The persona mix shift models at 27 percent and lands at 12 to 18, and the step down right is the gap

Map every user to the lowest plan that covers the work and a heavy E5 estate models a 27 percent list saving. Almost nobody collects it. The plans that fail are not the ones with bad arithmetic, they are the ones that do the arithmetic without securing the contractual right to act on it.

Prepared by Redress Compliance · August 16, 2026 · Microsoft advisory. 30 to 45 Microsoft 365 renewals benchmarked, 2024 to 2025.

E3 is the productivity core. E5 layers security, compliance, voice, and analytics on top. F3 drops the desktop apps and shrinks storage for frontline use. List runs roughly six to one from F3 to E5.

The original analysis states the direction and stops. Here is the arithmetic in full, because the size of the prize and the size of the shortfall both matter.

Note that the seat count never changes. Every dollar in that table comes from placement rather than from reduction, which is why the saving survives a headcount freeze and why it is available without cancelling anything.

The standard account team pitch is that E5 is the safe default and a deeper discount on a broad E5 base is the prize. That is wrong for a reason the benchmarking makes plain: in roughly 30 of the 45 estates reviewed, a quarter to nearly half of E5 seats never touched a single E5 only feature, so the discount was negotiated on top of capability nobody opened. A deeper percentage off an oversized, wrongly mapped baseline still overpays. But the more useful finding is the one that comes after you accept that argument, because the buyers who accepted it still did not collect what the model promised.

Three things eat the difference between 27 percent and the 12 to 18 that lands. The first is role exception. A persona map is built from feature telemetry, and telemetry says nothing about the user who needs Purview four times a year during a regulatory response, or the executive whose seat is a policy decision rather than a usage one. Those seats come back onto E5 during review, and they come back in clusters rather than singly. The second is the security floor. Most estates carry a written standard that puts a whole class of user, often anyone touching regulated data, on the advanced tier regardless of measured use. That standard is usually correct and it is rarely renegotiated to fit a licensing exercise.

The third is the one that is actually contractual, and it is the only one you can fix at the table. A mix shift is a reduction in the quantity of the expensive SKU, and an Enterprise Agreement does not generally let you reduce quantity whenever the analysis finishes. Without a negotiated step down right at anniversary, the work completed in month three of a three year term cannot be banked until month thirty six, and by then the estate has drifted again. This is why the same analysis produces a 27 percent number in the model and a 15 percent number in the ledger. The arithmetic was never the constraint.

So the sequence that actually pays runs in the opposite order to the one most teams use. Secure the step down right first, while the renewal is open and Microsoft needs the signature, then do the persona work against a contract that lets you act on it. Cost the add on route per persona rather than assuming the suite, since Defender, Purview, and Power BI Pro can each sit on an E3 base for the population that genuinely needs them. And treat the exception list as a negotiation input rather than an inconvenience: knowing that 400 seats will come back to E5 lets you commit to a number you can hold, instead of promising a board a saving the estate will refuse to deliver. The placement question more broadly is worked in Microsoft 365 licensing, the E3 and F3 boundary in F3 versus E3, and the wider library in the Microsoft practice.

Copilot needs a qualifying base plan. E3 and E5 qualify. F3 does not. That single prerequisite turns a $30 decision into something quite different for any frontline population.

• Negotiate the step down right first, at anniversary, while the renewal is open. Every other lever on this list depends on being able to act on its findings before the term ends.
• Map personas against feature telemetry, then build the exception list deliberately rather than discovering it during review. Security owners, analysts, voice users, frontline, deskless approvers, and the knowledge worker baseline cover most estates.
• Cost the add on route per persona. Defender, Purview, and Power BI Pro can sit on an E3 base, and the suite only wins when the population needing those capabilities is large enough to beat buying them individually.
• Move deskless and shift workers to F3, which was 10 to 20 percent of E3 seats in the estates we benchmarked and is the cleanest single move on the list.
• Stage the Copilot ramp against the base plan prerequisite, budgeting the F3 to E3 jump as part of the pilot rather than absorbing it later.
• Take the rate last. Negotiate the discount on the corrected seat count, because a deeper percentage on an inflated baseline is the outcome the account team is optimising for.

Across roughly 30 to 45 Microsoft 365 renewals benchmarked, the default E5 for everyone position was the single largest waste pattern:

The third pattern was the Copilot base plan jump. Pilots quietly moved F3 users onto E3 before any Copilot licence was bought, adding a cost nobody had budgeted and one that persists after the pilot concludes.

The seat that costs the most is an E5 licence assigned to a user who only ever opens Word, Teams, and Outlook. Finding those seats is straightforward. Being contractually able to move them is the part that decides whether the exercise pays.

Watch the briefing · 6:19Microsoft E5, Or NotHow to test the E5 case against what you already license, and where the add on route beats the suite.

What separates E3, E5, and F3?

E3 is the productivity core with full desktop Office and 1 TB storage. E5 adds advanced Defender and Purview, Teams Phone, and Power BI Pro. F3 is the frontline plan: web and mobile Office only, 2 GB storage, no Teams Phone or Power BI Pro. List runs about $36, $57, and $8 per user per month.

How much does a persona remap actually save?

On the worked 10,000 seat estate it models 27.4 percent, or $1.55m a year off a $5.66m list bill. Realised savings across the renewals we benchmarked landed at 12 to 18 percent, roughly $680,000 to $1.02m. Model the higher figure, commit to the lower one.

Why does the realised saving fall short of the model?

Three reasons. Role exceptions that telemetry cannot see, such as the user who needs Purview during an annual regulatory response. Written security standards that place whole user classes on the advanced tier regardless of measured use. And the absence of a step down right, which prevents acting on the analysis until the term ends.

What is a step down right and why does it matter most?

It is the negotiated ability to reduce seat quantities at the anniversary rather than only at term end. Without it, mix work completed in month three of a three year term cannot be banked until month thirty six, by which point the estate has drifted. It is the single lever that converts the model into money.

How many E5 seats are typically wasted?

Between 25 and 45 percent used no security or compliance feature beyond what E3 already carries, and in roughly 30 of the 45 estates benchmarked that pattern held. The discount everyone negotiates therefore sits on top of capability nobody opened.

Can Copilot run on F3?

No. Copilot requires a qualifying base plan and only E3 and E5 qualify. A frontline user must move from F3 to E3 first, which is a $28 base jump before the $30 Copilot line lands. The seat goes from $8 to $66 a month, and less than half the increase is Copilot.

When does the add on route beat full E5?

When the population genuinely needing advanced Defender, Purview, or Power BI Pro is a minority of the base. Buy the capability for that group on an E3 base and leave everyone else on E3. The suite wins only when the qualifying population is large enough to beat buying the components individually.

How many E3 seats should really be F3?

Between 10 and 20 percent in the estates we benchmarked. Frontline, shift, retail, field staff, and deskless approvers who never open desktop Office all fit F3 at roughly a fifth of the E3 cost, and this is the cleanest single move available.

Does the mix shift require cutting headcount?

No. Every dollar in the worked example comes from placement rather than reduction, and the seat count is identical before and after. That is why the saving survives a headcount freeze and why it can be pursued without cancelling anything.

In what order should the levers be pulled?

Step down right first, then persona mapping, then the add on costing, then the frontline shift, then the Copilot ramp, and the rate last. Negotiating the discount before correcting the seat count is exactly the sequence the account team prefers.