh compliance recompute, not an argument about which model is smarter.
Bottom line
A saving with no named baseline is not auditable, and an unauditable saving is the fastest way to lose a finance leader's trust. On the client where the headline was 562,032 dollars, only 171,552 dollars was the redesign. The other 69 percent was a real and separate argument about what Microsoft says users need today. Both are worth having. Neither survives being blended into one unlabeled number.
Frequently asked questions
Which number should we lead with?
Lead with both, labeled. If forced to pick one as the redesign's own result, the like-for-like figure is the defensible choice, because it isolates what the redesign changes from what the model disputes about today.
Does a large gap between the two mean one is wrong?
Not by itself. It means the usage model and Microsoft's compliance determination disagree about current requirements. That is worth resolving with a compliance recompute before quoting, because the resolution can move the number either way.
Why can't we just compare against the client's real invoice?
You can, but then the like-for-like comparison does not apply, because a modeled design measured against a real invoice is not like-for-like. Keep that column empty on an invoice basis and report the compliance comparison instead.