E1, E3, E5, the F frontline tiers, and Business: what each includes, who each is for, and the upgrade paths between them. Three knowledge checks along the way, and 1 clip from a senior cloud advisor.
This is a taught session, not a talking head. The instructor works through analyst grade slides, and three times the video stops on a question with four options on screen. Pause, commit to an answer, and the next slide explains which option is right and why each of the others is wrong. Once in the session the frame splits and a senior cloud advisor gives the view from inside real Oracle negotiations, and the instructor picks the clip apart when the slides return.
The full narration of this session, section by section, for reading and reference. Guest analyst clips are marked.
Welcome back, session eleven of forty, and this opens module three. Module two was the machinery: how an Enterprise Agreement is structured, how it counts, how it meters, and how it arrives priced on a sheet. Module three is what you put through that machinery, which is the Microsoft 365 stack. And I want to start with the claim that shapes the whole module. On most enterprise estates, the plan mix moves the bill further than the discount does. Not slightly further. On a ten thousand seat estate the difference between a persona aligned mix and a standardised one runs into the hundreds of thousands a year, held for the whole term, and it sits entirely inside decisions you control without needing the vendor to agree to anything. That is unusual leverage and most organisations spend it without noticing, by deciding the mix quickly and then negotiating the rate slowly.
Five takeaways. One, the map: five plan families and how they relate, E1, E3, E5, the frontline F tiers, and Business, where each starts and where each stops. Two, the list spread: roughly seven to one from F3 to E5, which is the arithmetic that makes allocation outrank discount. Three, personas over averages: most estates carry four to six recurring personas, and licensing to the average overpays for most of them while underserving the rest. Four, the upgrade paths: which moves are simple, which are one way in practice, and which quietly force a base plan change nobody budgeted for. Five, what the mix is worth: a persona aligned shift moves the annual list bill by roughly twelve to eighteen percent on a large estate, and unlike a discount it holds for the whole term because it changes what you buy rather than what you pay for it.
Put the feature matrices aside for a moment, because underneath them the stack answers three questions. First: does this person need the desktop applications. That is the biggest single fork. E1 and the frontline tiers give web and mobile Office. E3 and above give the installed desktop apps. A large share of misallocation in real estates is people paying for desktop applications on machines that never launch them. Second: how much security and compliance. E3 carries a baseline, E5 carries the advanced Defender, Purview, and identity layer, and that delta is the whole of next session. Note the framing though: it is a security decision with a price attached, not a licensing decision with a security story attached. Third: is this person at a desk at all. The frontline tiers exist for shift, retail, manufacturing, and field work. Shared devices, short sessions, small storage. Where the work genuinely looks like that, an enterprise tier is not a better licence, it is just a more expensive one.
The tiers side by side, with list prices, and treat these as anchors rather than as what you will pay. Microsoft 365 F3, around eight dollars per user per month: web and mobile Office, Teams, two gigabytes of OneDrive, basic security, built for frontline and shared device work. Office 365 E1: web and mobile Office with full Teams and Exchange but no desktop apps, sitting below E3 and above the F tiers. Microsoft 365 E3, around thirty six dollars: full desktop Office, a terabyte of OneDrive, Windows and mobility, baseline Defender and Purview. Microsoft 365 E5, around fifty seven dollars: everything in E3 plus advanced Defender, advanced Purview, Power BI Pro, and Teams Phone. Business plans, for organisations under the three hundred seat cap. And Copilot at thirty dollars as an add on, requiring an E3 or E5 base, which frontline plans cannot carry today. The ratio matters more than any single number. Seven to one from F3 to E5 means moving a thousand people one tier is worth more than several points of discount across the entire estate.
First check. A warehouse operation, two thousand shift workers, sharing tablets, never opening desktop Office, needing Teams and a payslip portal. What should they hold? A, E3, because everyone in the company should hold the same plan for simplicity. B, a frontline F plan, which is built for shared device short session work at roughly a fifth of the E3 rate, with the exceptions handled individually. C, E5, because warehouses carry security risk. D, E1, because they do not need desktop apps. Pause it. D is defensible and it is still not the best answer, so as you think, ask what the device and the session length imply about the plan that was designed for this.
The answer is B, and D deserves real credit because it identifies the correct fork: these people do not need desktop applications. It is still the wrong plan, because the frontline tiers were designed for precisely this shape of work and they price accordingly. Roughly eight dollars against roughly thirty six, across two thousand people, on a single decision. A is the simplicity argument, and I want to be careful here because operational simplicity is a genuine value. It is just regularly used to justify several hundred thousand a year, which is a poor exchange rate for administrative convenience, and nobody ever writes the cost of simplicity down where the person choosing it can see it. C confuses risk with plan tier. Warehouse operations do carry risk, and the response is targeted controls on the accounts holding sensitive access, not an estate wide upgrade. And the discipline is in the second half of that answer: handle the exceptions individually. A frontline population always contains supervisors and planners who need more, and naming those fifty is what makes the other nineteen hundred and fifty defensible.
Personas, not averages, and here are the six that recur in almost every estate. Knowledge worker: desktop Office, Teams, email, ordinary data, E3 baseline, and usually the largest group by a wide margin. Security or compliance owner: Defender, Purview, eDiscovery, insider risk, and E5 is genuinely right for these people, though this group is far smaller than the estate wide E5 population tends to be. Analyst: Power BI Pro and advanced analytics, so either E5 or E3 with Power BI Pro attached, and the arithmetic between those is a per persona calculation rather than a policy. Voice user: Teams Phone and a calling plan, same structure, and the answer depends on how many other E5 components that person would actually use. Frontline worker: shift, retail, field, shared devices, F tier, and typically ten to twenty percent of the seats sitting on E3 in estates that have never run this exercise. And the deskless approver, the light reader who signs things and never opens desktop Office, who is the most consistently over licensed person in any organisation.
Guest analyst I worked with a logistics group a couple of years ago where the licensing policy was one line long: everybody gets the same plan. And I want to be fair about how they got there, because it was not laziness. They had been through a messy audit years earlier, they had been told that inconsistency was what created exposure, and they had drawn the reasonable conclusion that uniformity was safety. So thirteen thousand people, all on the enterprise tier, including about four thousand who worked in depots and never touched a desktop computer at work. The first thing we did was not an argument. It was a query. Accounts with no desktop Office activity in ninety days. It took the administrator about an hour, and it came back with just over four thousand names, which was almost exactly the depot population, and nobody in the room disputed it because it was their own data. What changed the conversation was that the head of IT, who had been braced for a fight about security, looked at the list and said, none of these people should have been on this plan, and I have been defending it for four years. The mix correction was worth close to a million a year. But the thing he said afterwards was the more useful one. He said the policy had never actually been reviewed, it had only ever been inherited, and nobody had asked for the data because everyone assumed somebody already had.
Inherited rather than reviewed, and one query settled four years of assumption. That query is in your homework. Second check.
Check two. You propose moving two thousand five hundred seats from E5 down to E3, and IT objects that this reduces the security posture. What is the right response? A, withdraw the proposal, security objections end the conversation. B, ask which specific E5 security features those two thousand five hundred accounts have used in the last ninety days, because between twenty five and forty five percent of E5 seats typically use nothing beyond what E3 already carries, and the answer converts an opinion into a list. C, move them anyway, cost is procurement's decision. D, offer to move half as a compromise. Pause it, and notice while you think that the objection is testable, and that the test is a report rather than an argument.
The answer is B. The objection is legitimate and it is answerable, which makes it the best kind of objection you can receive. Across the Microsoft 365 estates we have benchmarked, between twenty five and forty five percent of E5 seats used no security or compliance capability beyond what E3 already carries. That is not an argument against E5. It is an argument for knowing which seats are which, and your admin centre already knows. A treats a testable claim as a veto, which is how unexamined cost survives for a decade inside an organisation that considers itself rigorous. C is worse in the opposite direction: a licensing change that removes real protection from real accounts is a bad outcome regardless of what it saves, and procurement does not get to make that call alone. D is the compromise that satisfies nobody, because it takes half the saving and leaves the same unanswered question about which half was correct. Run the report, produce the list, let the list decide. That is also how procurement earns the right to be believed next time it raises something.
The upgrade paths, and the important structural point is that the map is not symmetrical. Going up is a purchase. Coming down is a negotiation. Up is easy and priced two ways: a move up a tier can be bought as a step up SKU against the existing licence, or as a fresh licence at the new tier, and those price differently, so ask for both on the quote. Upward movement is frictionless by design, and that is a fact about the commercial model rather than about your requirements. Down needs a right agreed in advance. Reducing tier or count mid term is governed by your agreement, not by your intentions, and a step down right negotiated at signature is the thing that makes a later mix correction possible at all. Without it, the mix you commit to on day one is effectively the mix you carry for the term. And the third one, which module four will keep returning to: Copilot requires an E3 or E5 base and frontline plans cannot carry it, so a Copilot pilot among frontline staff silently implies a base plan upgrade first.
What the mix is worth, and this is a real worked shape from benchmarked estates rather than an illustration. Ten thousand users. Before, on a heavy E5 default: six thousand E5, three thousand five hundred E3, five hundred F3. After, mapped to actual personas: two thousand five hundred E5 on the roles that genuinely use the security and compliance delta, five thousand E3 as the knowledge worker baseline, and two thousand five hundred F3 for frontline and deskless roles. The effect on the annual list bill is roughly twelve to eighteen percent lower. On a five million dollar list bill that is in the order of six hundred to nine hundred thousand dollars a year, and it holds for the term. Read the last two rows carefully, because here is the property that matters: that reduction happens before the discount conversation and it survives the discount conversation. It is a change to what you are buying, not to what you are paying for it. Which is the whole argument for why the mix outranks the rate.
Last check. Your account team offers two extra discount points if you standardise the whole estate on E5. Your persona work says two thousand five hundred of ten thousand actually need it. What is the position? A, accept, two points across ten thousand seats is a large number. B, decline, because the mix change is worth twelve to eighteen percent of the list bill while the discount offered is two points on a much bigger basket, so the offer is an expansion wearing a saving's clothing. C, accept for three years and correct the mix at renewal. D, counter with three points. Pause it, and price both options in cash rather than in percentages before you answer.
The answer is B. This is the standardisation offer, it arrives at most large renewals, and it is a genuinely good offer for the party making it. Price both sides in cash. Standardising ten thousand seats on E5 raises the basket substantially, and two points off a substantially larger number is arithmetically worse than a persona aligned mix at your existing rate. A fails because it compares a percentage against nothing at all. D is the same error with more effort attached, and it concedes the shape of the deal while arguing about its size, which is the most common way a negotiation is lost politely. C is the most seductive and the most expensive, because it assumes a mid term correction that your agreement may not permit without a step down right, and because three years of the wrong mix is exactly the outcome the offer was designed to secure. Bring the persona mapping into the meeting, price both options per seat, and put the two numbers side by side in writing. Numbers on a page are much harder to talk past than positions in a room.
Building the map for your estate, three steps, and this is a data exercise rather than a debate. One, export what is assigned: every account, the plan it holds, last activity, and the workloads it actually uses, from your own admin centre rather than from the vendor's proposal. That export is the foundation of the rest of this module and most of module seven, so it is worth getting properly. Two, map personas to the data: four to six personas defined by the work rather than by the department, every account assigned to one, and produce the count. The gap between that count and your current plan distribution is the size of the prize in seats, before anybody prices anything. Three, price three scenarios: current mix, persona aligned mix, and the vendor's proposed mix, all at the same rate, three totals on one page. On a ten thousand seat estate give this a fortnight of an analyst's time. The output holds for the whole term and it is reusable at every true up.
Session eleven, three sentences. One: the stack answers three questions, desktop apps or not, how much security and compliance, and whether the person sits at a desk at all, and those three answers place almost every user in your estate without opening a feature matrix. Two: list runs roughly seven to one from the frontline tier to E5, so allocating people to the right plan moves more money than any discount you are likely to negotiate. Three: map personas from your own assignment data and price three scenarios on one page, because a mix correction changes what you buy and therefore survives the discount conversation instead of being traded away inside it. Next session takes the single biggest fork on that map and prices it properly.
Homework, about an hour, and this week you draw your own map. One, pull the assignment export: every account, its plan, and last activity, and if you cannot get it this week then find out who can and put a date on it, because everything downstream depends on it. Two, count the plans, how many on each tier today, and most people are surprised by that number before any analysis begins. Three, name your personas, four to six, defined by the work, with one sentence each, written clearly enough that somebody else could apply them to the same export and reach the same answer. Four, find the deskless: accounts on an enterprise tier with no desktop Office activity in ninety days. That single query typically finds ten to twenty percent of an estate, and it is the query from the story earlier. Five, price two scenarios, current mix and persona aligned mix, at today's rates, one page, two totals, and the difference written at the bottom.
Five reads before next session, all free on redress compliance dot com. First, E3, E5, and F3, picking the right plan per persona, which carries the full allocation arithmetic and the worked mix shift we walked through. Second, the CIO playbook on selecting the right Microsoft 365 enterprise plan, which is the same decision with the governance wrapped around it. Third, the Microsoft 365 F SKU licensing guide, covering the frontline tiers in detail including where F1 and F3 diverge, which matters more than most estates realise. Fourth, Microsoft 365 Business against Enterprise, on the three hundred seat cap and what it means for acquired subsidiaries that arrive holding Business plans. And fifth, the Microsoft licensing guide, for where the plan stack sits inside the wider programme. Next session is E3 against E5, priced. The security, compliance, identity, and voice delta valued component by component against what you already own, and the retirement test that decides whether the step up actually pays. See you there.