The controls, the owners, and the monthly page that prevents a surprise. Three knowledge checks along the way, and 4 clips from a senior licensing analyst.
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. 4 times in the session the frame splits and a senior licensing analyst gives the view from inside real SAP 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 fifteen, and this closes module three. Over the last four sessions we have met three meters and a unit you cannot count: Data Cloud credits, the pipeline that drives them, agent conversations, and the AI lines that sit half in your edition and half on a meter. Today we put controls around all of it. And I want to be careful here, because the instinct in most organisations is to add an approval step, and that is precisely the control that will not work. So today: why your existing governance misses this entirely, where the gate actually belongs, the four owners you need to name, the monthly page, and thresholds that come with responses attached. Three knowledge checks. Let's begin.
Five objectives. First, explain why your controls miss, because licence governance catches requests and consumption is not requested, it is configured. Second, put a gate where the decision is: a short consumption check inside the change process, not a new approval board. Third, name four owners: consumption, content, commercial and configuration, and most estates can name none of them. Fourth, run one monthly page, five lines covering every meter, with a projected exhaustion date on each. And fifth, set thresholds with responses, because a number nobody acts on is not a control, it is a notification.
So, why the usual controls miss. Your governance was built to catch a request. There is no requester, because a schedule changed in an admin screen is not a request and meets no gate. There is no approval, because the person who moved the number had every right to move it. There is no transaction, so nothing to review, nothing to sign, nothing that reaches procurement. And the signal is delayed, since the effect appears in consumption reporting weeks after the cause. Let me say why adding an approval step makes this worse rather than better.
Guest analyst clip.
So the answer is not more approval. It is putting a small check where the decision is actually made, which is the technical change process, and keeping it small enough that nobody has any reason to work around it.
Right, the gate. Four questions, added to a process you already run. Does this touch a meter? Most changes do not, so most take five seconds and you want a clear no or a named meter. What is the order of magnitude? Precision is not needed, direction and scale are, so roughly a tenth, similar, or several times current. Is there a cheaper design? Frequency, filtering and scope are usually adjustable, so you want the alternative considered and why it was rejected. And who is told? The consumption owner, before it ships rather than after, named and recorded on the change. That is four lines on an existing change record. It is not a board, it is not a committee, and it fails immediately if you make it heavier than the change itself.
First knowledge check. Why does a standard licence approval process fail to control Data Cloud spend? A, because Data Cloud spend is too small to notice. B, because consumption is configured rather than requested, so nothing enters the process. C, because the approvers lack technical knowledge. D, because the vendor does not provide usage reporting. Pause here and pick an answer before you continue.
B. An approval process can only act on the things that arrive at it, and a refresh interval changed inside a working system never arrives. A is the opposite of the problem, because these lines are large and growing. C blames the people rather than the design, and the same approvers handle it perfectly well once the change gate puts the question in front of them. And D is not true, and reporting is necessary rather than sufficient, because a report with no owner and no threshold changes nothing at all.
So, four owners, named, in the same document. The consumption owner watches every meter monthly and can pause a job, and that is one person rather than a team. The content owner answers for containment and grounding quality, which is what makes AI worth its meter in the first place. The commercial owner holds the commitments, the allotments and the renewal dates, and in most organisations listening to this, that is you. The configuration owner approves changes that move a meter and is close enough to the platform to judge them. And write the names down, because a role with no name attached is an intention, and the document is the control. Let me be blunt about what happens when these are unnamed.
Guest analyst clip.
Which brings us to the page itself. One page, every meter, five lines each. Consumed against commitment, for credits, conversations, API blocks and storage, with the same five lines for each. The slope rather than the total, month on month and plotted, because a total only ever rises and tells you nothing.
Then the projected exhaustion date at the current rate, which is the line that turns data into a decision. What changed this month, so new sources, schedules, agents and segments, because that line explains every movement above it. And the decision: continue, tune, or buy, written down monthly by the person who owns the budget.
Second knowledge check. Which is the most useful single number on a consumption report? A, total consumed to date. B, percentage of the commitment used. C, the projected date the commitment runs out at the current rate. D, consumption compared against the same month last year. Pause here before you continue.
C. A date can be compared against your renewal date, which immediately tells you whether this is a negotiation or an emergency purchase, and no other number does that. A only ever rises and prompts nothing. B is the number most reports lead with, and sixty percent consumed is either comfortable or alarming depending entirely on the slope, which the percentage hides. And D is useful context in a mature estate and useless in the first two years, which is exactly when these lines cause the most trouble.
Now, thresholds, and every one of them gets a response written beside it. Exhaustion inside the term: projected to run dry before term end, and the response is to open the commercial conversation now. A slope change over twenty percent, meaning a month on month jump, where the response is to find the configuration change that caused it. Containment below plan for agent programmes, where the response is tune or narrow, with a named owner and a date. Utilisation under half on any allotment or per user AI line, where the response is a reduction case for renewal. And no threshold without a response, because otherwise it is a notification, and notifications are what people filter. Let me explain why that last rule matters more than the numbers do.
Guest analyst clip.
So write the response first and the number second. It sounds backwards and it produces much better thresholds, because you quickly discover that some of the alerts you were planning have no action behind them at all, and those ones should simply not exist.
Quarterly, then, is where the monthly page turns into a position. Roll up the four meters, one slide each: where you are, where the run out date sits, and what you propose. Compare against the forecast you built, meaning your bottom up model from session twelve, tested against reality, which is how the next one gets better. Update the renewal position, so which lines grow, which reduce and which get restructured, kept current rather than written late. Log what you learned, so the next commitment is sized from your own history rather than from somebody's benchmark. And keep it to one hour and four slides, which is possible precisely because the monthly page already did the work.
Five governance failures. A committee instead of a gate, because heavy governance gets routed around and the routing around is invisible to you. A dashboard with no owner: excellent data, refreshed nightly, opened by nobody with the authority to act on it. Thresholds without responses, so alerts that fire, get acknowledged, and change nothing about the trajectory. The technical team reporting to itself, where consumption is reviewed inside IT and never reaches the person who signs. And governance starting at the surprise, set up in month eleven after the overage, when the commitment has already been reset.
Last knowledge check. When should consumption governance be set up? A, before go live, alongside the first commitment. B, after the first quarter, once there is data to look at. C, at the first overage, when the need is proven. D, before the renewal, when the position matters. Pause here and pick an answer before you continue.
A. The cheapest month to build this is the month before anything is consuming, because the owners are still assembled, the design decisions are fresh and reversible, and nobody is defending a number yet. B loses the baseline, and the first quarter is when the steepest changes happen. C is the most common answer in practice and it is the most expensive, because by then the overage has been paid and the next commitment is being proposed from it. And D is too late by a year, and it leaves you assembling a position instead of holding one. Let me put the whole rhythm together.
Guest analyst clip.
So, what module three adds up to. At every change: four questions on the change record, answered in a minute or escalated in five. Every month: one page, every meter, the slope, the run out date, and a written decision. Every quarter: four slides to the commercial owner, and the renewal position updated. Every renewal: your own consumption history against their proposal, which is the strongest position available to you in any consumption negotiation. And all of it owned by four named people: consumption, content, commercial and configuration, written down and kept current, because the document is what survives the people moving on.
Three sentences. Licence governance was designed to intercept a person asking for something, and consumption never asks, so the control has to move to where the decision is actually made, which is four short questions on a change record rather than a new approval board. The monthly page is the whole control: consumed against commitment for every meter, the slope rather than the total, the projected exhaustion date, what changed this month, and a written decision from the person who owns the budget. And every threshold needs a response written beside it or it is only a notification, and the cheapest month to build all of this is the one before anything is consuming. That closes module three.
Homework before session sixteen, about two hours. One, name your four owners: consumption, content, commercial, configuration, with real names, and note every gap, because the gaps are the finding. Two, draft the monthly page, one page, every meter you have, five lines each, filled in as far as you can today. Three, add four questions to your change process, the gate from earlier, on the existing change record and nothing heavier. Four, write a response beside each threshold you already have, and if a threshold has no response, either write one or delete the threshold. And five, book the quarterly review in the diary with the commercial owner, before the next renewal window opens.
Five guides, all on redresscompliance dot com. Negotiating Salesforce AI and Data Cloud licensing covers the control strategies and the terms that make them enforceable. Salesforce licence optimization sets out the wider operating rhythm this session plugs into. The Salesforce AI credits consumption model describes what you are governing and how the meters behave. The Salesforce Data Cloud pricing hub collects the Data Cloud references in one place. And Salesforce hidden costs lists, line by line, what governance is there to catch.
That is session fifteen, and module three is complete. The thing to take away is that consumption governance is not heavier governance, it is governance in a different place, and four questions on a change record will do more for you than any approval board ever will. Next time we open module four on the wider portfolio, starting with MuleSoft: Anypoint Platform, the capacity model, and how integration spend scales with the estate. See you then.