The operating model that keeps the meter, the commitment, and the rewards loop honest, every month. 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 ten of thirty, and the module two finale. Look at what this module handed you: the meter and its units, the commitment and its clauses, BYOL and its ledger, and the rewards loop with its napkin. Four levers, each worth real money. Today's session is about the uncomfortable truth underneath all four: none of them work in an ungoverned estate. The sizing method needs a baseline, the ledger needs maintenance, the rewards need a redemption cadence, and the negotiation calendar needs a distance number, and every one of those is a governance artifact. So today we build the operating model: compartments, tags, budgets, alerts, and the two meetings that run the whole machine for about thirty hours a year. It's the least glamorous session in the module and, over a three year horizon, quite possibly the most valuable. Let's make governance pay.
Five takeaways. One, the frame: cost governance is a commercial control, not an ops chore, and I'll prove that claim lever by lever in about two minutes. Two, the structure: compartment design and tag policy that make every dollar attributable on the day it's spent, by design, instead of reconstructed by archaeology at quarter end. Three, the controls: budgets, alerts, and quotas, the guardrails that catch drift in days instead of quarters, and the routing rule that decides whether any of them matter. Four, the rhythm: the monthly hour and the quarterly sitting, with the exact agendas, because vague meetings decay and precise ones survive. And five, the pack: module two's entire discipline compressed into one operating pack, five artifacts, two meetings, four escalation plays, that you could hand to a brand new team and have them running the estate's OCI commercials by Friday.
First, the claim that governance is where the savings come from, argued from this module's own material. Session seven: the commitment sizing method starts from measured consumption by service family, trailing six to twelve months. An ungoverned estate cannot produce that baseline, so it guesses, and you know from the forfeit arithmetic what guessing costs. Session eight: the BYOL ledger and the quarterly toggle audit are pure governance artifacts, and without them the biggest discount in OCI degrades into the advisor's double payment, six figures a year, in either flavor. Session nine: the rewards loop needs accrual tracking, a redemption cadence, and an expiry alarm, and in ungoverned accounts, free money quietly times out at month twelve. And session five: the negotiation calendar runs on the distance to commit, monthly, because the estate that knows its number engages Oracle early and negotiates, while the estate that doesn't gets processed at renewal week. Four levers, four dependencies, one conclusion: governance isn't overhead on the savings. Governance is where the savings come from. Everything else today is just the mechanics of making that true.
The structure, and the design principle is one word: attribution. Every dollar attributable on the day it's spent. Compartments first: design the tree by who answers for the spend, teams and environments, not by network topology, because cost analysis groups by compartment, and you want that grouping to match the org chart that actually pays the bill. Production and non production separated, always, because half your cost questions start with which one was that. Tags second, and the rule of three: owner, project, environment, as defined tags with allowed value lists, applied at creation. Defined tags, not free text, because free text rots, prod, production, and PROD are three different strings and one analyst's lost afternoon. Third, enforcement by mechanism, not memo: tag defaults on compartments apply automatically, and policy can require tags at creation. Any tagging standard that relies on humans remembering is a standard with a half life of one quarter. Fourth, quarantine the untagged: a weekly report of untagged resources, routed to compartment owners, with a deadline, because session six's law, untagged spend is unexplainable spend, needs an enforcement loop to be a law rather than a poster. And fifth, timing: designing this in week one is a meeting, retrofitting it onto a grown estate is a project. If your estate already grew wild, the retrofit is still worth it, and it's this quarter's project, not never.
First check. An estate has one root compartment with every resource in it, tagged sporadically according to a memo from last spring. The bill jumped eighteen percent and nobody can say which project drove it. The fix that actually holds: A, a sternly worded tagging policy document. B, compartments redesigned around accountability, defined tags with enforced values and defaults, and a weekly untagged report with named owners. C, exporting the bill to a spreadsheet and annotating it by hand each month. Or D, buying a third party cost tool and pointing it at the ungoverned account. Pause here. What survives contact with humans: memos, spreadsheets, or defaults?
The answer is B, and the principle is worth stating cleanly: attribution must be produced by the platform at creation time, not reconstructed by people after the fact. Defaults and required tags do the work automatically, compartments make the grouping match accountability, and the weekly untagged report catches the leaks with a named owner and a deadline. Now the failure modes, each instructive. A, the policy memo, fails because nothing enforces it; it's the same memo from last spring that produced sporadically tagged in the first place. C, the hand annotated spreadsheet, technically works and costs an analyst's week every month, forever, which is exactly the toil the meter was supposed to eliminate; archaeology is not a process, it's a punishment. And D, the cost tool on an ungoverned account, is the subtle one: tools visualize whatever structure exists, so pointing one at chaos gives you beautifully rendered chaos. Structure first, tools second, and honestly, once compartments and tags are right, the native cost analysis takes you a very long way before any tool is needed at all.
The controls, five of them, from measurement to enforcement. One, budgets per compartment: a monthly budget on every spending compartment, sized from the trailing average, alerting at eighty and one hundred percent of forecast. Nothing fancy, just tripwires with names attached. Two, the estate budget, the one that matters most: a single budget tracking total burn against the commitment's monthly run rate. That is the distance to commit, automated, the number sessions seven and nine keep asking for, produced continuously instead of computed annually in a panic. Three, anomaly alerts on unusual daily spend, because they catch the leaked egress, the runaway job, the wrong toggle, in days. Recall the advisor's bill review: forty percent growth, entirely visible, invisible for eighteen months. Anomaly detection turns eighteen months into a week. Four, quotas where it hurts: compartment quotas capping what can even be provisioned, GPU shapes, the expensive services, non production sprawl. A quota is a budget that enforces itself, and for dev compartments it's kinder than the alternative, which is the quarterly orphan hunt finding what the interns left running. And five, the rule that decides whether any of this matters: every alert lands with a named owner, not a distribution list. Session nine said it about rewards, and it generalizes to everything: detection is cheap, routing is the control. An alert nobody owns is a log entry with ambitions.
The rhythm, two meetings, and I'll give you the agendas verbatim because precise meetings survive and vague ones die. The monthly hour: cost analysis by compartment and tag, top movers explained. The distance to commit, from the estate budget. Rewards: earned, redeemed, expiring, three numbers. The five surprise meters walked, egress, volume performance, toggles, cross region, orphans. And the untagged report, trending to zero. Output: one page, the estate report, burn, distance, rewards, findings. One page, every month, no exceptions. The quarterly sitting, same hour, deeper: the BYOL reconciliation, cloud ops brings the toggle report, licensing brings the entitlement ledger, they reconcile, the advisor's standing meeting from session eight. Support status verified on allocated sets. The orphan hunt. The policy snapshot diff from session four. And commitment posture: trajectory versus commit, true up trigger checked, distance to the renewal calendar noted. Output: ledger updated, toggles corrected, findings priced. Plus annually, the portfolio review from session eight, feeding next year's sizing per session seven's method. Total cost of the entire apparatus: about thirty hours a year. I've watched single findings from one quarterly sitting pay for a decade of those hours. That's the pitch, and it shouldn't be a hard one.
Check two. A team proposes dropping the monthly review, because a new dashboard now shows spend in real time, so the meeting is redundant. The accurate objection: A, none, a live dashboard supersedes a monthly meeting. B, dashboards display, they don't decide: the review exists to act, on distance to commit, expiring rewards, toggle findings, untagged spend, and unowned dashboards decay into wallpaper. C, dashboards are inaccurate, only invoices count. Or D, fine, but the meeting should be weekly instead. Pause here. What did the dashboard decide last month?
The answer is B. The monthly hour was never a data delivery mechanism, it's a decision mechanism: someone commits to opening the true up conversation, someone redeems the rewards expiring in sixty days, someone chases the untagged compartment, someone prices the toggle finding. Those are verbs, and verbs don't come from screens. Here's the empirical version: every war story in this module featured an estate where the data was available the entire time. The advisor's client had cost analysis sitting unread for eighteen months while forty percent accumulated. What those estates lacked was never data, it was an owned hour where a named person had to look and act. That's what the meeting is. Keep the dashboard, genuinely, it makes the hour faster and the anomaly alerts sharper. But the hour is the control, and the dashboard is its instrument, not its replacement. And D, making it weekly, over corrects into noise: weekly reviews of monthly signals train everyone to ignore them. Monthly decides, quarterly reconciles, annually re-plans. The cadence is the design.
Now the money flows around the program, because governance that isn't funded politically dies faster than governance that isn't funded financially. One, showback first: publish monthly cost by compartment and tag to the teams that spend it. Just visibility, no invoices, and behavior changes anyway, nobody wants to top the leaderboard for orphaned volumes two months running. Two, chargeback only where it matters: full internal billing suits mature estates, but premature chargeback breeds tag gaming and shadow spend, teams optimizing the attribution instead of the architecture. Showback with an escalation path covers most estates for years. Three, the rewards routing, session nine's budget question, and it must be settled in writing: who benefits from redeemed rewards, who gets credited for the consumption that earned them. One paragraph, agreed with finance, or the program starves in a turf war, and I keep repeating this because turf kills more rebates than arithmetic ever has. Four, the commitment's owner: the commit is an estate level obligation, but individual teams only ever see rates, not forfeits. One named person carries the distance to commit, or effectively nobody does. And five, the habit that funds everything: savings get receipts. Every finding, the toggle fix, the orphan purge, the rescued rewards, the early restructure, written down with its annual value, in a findings ledger. That ledger is about to star in the last check.
Before the check, here's our advisor on what this looks like when it's real, the governance pack he actually hands to clients.
Guest analyst When I finish an OCI commercial engagement, I leave the client with a pack, and it is deliberately thin. Five documents and two calendar invites. The documents: the consumption baseline by service family. The commitment order with every negotiated clause highlighted, because the clauses you forget you have are the clauses you fail to invoke. The BYOL ledger, with decommission dates. The rewards identity on one page, consumption times rate against the bill, with the current coverage percentage. And the findings ledger, every fix with a dollar value and a date. The invites: the monthly hour and the quarterly reconciliation, each with a named chair and the agenda in the invite body, so the meeting survives personnel changes. Here is why thin matters. I have seen fifty page governance frameworks die in the drawer they were printed into. The pack survives because every page answers a question somebody actually asks: what do we spend, what did we promise, what do we own, what are we owed, and what have we saved. Five questions, five pages, two meetings. When the CFO asks any question about Oracle cloud money, the answer is in the pack, dated this month. That is what governed looks like, and it is genuinely achievable by February for an estate that starts in January.
Five questions, five pages, two meetings: what do we spend, what did we promise, what do we own, what are we owed, what have we saved. If your estate can answer those five, dated this month, you're governed. If it can't, the gaps are your homework list.
Last check of module two. After a year of the rhythm, the findings ledger reads: toggles fixed, one hundred eighty thousand a year. Orphans purged, forty thousand. Rewards rescued from expiry, ninety five thousand. A forfeit avoided by an early restructure, three hundred thousand. Leadership asks whether the governance hour is worth continuing. The answer: A, unclear, savings are anecdotal. B, six hundred fifteen thousand dollars of receipted findings against roughly thirty hours of meetings: the ledger is the answer, and it also prices the estate's credibility in the next commitment negotiation. C, the program paid off, so it can safely stop now. Or D, governance should be outsourced now that the process is proven. Pause here. What does the ledger add up to, and what does stopping restart?
The answer is B, and this check is really about why receipts exist. The worth it question arrives annually, reliably, usually during budget season, and the findings ledger answers it in one line: six hundred fifteen thousand against thirty hours, which makes the monthly hour the best paid meeting in the company by a comfortable margin. Without receipts, the honest answer is A, anecdotal, and anecdotal programs get cut. Now C, stopping because it worked: stopping restarts the decay that produced the findings. Toggles drift when templates change, orphans accumulate wherever developers exist, rewards expire on their own clock, the same physics that made the money runs continuously, and governance is the friction against it, remove the friction and the physics resumes. And D, outsourcing: a provider can absolutely run the mechanics, the reports, the reconciliations, but the decisions in the escalation plays, the true up, the restructure, the budget question, are the estate's own commercial posture, and posture can't be rented. Keep the rhythm, keep the ledger, and let the ledger renew the mandate every year. That's self funding governance, and it's the only kind that survives.
The module two operating pack, the whole module in one handover, per the advisor's format. The five artifacts: the consumption baseline by service family, session six. The commitment order with its clause checklist highlighted, session seven. The BYOL ledger with decommission dates, session eight. The rewards identity and this month's coverage percentage, session nine. And the findings ledger with receipts, today. The two meetings: the monthly hour, burn, distance, rewards, meters, tags, one page out. The quarterly sitting: BYOL reconciliation, commitment posture, policy diff, ledger out. Named chairs, agendas in the invite body. And the four escalation plays, pre agreed so they execute calmly: true up trigger hit, run session seven's negotiation, whole relationship on the table. Forfeit projected, session seven's rescue, opened early while the churn risk is Oracle's problem too. Rewards expiring, redeem this week, no meeting required. Toggle mismatch, session eight's afternoon fix, ledger updated. That's module two, complete: the meter, the commitment, BYOL, the rewards loop, and the governance that keeps all four honest. A team that holds this pack can run an OCI estate's commercials, and next module, we take the same discipline to clouds Oracle doesn't own.
Session ten, three sentences. One: governance is a commercial control, every lever this module built, sizing, BYOL, rewards, negotiation timing, consumes numbers that only compartments, tags, budgets, and owned reviews produce, so governance isn't overhead on the savings, it's the source of them. Two: two meetings run the machine, the monthly hour that decides and the quarterly sitting that reconciles, thirty hours a year, and the findings ledger with receipts is what proves the program and renews its mandate every budget season. Three: dashboards display, reviews decide, and every alert routes to a named owner with a pre agreed play, or the detection was decoration. Module two is complete. Next session opens module three, and the terrain changes: Oracle software running on clouds Oracle doesn't own. We start with the document that governs all of it, the authorized cloud environment policy, how Oracle counts licenses on AWS, Azure, and Google Cloud, the two vCPUs to one license arithmetic, and what it does to every hyperscaler decision your estate will ever make. Same discipline, new map. See you there.
Homework, about an hour, and this week it's pure construction: stand up the rhythm. One, book the meetings: the monthly hour and the quarterly sitting, recurring calendar invites, named chairs, the agenda from the rhythm table pasted into the invite body, this week, before the intent decays. Two, set the estate budget: one budget tracking total burn against the commitment's monthly run rate, alerting at eighty and one hundred percent, routed to the commitment's named owner. Fifteen minutes in the console, and the distance to commit becomes automatic. Three, run the untagged report: count untagged resources today, write the number down, that's the governance program's first KPI, and its target is zero. Four, start the findings ledger: open a document and backfill every fix module two's homework already found, the toggle corrections, the orphan purges, whatever the BYOL comparison surfaced, with annual values. Most estates that did the homework honestly already have six figures on that page, which means the program is funded before its first official meeting. And five, write the four escalation plays as one liners: true up, forfeit, expiring rewards, toggle mismatch. Agreed now, calmly, so they execute later without a meeting about the meeting. That's the pack started. Module three awaits.
Five reads before next session, all free on redress compliance dot com. First, the OCI licensing and cost guide, the commercial machinery this operating model exists to govern, worth a final pass now that you've seen every lever. Second, the Support Rewards guide, the accrual and expiry mechanics your monthly hour now tracks as three numbers. Third, the Oracle BYOL comprehensive guide, the entitlement rules behind the quarterly reconciliation. Fourth, Oracle cloud contracts and credits for CIOs, the executive version of the operating pack, useful for the leadership conversation the findings ledger will eventually win. And fifth, the one that matters most for what's next: Oracle database licensing in cloud environments, the counting rules for Oracle software on other people's clouds, because that's exactly where module three begins. That's session ten, and that's module two: measured, committed, owned, rewarded, and governed. See you in module three.