Oracle OCI procurement. The five artifacts.

An OCI deal is won or lost before signature, in the artifacts on the table. Five belong there: a requirements matrix, a commitment model, reference checks, an order document clause list, and an evaluation scorecard.

This toolkit defines each artifact, who builds it, and the decision rule it enforces. Running the estate after signature is a different discipline with its own page, the OCI FinOps operating model; this one covers the buying event.

• The OCI licensing reference covers BYOL rules and OCPU counting.
• The OCI cost optimization page covers the levers that work after go live.
• The BYOL versus license included framework covers the database licensing decision inside the deal.

What belongs on the table before you sign an OCI order?

Five artifacts, each answering one question Oracle's paperwork will not ask on your behalf. Build them in the sequence below and the pricing calls take care of themselves.

The three ways OCI deals close

• Signed clean. The first quote, accepted, with no benchmark and no markup. The most common pattern and the most expensive.
• Discount memory. Last cycle's rate plus a small concession, negotiated from what Oracle remembers you accepted before.
• Artifacts on the table. Benchmarked, scoped, and papered. The only pattern that protects the multi year picture, and the one this toolkit produces.

The twelve week runway

Twelve weeks from kickoff to signature is the runway that holds. Oracle's quarter end is leverage for the buyer who arrives prepared and a trap for the one who lets the clock compress the clause list.

If the timeline slips, slip the signature, not the artifacts. A deal that must close this quarter on Oracle's calendar is a deal priced on Oracle's assumptions.

What goes into the OCI requirements matrix?

Every workload, every region, every compliance constraint, and every exit requirement, written as testable lines an Oracle response can be scored against. Adjectives are not requirements; numbers and named services are.

The sections that matter

• Workload inventory. Databases including any Exadata dependency, middleware and application servers, compute, storage, and network profiles with sizing evidence.
• Service by region availability. The OCI catalog is not uniform across regions. Verify each required service in each target region against the published region documentation.
• Sovereignty and compliance. Data residency, government or sovereign region needs, and any regulator specific attestation, each tied to a named workload.
• Deployment model. Public region, dedicated region, or Cloud at Customer; the Cloud at Customer versus OCI comparison frames that choice.
• Migration dependencies. Middleware disposition changes both cloud sizing and license need; the middleware migration business case covers the economics.

The lines buyers forget

• Exit assistance. Data egress support and a defined wind down period, priced at signature rather than at departure.
• SLA claims process. Not the SLA number, the mechanics of claiming against the published Oracle Cloud SLAs.
• Support model. Severity response expectations and escalation paths, tested against references rather than assumed.
• Disaster recovery pairing. The second region, its service coverage, and the cross region traffic the design generates.

The decision rule: a requirement that is not in the matrix cannot be scored, so it will not be priced, and it becomes a change order later. Write it now or pay for it twice.

How should the commitment model be built for the negotiation?

As a three scenario consumption model you build yourself: floor, expected, and stretch, month by month by service, with the floor as the number you commit to. The sizing philosophy and the expiry math live on the Universal Credits sizing page; the artifact itself has five tabs.

The five tabs of the model

1. Workload inventory with owners and go live dates, tied to the requirements matrix.
2. Monthly consumption curve by service, priced from the published price list before any discount conversation.
3. BYOL offsets, each backed by an entitlement record, never by an assumption.
4. Scenario switches: floor, expected, and stretch, driven by migration dates that can slip.
5. Sensitivity: what each ninety day slip does to year one consumption, because one usually happens.

A worked example: the ramp that saves the floor

Suppose the migration plan implies 1.8 million dollars of year one consumption, but the floor case, counting only workloads with signed off migration dates, supports 1.1 million. Oracle proposes a flat 1.8 million commitment against the plan.

The floor plus a written overage rate outperforms the big commit in every slip scenario, and migrations slip more often than they accelerate. That asymmetry is the whole argument.

Where BYOL enters the model

BYOL lines change the rate and the risk at the same time. Every flagged workload needs its entitlement evidence attached to the model, because the flag will outlive the spreadsheet and surface in an audit.

Pay As You Go is the control case. Every commitment scenario is compared against simply paying list with no commitment, which is the honest baseline the account team will never show you.

Share the floor, negotiate at the floor, and keep the stretch case in your own file. Oracle sizes the deal to what you show it.

How do you run the hyperscaler benchmark and the reference checks?

Benchmark on your own workload numbers, not on published rate cards, and normalize compute units before comparing anything. The per workload comparison for Oracle estates is covered in OCI versus AWS for Oracle workloads, and the Benchmark Program supplies deal evidence rather than survey averages.

References are the second evidence artifact and the cheaper one. Insist on three production customers of comparable scale, and source at least one yourself through user groups or peers rather than taking only Oracle's curated list.

Normalizing the benchmark file

• Compute units. OCI bills x86 compute in OCPUs, each carrying two vCPUs; convert before comparing any hourly rate.
• Like for like shapes. Match processor generation and memory ratios, not just core counts.
• Storage assumptions. State the IOPS and throughput profile per workload; storage tiers do not map one to one across clouds.
• Traffic profile. Model each workload's egress and cross region pattern rather than assuming the marketing allowance covers it.

Eight questions to ask a reference customer

1. How accurate has billing been, and how were disputes resolved?
2. Have you claimed against an SLA, and what did the process require?
3. What did a severity one support case actually look like?
4. Where did the migration diverge from the plan Oracle presented?
5. How did the account team behave at your first renewal?
6. Which ancillary charges surprised you in the first year?
7. How long did the first production workload take from signature?
8. What would you negotiate differently with what you know now?

Question five is the one that matters. Day one behavior is marketing; renewal behavior is the product you are actually buying.

Which clauses must be in the OCI order document?

Ten, and the test for each is the same: if the order document is silent, what happens? The governing framework is Oracle's cloud services contract set, and silence almost always resolves in Oracle's favor. Which document binds, and how the pieces stack, is covered in the cloud contracts briefing for CIOs.

The clause Oracle will not volunteer

Unused credit treatment. The default is forfeiture, the account team knows it, and the topic will not come up until you raise it. Raise it in writing, in the first markup, while you still have a signature to trade.

One discipline covers all ten: nothing agreed on a call exists. No side letters, no emails as amendments; the order document is the deal.

Running the markup itself

• Send the clause list before the first draft arrives, so Oracle's paper is written knowing what will be checked.
• Track every clause in a simple grid: requested, countered, agreed, with the draft version where each landed.
• Trade clauses against signature timing, not against each other. The calendar is your concession; the clause list is not.

How does the evaluation scorecard keep the decision honest?

By existing before the first quote arrives, with weights the sponsor has signed. Scored after the fact, it is a rationalization of a decision already made on rate.

Decision rules that survive an audit committee

• Terms floor. If commercial terms score below an agreed floor, there is no signature, whatever the rate says.
• No pending evidence. A reference that cannot be produced scores zero, not "to follow".
• Failed requirements move, they do not fade. Each one goes to the clause list or the price, never to hope.
• Two scorers. Independent scoring, disagreements adjudicated in writing by the sponsor.

What happens to the toolkit after signature?

It becomes the operating baseline and the renewal file. The commitment model turns into the burn plan for the weekly review, the clause list feeds the contract calendar, and the scorecard is rerun before renewal with real consumption data.

• Commitment model hands over to the FinOps operating model as the pro rata plan.
• Clause list becomes the contract calendar: notice dates, ramp steps, and review points, owned by procurement.
• Scorecard and references become the renewal baseline inside the Renewal Program.

Where the common advice on buying OCI is wrong

The common advice says OCI is the aggressive challenger on price, so the real work is technical validation and the commercial side will look after itself; push for the deepest rate and sign. We disagree. The rate is the least durable part of an OCI deal, because the other nine clauses decide what it is worth. A deep rate on credits forfeited at period end, with no ramp, no renewal protection, and no exit assistance, costs more than a modest rate inside a well papered order. In the negotiations we supported, the money was won in the clause list and the commitment curve, and the buyers who chased the headline rate signed the most expensive deals in year two.

15 to 25

OCI negotiations supported

10

Clauses on the order document checklist

12

Weeks of procurement runway that hold

Source: Redress Compliance advisory engagement file, OCI procurement 2024 to 2025.

How Redress engages on an OCI purchase

Redress runs the buying event buyer side: the five artifacts, the benchmark evidence, and the order document markup, through the Benchmark Program and the Software Spend Assessment. Vendor Shield keeps the file warm between events, and benchmarking underpins the rate conversation.

The condensed versions are free: the OCI procurement white paper and the ten OCI negotiation levers. Redress is independent: read about us, the management team, our locations, or contact us.

What should a buyer do next?

1. Set the signature date twelve weeks out and put the five artifacts on that calendar with owners.
2. Write the scorecard first and get the sponsor's signature on the weights before any quote exists.
3. Build the requirements matrix with the region and service availability check done line by line.
4. Model the commitment yourself in three scenarios, and let Oracle see only the floor.
5. Book three reference calls, at least one sourced outside Oracle's list, and score the answers.
6. Mark up the order document against the ten clauses, starting with unused credit treatment.
7. Refuse the quarter end squeeze. If the clause list is not closed, the signature moves, not the terms.
8. Hand the file to operations on day one, so the renewal negotiation starts with two years of evidence.

Frequently asked questions

What is in an OCI procurement toolkit?

Five artifacts: a requirements matrix, a three scenario commitment model, structured reference checks, a ten clause order document checklist, and a weighted evaluation scorecard. Together they move the negotiation from Oracle's quote onto the buyer's evidence.

Do unused Oracle Universal Credits roll over?

No, not by default. Unused prepaid credits are forfeited at the end of the commitment period unless the order document says otherwise. Rollover or an extension window must be negotiated and written in, and it is one of the highest value clauses on the list.

How long should an OCI procurement take?

Twelve weeks from kickoff to signature for a material commitment. Scorecard by week two, requirements matrix by week four, commitment model by week six, references and clause markup by week eight, leaving four weeks to negotiate on your paper.

What discount should we expect on an OCI commitment?

There is no honest universal number. The achievable rate depends on workload mix, commitment size, term, timing, and the credibility of your alternative, so benchmark against real closed deals for your profile rather than a rule of thumb, and weigh terms as heavily as rate.

Is BYOL on OCI an audit risk?

Not when the entitlement evidence is attached to every flagged workload. The risk comes from BYOL flags set on assumptions: lapsed support, metrics that do not map, or licenses double counted between on premises and cloud. Evidence in the commitment model is the defense.

Should the OCI commitment end on the same date as an Oracle ULA?

Sometimes. Aligning the dates concentrates your leverage into one negotiation, but it also concentrates Oracle's, and a certification dispute can then hold the cloud renewal hostage. Align only if your ULA position is clean and documented.

Do we need reference checks if we already run Oracle on premises?

Yes. Running Oracle databases says nothing about OCI operations: billing accuracy, support response, SLA claims, and migration reality are cloud questions. Three production references answer them in an afternoon each.

Who should build the commitment model, finance or engineering?

Finance owns the artifact and engineering owns the inputs. Consumption estimates and migration dates come from the teams doing the work; the scenarios, the pricing, and the disclosure decision sit with the finance partner so the model survives contact with the negotiation.