EA, MCA, and CSP decided with a framework and numbers: the seat split, the break even, and the portfolio answer. 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 five of forty, and the one that closes module one. We have met all three vehicles now: the Enterprise Agreement with its three year hold and its one way true up, the Microsoft Customer Agreement where nothing expires and nothing is protected, and CSP where the partner is part of the product and the term is the product. Today we decide between them, properly, with a framework and numbers rather than a preference. And the single most important thing in this session arrives right at the start, so let me say it before anything else: decide the seat split before you ask Microsoft for a rate. Across the vehicle decisions we advised in 2024 and 2025, the cheapest outcome was almost never a single vehicle, which means the question is not which one, it is which populations go where. Today: the six step framework, the five inputs, a worked comparison on a real shaped estate, the dual track, and when to run all this. Let's decide.
Five takeaways. One, the framework: a sequence that ends in a portfolio decision rather than a preference, and it deliberately begins with your own data instead of somebody's quote. Two, the inputs: five facts that decide the answer, every one of which you can establish without speaking to a vendor at all. Three, the worked case: a five thousand seat estate priced across all three vehicles with the arithmetic shown rather than asserted, because a framework you cannot compute is a slogan. Four, the dual track: stable core on one vehicle, variable pool on another, and why that pattern returned nine to sixteen percent against a single vehicle across the decisions we advised. And five, the timing: when to run this, how long it takes, and why running it after the proposal arrives is running it too late. By the end you will have the shape of your own answer, and the homework produces the two tables that make it real.
The framework, six steps, and the order matters more than any individual step. One, segment the seats by behaviour: stable, variable, frontline, external, because the split is the decision and everything after it is just pricing. Two, measure churn per segment over twelve months, since churn rather than headcount is what decides which term fits, as session four established at some length. Three, establish the tier mix each segment actually needs, because mix moves more money than vehicle choice, which was session one's opening lesson and module three's whole subject. Four, price each segment on each candidate vehicle, so you end up with three prices per segment rather than one price per estate. Five, add the structural terms, the hold, the price protection, the reduction rights, because a rate without a term is half a comparison and the half that flatters the seller. And six, decide the portfolio, and only then take it to Microsoft. Now look at what steps one through five have in common: not one of them requires a vendor conversation. That is deliberate. A buyer who arrives at step six with the answer is negotiating the terms of their own proposal. A buyer who arrives at step one is responding to somebody else's.
The five inputs, and all five are yours to establish before anyone quotes you anything. Stable seat count: how many seats will genuinely still be there in three years, and here is the number that does the most work in this whole module, the EA threshold is five hundred users or devices, but the EA only pays off well above that, near two thousand four hundred stable seats. So this single figure picks your candidate vehicles before anything else does. Churn and shape: how much of the estate turns over, and whether it moves seasonally or steadily, because high churn punishes committed vehicles for the exact reason session two gave, the true up counts only upward. Direction of travel: growing, flat, or shrinking, and how confident you honestly are, because a shrinking estate on a three year enterprise wide commitment is paying for its own decline and no discount recovers that. And two more in the note that complete the set: the tier mix each population needs, which module three builds properly, and your appetite for administration, which is a real input rather than a soft one, because a dual track is more work than a single agreement and that work has to belong to a named person or it does not happen.
First check. Your account team asks for your seat count so they can prepare EA and MCA proposals. You have not segmented the estate yet. The right response: A, send the total headcount, the segmentation can happen while they prepare. B, do the segmentation first: a single total invites a single vehicle proposal, and the split is the decision the framework exists to make, so handing over one number concedes the shape of the deal before it is discussed. C, refuse to share any numbers until pricing is agreed. Or D, send the count and ask for a discount indication to anchor the conversation. Pause here. What shape of proposal comes back when a vendor receives one number?
The answer is B, and the mechanism is worth naming precisely: a proposal is shaped by the question it answers. Send one total and you get one enterprise wide proposal, and once that document exists it becomes the thing everybody negotiates against, including you, because it is easier to argue about a number on a page than to argue for a different page. Segmenting first costs you a few days and changes what arrives: proposals that price your populations the way those populations actually behave. A sounds efficient and quietly loses the framing, which is the most common way this goes wrong, nobody decides to concede the shape, they just answer a reasonable question quickly. C is unnecessarily adversarial and slows a process you already control, because the vendor sees your counts eventually and the leverage was never in withholding them, it was in knowing what they mean before they do. And D compounds A by anchoring on a percentage before the structure is settled, which is session one's error wearing a new coat. So the rule that closes module one, and I will repeat it until it is boring: decide the seat split before you ask Microsoft for a rate.
The comparison, worked, on a five thousand seat estate. Three thousand two hundred permanent staff: flat, high confidence over three years, and that lands them on the EA, comfortably above the break even where the rate wins by six to fourteen percent. Nine hundred seasonal: peaks for four months then leaves, so CSP monthly, where the count falls when the people do. Six hundred frontline: stable but needing a lighter tier, and either vehicle works because for this population the tier choice matters considerably more than the vehicle choice, which is module three's territory. Three hundred contractors: constant churn and external identities, CSP monthly, and here is a principle worth adopting, never let churn enter a committed count. And four hundred Copilot pilot users: adoption unproven, so CSP, where commitment risk is lower while the business case is still being made, which is directly relevant when we get to module four. Now the arithmetic that matters. Run that estate as one EA and you commit five thousand seats for three years, including twelve hundred that churn and four hundred whose value nobody has demonstrated yet. Run it as a split and you commit three thousand eight hundred. That difference is the same arithmetic that produced the nine to sixteen percent advantage in our benchmark set. Our guest analyst has run this exact play.
Guest analyst The dual track that convinced me for good was a retail group, about seven thousand seats, coming up to an EA renewal. Their existing agreement covered everything, all seven thousand, one enrollment, one commitment, which is how it had been done for nine years and nobody had questioned it. We spent three weeks on segmentation before we spoke to Microsoft at all, and the shape that came out was stark: four thousand one hundred permanent head office and management seats that had barely moved in five years, and roughly two thousand nine hundred store and seasonal seats with turnover north of sixty percent a year. They had been buying two thousand nine hundred committed three year seats to serve a population that replaced itself twice inside the term. We took the permanent core to the EA, kept the volume tier by being honest about the count, and put the store and seasonal population onto CSP monthly. First year saving was about fourteen percent of the run rate. But the part I did not expect was the second payoff. Eighteen months later they closed forty stores. Under the old structure that would have been eight hundred seats paid for until the renewal, roughly one and a half years of nothing. Under the split, those seats came off inside two months. The split paid once on the rate and then paid again on an event nobody had forecast. So the question I now ask every client before a renewal: which of your seats will still exist in three years, and which of them are you about to buy three years of anyway?
Fourteen percent on the rate, and then eight hundred seats released in two months when forty stores closed, an event nobody had forecast. The split paid twice. Which of your seats will still exist in three years, and which are you about to buy three years of anyway? Second check.
Check two. In that worked estate, the account team offers a deeper volume discount if all five thousand seats go on the EA. The extra discount is three points. The analysis: A, take it, three points across five thousand seats beats a complicated split. B, price it: three points applies to the whole base, while the split avoids committing twelve hundred churning seats and four hundred unproven ones for three years, so compare the discount value against the cost of the seats you would be committing. C, refuse on principle, splits are always better. Or D, take it and reduce the seats at the first anniversary. Pause here. Three points on five thousand seats, against three years of sixteen hundred seats you would not otherwise buy. Which number is larger?
The answer is B, and the instruction inside it is do the arithmetic rather than the instinct. Three points on five thousand seats is a percentage of your annual base. Sixteen hundred committed seats you do not need is roughly a third of that base, paid three times over. In most estates of this shape the second number dominates comfortably, and notice that this is exactly session one's discount versus mix comparison appearing in a different place, which is the pattern of this entire course: the visible percentage against the invisible quantity. Now D, which I want to kill properly, because it is the one people actually say in meetings: session two established there is no downward true up, so reducing at the first anniversary is simply not available, and if a seller implies that it is, ask for it in writing and watch the sentence change. C is the mirror image of A, a rule substituting for a calculation, and sometimes the single vehicle genuinely does win, in which case take it, knowing precisely why. The general shape to carry forward: a discount offered in exchange for committing seats you would not otherwise commit is priced by the seats, not by the percentage.
The dual track, five points, running two vehicles on purpose. The pattern: stable core on the committed vehicle where its rate wins, variable pool on the flexible one where the count can fall, and across thirty five to fifty vehicle decisions we advised, that returned nine to sixteen percent against a single vehicle. The line: draw it where behaviour changes, not where the org chart does, because the question is whether a seat will still exist in three years, not which department owns it, and those two questions have very different answers. The cost, stated honestly: two vehicles is more administration, two renewal calendars, two sets of terms, two relationships, and that work is real and needs an owner, which is module seven's subject. The review: the split is not permanent, populations move between segments as the business changes, so the line gets redrawn annually, ideally at an anniversary rather than in the middle of a renewal. And the honest exception, which I want to state clearly because a course that only sells one answer is not worth your time: for a genuinely stable, standardised, single shape estate, one vehicle is right and a split is pure overhead. That estate is rarer than most buyers assume, and step two of the framework tells you within an hour whether you are it.
Timing and sequencing. Twelve months out is ideal: the segmentation, the churn measurement, and the tier work all take real time, and the benchmark from session one applies directly, arriving a hundred and eighty days out with a credible alternative was worth five to nine points on the final discount. When the proposal arrives is late but not lost: run the framework in parallel rather than negotiating the document you were handed, and ask for the timeline you need, because a vendor who wants the deal will usually give it, and a vendor who refuses has told you something useful about how the rest of the negotiation will go. And after signature is too late for this term and exactly the right moment to start for the next one, which is why this framework belongs in your annual review rather than in the renewal panic. One more thing, in the note, and it deserves saying plainly: Microsoft is steering renewals toward CSP and the MCA for enterprise. That is a genuine direction of travel and it is also a starting position, and those two facts are compatible. The framework is simply how you decide whether their preferred destination happens to be yours, which sometimes it is.
Last check of module one. Your estate is one thousand seven hundred stable seats, flat, heavily standardised on Microsoft, with almost no churn. Which vehicle does the framework choose? A, the EA, because standardisation suits an enterprise wide commitment. B, likely CSP or the MCA: below roughly two thousand four hundred stable seats the EA rate advantage has not arrived yet, and low churn means the flexibility costs little to keep, so the committed vehicle would buy a discount tier the estate is too small to earn. C, a dual track, since splits always win. Or D, whichever Microsoft proposes, since the estate is too small to negotiate. Pause here. The EA threshold is five hundred seats. Where does the EA actually start paying off?
The answer is B, and this check exists for one distinction: eligibility is not advantage. The EA threshold is five hundred users or devices, so this estate can sign an EA, and the EA only pays off well above that, near two thousand four hundred stable seats, so this estate signing one commits three years to earn a rate that CSP or the MCA can match or beat today. Eligible and advantaged are different words. A treats standardisation as decisive when it is genuinely secondary: heavy standardisation makes an EA administratively neat, which is a real benefit and not a commercial one. C is the dual track applied without the condition that justifies it, and I put it in deliberately because the previous slide argued hard for splits, with almost no churn there is no volatile pool to separate, so the split is overhead with no offsetting saving, which is the honest exception arriving one slide later. And D is defeatism with a real fact inside it, smaller estates do have less leverage on rate, and that is precisely the argument for spending your effort on choosing the right vehicle, where the advantage is structural and available to everyone, rather than on chasing a discount tier your size cannot earn.
Module one, complete, and worth a minute on what you now hold. The landscape: who is in the transaction and where each margin sits, the five inputs that set your price, and the environment your next renewal happens inside, two price waves inside twenty four months, compressed volume tiers, and an active push away from the EA. The three vehicles: the EA's three year hold with its one way true up, the MCA's evergreen terms where nothing expires and nothing is protected, and CSP's partner channel where the term is the product and the margin is measurable with two phone calls. And the decision: a framework that ends in a portfolio rather than a preference, segment by behaviour, measure churn, price each segment on each vehicle, add the structural terms, and arrive at Microsoft with a position instead of a question. If you do nothing else from module one, do the segmentation, because every module that follows assumes you know which of your seats are permanent and which are passing through. Module two opens next week with the EA's mechanics in detail: enrollment anatomy, price levels, and the structural decisions that quietly set three years of pricing before anyone talks about a discount.
Session five, three sentences. One: the framework runs in six steps and the first five need no vendor conversation at all, because a buyer who arrives with the answer negotiates their own proposal instead of responding to someone else's. Two: segment by behaviour and price each segment on each vehicle, since a single total invites a single vehicle proposal, and the split is the decision, worth nine to sixteen percent against committing an entire estate to one machine, and worth considerably more than that when an unforecast event arrives. Three: eligibility is not advantage, the EA threshold is five hundred seats and the EA pays off near two thousand four hundred stable ones, so an estate in between should spend its effort on choosing the vehicle rather than chasing a tier it cannot earn. Module one is done. Next week, module two, and the EA's mechanics in detail. See you there.
Homework, about an hour, and this week you run steps one and two for real. One, segment the estate: stable, variable, frontline, external, four numbers from HR and identity data, and write a one line definition of each so the next person can reproduce your work, because a segmentation nobody can reproduce gets rebuilt from scratch every renewal. Two, measure the churn: adds and removals per segment over twelve months, and I will make a claim about this table, it decides more of your vehicle answer than any conversation with a vendor ever will. Three, find your break even position: stable seat count against the roughly two thousand four hundred mark, above, below, or close enough that both vehicles deserve pricing. Four, name the direction: growing, flat, or shrinking over three years, with the confidence stated honestly rather than optimistically, because a shrinking estate on a committed vehicle pays for its own decline. And five, write the one page position: which populations belong on which vehicle and why. That page is what you carry into any vehicle conversation, and it is the specific thing that stops you being handed a shape instead of choosing one.
Five reads before next session, all free on redress compliance dot com. First, the CSP versus EA pillar for 2026, which is the portfolio framing and carries the dual track data behind today's session, including the nine to sixteen percent finding. Second, the CIO playbook on evaluating Microsoft renewal proposals across EA, MCA, and CSP, which gives you the questions to ask of each vehicle when the proposals actually arrive. Third, CSP versus EA side by side, for the break even analysis dimension by dimension. Fourth, the Microsoft licensing guide, for the wider context around the decision. And fifth, the Microsoft cloud agreements and subscriptions playbook, which sits underneath the whole of module one and is worth a second read now that you have all three vehicles in your head. That is session five, and that is module one: the frame before the number, the split before the rate, and a position instead of a response. Next week, module two, the EA in mechanical detail. See you there.