MACC and Azure spend under the agreement: how the commitment works, how it interacts with the licensing lines, and the forfeit risk. Three knowledge checks along the way, and 1 clip from a senior cloud advisor.
The presenter in this session is an AI generated avatar. The curriculum and guidance are real, produced by Redress Compliance analysts from our consulting engagements and market network.
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 twenty one of forty, and this opens module five. Module four was Copilot, top to bottom. Module five is everything else riding on the agreement, and we start with the line that is usually the largest single number in a Microsoft deal and gets the least scrutiny, which is the Azure consumption commitment. I will give you the framing straight away, because it explains why this line goes wrong so consistently. A budget is a limit you try not to exceed. A commitment is a floor you have to reach. Those are opposite disciplines, and almost everybody who works with cloud spend has spent their career on the first one. So the instincts are all pointed the wrong way. Across roughly twenty five to thirty five enterprise agreements carrying one of these, the commitment number was the most consequential and least scrutinised figure in the entire deal, which tells you where the attention should go.
Five takeaways. One, what it is: a contractual spend floor, where you promise to consume a set Azure amount over the term and Microsoft prices the relationship around that promise. Two, the core risk: a shortfall payment when you commit to more than you consume, and the discount is easy to celebrate at signature while the shortfall is easy to forget until the term closes. Three, how to size it: trailing twelve month actuals plus funded growth, with speculative workloads discounted heavily, rather than the account team's forward model. Four, what counts: eligible marketplace purchases can count toward the commitment, and that was missed in roughly two out of five estates, which changes the arithmetic materially. Five, who watches it: no named owner tracked burn down in about sixty percent of cases, so shortfall risk surfaced only near term end, at the point where nothing can be done about it.
What a MACC actually is, three properties that decide how you treat it. A spend floor rather than a budget: you promise to consume a set amount and Microsoft prices around that promise, and as I said, a budget is a limit you try not to exceed while a floor is an amount you have to reach. That inversion catches people out repeatedly. Terms run one to five years, usually with an annual or a full term number, and the longer the term the more work the forecast is doing, because a forecast about cloud consumption three years out is a fundamentally different kind of object from a forecast about seat counts. And unmet commitment converts to a payment, which is the shortfall: money paid for consumption that never happened, at the end of a term, with nothing received in exchange. That is why sizing discipline is the whole game here and the discount is genuinely secondary, which is the reverse of how these get negotiated.
Sizing it from your own data, line by line. Trailing twelve month actuals: the foundation, take it as read, because it is the only number in the model that has already happened. Funded growth: add it where a budget and an owner exist, and funded means somebody holds the money, not that somebody holds the intention. Speculative workloads: discount heavily or exclude, because migration plans slip and a slipped migration is a shortfall. Eligible marketplace spend: add it, and check the eligibility rules, because it was missed in roughly two out of five estates and it counts toward the commitment. And the vendor forward model: treat that as an input, never as the answer. Two numbers to hold beside each other. Initial commitments were oversized by fifteen to thirty percent against trailing consumption in more than half the deals reviewed, and commitments sized on the vendor's forward forecast overran trailing twelve month consumption by twenty two to thirty eight percent.
First check. The account team's model shows Azure spend growing forty percent a year. Your trailing twelve months grew twelve percent. What do you commit to? A, the forty percent model, it reflects the transformation plan. B, trailing actuals plus growth that is genuinely funded, because forward models overran trailing consumption by twenty two to thirty eight percent and the gap converts to a shortfall payment. C, the midpoint between the two, as a compromise. D, the forty percent model, but ask for a bigger discount. Pause it, and while you think, ask yourself who pays if the transformation plan slips by two quarters, which transformation plans routinely do.
The answer is B, and the answer to who pays is you, in full, at the end of the term, for capacity you never used. Commitments sized on the vendor's forward forecast overran trailing twelve month consumption by twenty two to thirty eight percent, and more than half of initial commitments were oversized by fifteen to thirty percent against trailing consumption. Notice the word funded in that correct answer, because it is doing real work: funded means somebody holds the budget and owns the delivery date, not that a slide exists showing the workload moving. C splits the difference between a measurement and a projection, which produces a number with no basis in either of them. D is the most expensive answer on the slide and also the most tempting, because it converts an oversized commitment into a discount conversation, and a deeper discount on spend you never make is worth precisely nothing. Size low. Increases mid term are easy, and that asymmetry is coming up shortly.
What counts toward it, five things to check before you size anything. Eligible marketplace purchases: third party software bought through the Azure marketplace can count toward the commitment, and this was missed in roughly two out of five estates. Not everything in the marketplace qualifies, though, because the eligibility rules are specific and published, so check the actual rules against your actual purchases rather than assuming in either direction, since both assumptions are expensive. Copilot consumption: pay as you go credits decrement the Azure commitment, from session seventeen, so part of your AI spend is already helping you burn down. What does not count: establish that explicitly at signature, because the gap between what you assumed counted and what does gets discovered late. And which entity is spending, because in a group structure subsidiary spend may or may not count depending on how the agreement is written. That first point is the cheapest win in this session.
Guest analyst The most avoidable eight figure problem I have watched develop was an Azure commitment at a European retail group. Three year commitment, sized in a good faith workshop with the account team, built on a migration programme that everybody genuinely believed in at the time. And then the thing that always happens happened, which is that the migration went slower than planned. Not dramatically. About two quarters behind, which in a three year programme is unremarkable. What made it expensive was that nobody was watching the number. There was no owner. Cloud consumption reporting existed, it was good reporting, but it answered the question how much are we spending, and nobody was asking the different question, are we on track against what we promised. Those sound similar and they are not. They found out at month thirty one, when somebody in finance was preparing for the renewal and worked it out, and by then the projected shortfall was in the millions with five months to close it. Now here is the part I want people to take away. When we went back through it, two things could have fixed it entirely if they had been seen at month eighteen. They were making substantial marketplace purchases that were eligible and simply not being counted toward the commitment. And there was a workload scheduled for year three that could have been pulled forward into year two at no extra cost. Both fixes existed the whole time. Neither was available at month thirty one. The shortfall was not caused by the migration slipping. It was caused by nobody holding one chart.
Both fixes existed at month eighteen and neither existed at month thirty one. One chart, one owner. Second check.
Check two. You are eighteen months into a three year commitment and nobody has tracked burn down. What is the risk? A, none, the total will resolve itself by term end. B, shortfall risk that surfaces too late to correct, which is what happened in about sixty percent of cases where no named owner tracked burn down against the commitment curve. C, only that you might exceed the commitment. D, the risk is real but there is nothing to do until term end. Pause it, and ask yourself what you could still do at eighteen months that you could not do at thirty three, because that gap is the answer.
The answer is B. At eighteen months you can still move workloads forward, bring eligible marketplace purchases inside the commitment, accelerate a migration that was scheduled late, or open a renegotiation with enough term left for it to conclude properly. At thirty three months you can do none of those and the shortfall is simply arithmetic. That difference is the entire value of tracking burn down, and no named owner tracked it in about sixty percent of the cases we reviewed, which is exactly why shortfall risk surfaced near term end when it had already hardened. D is the belief that produces that outcome and it is wrong on the facts, because mid term is precisely when the levers exist. C inverts the risk, since exceeding a commitment means consuming more than you promised, which costs nothing beyond the consumption itself. And A is the assumption that a number nobody is watching will land where it should, which is the assumption behind the story you just heard.
Up is easy, down is hard, three consequences of a one way door. Conservative sizing is cheap insurance: if you undershoot you increase, and increasing is easy because it is more revenue for them, while if you overshoot you are asking to be released from a promise, which is a favour rather than a transaction. That asymmetry alone should decide your initial number, before any other consideration. The discount is not worth the lockup: a larger commitment buys a better rate and the arithmetic only works if you consume it, so model the shortfall scenario beside the discount scenario, both at your own numbers, and put them side by side in the paper where the decision gets made. And ramps exist and go unrequested: a low first year floor stepping up across the term matches how migrations actually land, and buyers without ramp terms overcommitted by fifteen to twenty five percent in year one, which is a pure sequencing loss and entirely avoidable. Ask for a mid term re baseline right too.
Burn down governance, what the monthly review actually looks like, and it is one chart. Consumption to date against the straight line to the commitment: above the line, keep going, below it, act now. Projected term total against the committed number, because a gap forecast is a mid term negotiation rather than a year three problem. Eligible spend not yet counted, against the eligibility rules, which is where marketplace purchases that should be inside the commitment get found. Workloads scheduled but not landed, against their planned dates, because a slipped migration is a shortfall in slow motion and it is visible long before it is a number. And the named owner against their standing monthly slot, because nothing above happens without that row. A named finance or procurement owner tracking burn down monthly is the whole control. The estates that did this did not have shortfall surprises. That is not a sophisticated finding and it is the difference between two very different term ends.
Last check. Your burn down chart at month twenty of thirty six projects a twelve percent shortfall. What do you do? A, wait, cloud consumption is volatile and it may recover. B, act now: check eligible marketplace spend that is not being counted, pull forward funded workloads, and open a mid term conversation while there is still term left to renegotiate against. C, accept the shortfall payment as the cost of the discount. D, increase consumption artificially to meet the floor. Pause it. Sixteen months of term remain, which is enough time for three of these four to be genuine options rather than theoretical ones.
The answer is B, and notice the order inside it, because the order is the advice. Start with the cheapest move, which is checking whether eligible marketplace spend is already happening and simply not being counted, since that was missed in two out of five estates and costs nothing to correct. Then pull forward funded work that was scheduled late. Then, with those two done and the residual gap known, open the conversation, because a vendor would much rather restructure a live commitment than collect a shortfall from a customer they intend to renew. A is the reasoning that turns a twelve percent gap at month twenty into a settled fact at month thirty six. C accepts a payment for nothing while the term is still running and the levers are all still available. And D is the worst answer on the slide: consuming for the sake of the meter converts a shortfall payment into the same money spent on infrastructure nobody needed, and that infrastructure usually outlives the term.
The commitment method, three steps, and the third is the one that gets skipped. One, build from actuals: trailing twelve months, plus funded growth with owners and dates, plus eligible marketplace spend, minus a heavy discount on anything speculative. That is your number, and the vendor's model is something to compare it against rather than a starting point to negotiate down from, which is a meaningful difference in posture. Two, buy the shape: a low first year floor with a ramp across the term, and ask for a mid term re baseline right, because buyers without ramp terms overcommitted by fifteen to twenty five percent in year one and the ramp costs nothing to request while the commitment is still being set. Three, name the owner: one person in finance or procurement, a monthly slot, a burn down chart. That is the step missing in about sixty percent of cases and it is what keeps every other decision here influenceable.
Session twenty one, three sentences. One: a MACC is a spend floor rather than a budget, and an unmet commitment converts to a shortfall payment, which makes sizing discipline the whole game and the discount a secondary consideration. Two: size from trailing twelve month actuals plus genuinely funded growth plus eligible marketplace spend, because forward models overran trailing consumption by twenty two to thirty eight percent and more than half of initial commitments were oversized by fifteen to thirty percent. Three: increases are easy and decreases are hard, so commit low, ask for a ramp, and name one owner to track burn down monthly, which is the control missing in about sixty percent of cases. One connection worth carrying forward: Copilot pay as you go decrements this commitment, so those two lines belong in the same monthly review rather than with two teams who never compare notes.
Homework, about an hour, and this week you find your burn down. One, find the commitment: the committed number, the term, and the end date, and if it takes more than a day to establish then that is finding number one and worth reporting on its own. Two, plot consumption to date: actual spend against a straight line to the commitment, one chart, and you will know within minutes whether you have a problem. Three, check the marketplace question: are you making eligible marketplace purchases that are not being counted, because two out of five estates were. Four, list the unlanded workloads: anything in the sizing model that has not migrated yet, with its planned date, because those are your shortfall in slow motion. Five, name the owner: one person, one monthly slot, and if nobody owns it today then propose yourself and put the first meeting in the calendar this week.
Five reads before next session, all free on redress compliance dot com. First, the Azure MACC sizing guide, which carries the full sizing method, the eligibility rules, and burn down governance. Second, MACC commit to consume negotiation, on the negotiation itself including ramps and shortfall terms. Third, the MACC mid term renegotiation playbook, which is exactly what to do when the burn down chart shows a gap with term still remaining, and that is the most actionable of the five for anybody already inside a commitment. Fourth, Copilot Credits and the Azure commitment, on how the AI consumption line interacts with the commitment. And fifth, Azure licensing and cost optimisation, for the consumption side, which is what the commitment is ultimately a promise about. Next session is Windows, servers, and core licensing: core based licensing, CALs, and Azure Hybrid Benefit, the on premises estate that still funds a large part of the bill. See you there.