We read other firms' migration quotes for a living, usually because a buyer wants a second opinion before signing. The patterns repeat so reliably that the review has become a checklist, and this article is that checklist made public. The premise is simple: a quote is a map of what the seller believes the work is. Missing lines mean missing beliefs, and missing beliefs become change requests. As part of our series on what an OCI migration really costs, this is the buying side companion: what the complete cost picture looks like when it lands on your desk as a proposal.
The lines an honest quote must have
A named assessment phase with a deliverable. If the quote was produced without an assessment, the price is a guess wearing a suit. An honest quote either follows a real discovery exercise, the kind whose contents we cataloged in the 40 point assessment checklist, or its first line is one, with the explicit right to reprice after it. Fixed prices issued before anyone has inventoried the estate are not confidence, they are padding, and you are the one paying for the pad.
Per workload method and window. Each system, or at least each wave, should show its migration method and its expected outage window. This single discipline exposes more dishonesty than any other: a quote that lists GoldenGate replication for every database either reflects an estate of genuine tier one systems or a seller pricing capability nobody asked the outage arithmetic to justify. The tooling economics are covered in migration tooling costs; the honest version of this line defaults to the free tier and escalates only where the downtime price demands it.
Rehearsals, named and counted. At least one full cutover rehearsal per wave for serious systems, with its environment costs and its people hours visible. A quote with no rehearsal line has either hidden the cost inside a day rate or, worse, not planned to rehearse.
The overlap budget. Dual running is a certainty, not a contingency, and we gave it its own article in budgeting dual running. The quote should show the parallel running window per wave, who pays for the OCI consumption during sync and rehearsal, the fallback replication tail, and the decommission tasks at the end. Absence of a decommission line predicts, with depressing accuracy, a source estate still billing six months after go live.
Validation and acceptance criteria. Row counts, performance baselines, application smoke tests, and who signs what before the old system is switched off. This is where quality lives, and it is chronically the first casualty of a price negotiation precisely because it is invisible until it is missing.
Exclusions, in plain language. Every quote excludes things. Honest ones say so on the page: application code changes, third party vendor fees, network circuits and their lead times, licensing advice, source side egress charges. The dishonest version is silence, because silence becomes a change request priced at distress rates mid project.
Reading the commercial shape
How the price is constructed matters as much as the lines. Three shapes dominate, and each moves risk differently.
| Commercial shape | Who carries delivery risk | Where it is honest | Where it hides trouble |
|---|---|---|---|
| Fixed project fee | The seller | After a real assessment, with scope and acceptance criteria written down | Fixed prices issued before discovery, padded for the unknown you pay for either way |
| Time and materials | The buyer | Genuinely uncertain scope, exploratory phases, estates nobody can inventory yet | Open ended day rates with no estimate corridor and no incentive to finish |
| Milestone based | Shared | Long multi wave programs, payment tied to accepted cutover per wave | Milestones defined by activity rather than outcome, paid for effort, not arrival |
The pattern we recommend to buyers is assessment first as a small fixed engagement, then a fixed fee for the migration proper, repriced against the assessment findings, with milestones per accepted wave. That sequence keeps every party honest: the seller prices against evidence rather than fear, and the buyer pays for outcomes rather than hours. It also matches how we structure our own work: a fixed project fee for the bounded build, a managed monthly retainer where the engagement continues into operations, and an optimization fee on verified savings where the work is reclaiming money rather than moving systems.
The padding signals
Padding hides in the opposite direction from omission: lines that are present but should not be, or sized beyond what the estate justifies. Replication licensed for everything is the classic. Others worth interrogating: project management priced as a large flat percentage on top of every other line, contingency stacked at both line level and total level so the same risk is paid for twice, premium tooling line items for capabilities OCI includes free, and post migration support periods sized in months when the steady state plan already includes a managed service. None of these are dishonest in isolation. The test is whether the seller can explain each one against your estate rather than against their template, and the sellers who can explain are usually the ones who did the assessment properly.
Two quotes, one estate: a worked comparison
The method earns its keep when quotes disagree wildly, so walk a composite. An estate of twenty databases and eighty servers receives two proposals. Vendor A quotes 100 units, fixed, with a one page scope. Vendor B quotes 240 units after a paid assessment, with eleven pages of line items. Procurement's instinct is that B is gold plating. The line item comparison says otherwise.
Vendor A's page contains no assessment, no rehearsals, no overlap budget, no decommission tasks, and a single sentence about validation. Priced at the rates implied by A's own day rate card, the missing lines come to roughly 110 units, work the estate will need whether or not it appears on a quote, delivered later as change requests at distress prices or absorbed as internal effort nobody budgeted. Vendor A's true comparable number is therefore around 210 units, with the change request friction still to come. Vendor B's 240 includes two findings worth challenging in the other direction: replication licensed for eight databases when the outage arithmetic supports three, and a six month post migration support block the steady state operating plan duplicates. Both lines survive scrutiny as padding, worth perhaps 25 units. Honest comparable totals: roughly 210 against 215, from headline numbers that differed by a factor of 2.4. The decision then rests where it should, on delivery evidence and risk transfer, not on a cover page artifact.
The exercise takes an afternoon with a spreadsheet and pays for itself more reliably than any negotiation tactic. It also produces a useful side effect: the normalized line item sheet becomes the master scope for whichever vendor wins, closing the gaps before they become disputes.
The questions that expose a weak quote
Where time is short, five questions in a vendor meeting do most of the comparison work. Which assessment produced these numbers, and may we see its inventory? What is the migration method and outage window for our three most critical systems, specifically? Where in the quote is the rehearsal for each of them? Who pays for the parallel running period, and until what date? And what happens to the price when discovery contradicts an assumption, walked through on a real example? Confident, specific answers correlate strongly with programs that finish near their number. Deflection to the appendix correlates with the other kind. The deeper diligence behind these questions, reference checks, delivery evidence, and the engagement structures that keep incentives aligned, is the standing work of our OCI consulting practice, which reviews migration proposals as an independent second opinion precisely because the afternoon it takes is cheaper than the quarter it saves.
A framework for comparing quotes that disagree
- Normalize the scope before the price. List every line across all quotes in one sheet, mark present, absent, or excluded per vendor. Most three to one price gaps disappear into scope gaps.
- Price the absent lines yourself. Missing rehearsals, missing overlap, missing decommission: estimate them and add them to the cheap quote. Compare totals, not cover pages.
- Demand per workload method and window. Any quote that cannot show its method per wave has not done the thinking, whatever its price.
- Check the tooling tier against the outage arithmetic. The downtime price per system, from pricing downtime, should justify every paid replication line. Reject capability priced without a window to justify it.
- Read the exclusions as a second quote. Everything excluded is a cost you carry. Sum it. The honest vendor's exclusion list is usually longer, and their total is usually truer.
- Test the change request mechanics. Day rate for changes, approval flow, and what happens when discovery inside the project contradicts the assessment. This paragraph is where bad projects become expensive ones.
- Weigh evidence of repetition. Migration economics reward practice. A vendor on their 500th engagement prices the learning curve in your favor; a vendor on their fifth prices their education into your fee.
Reading the assumptions register
Between the line items and the exclusions sits the quietest section of any proposal: the assumptions. It rarely gets read and it decides more disputes than the rest of the document combined, because every assumption is a conditional discount. The source environment is documented and accurate. Network bandwidth between sites meets a stated number. Application owners are available within five business days of a request. Test data can be copied without masking. Freeze periods will not exceed the dates listed. Each one of these is a price reduction taken in advance against a condition the buyer is now responsible for delivering, and the change request that follows a broken assumption is priced at the vendor's leisure.
The buyer's defense is to cost the assumptions like risks: walk the register asking which of these are actually true today, which are probably true, and which are aspirations wearing declarative grammar. The documentation assumption fails at most estates; the availability assumption fails at all of them during quarter close; the masking assumption fails anywhere a regulator has visited. Every assumption unlikely to hold should either be corrected before signature, with the price adjusted honestly, or converted into a named contingency with a number on it. A vendor who resists making assumptions explicit is telling you where their margin protection lives, and a buyer who signs an unread assumptions register has agreed to a second negotiation at a worse moment. Ten minutes per page of assumptions is the cheapest diligence in the entire procurement.
What independence buys in this conversation
One structural note that buyers underweight: who profits from the answer. A quote from a reseller earns margin on what you consume; a quote from a vendor's delivery arm earns credit toward the vendor's targets. Neither is automatically wrong, but both have a thumb on the scale when the honest answer might be a smaller estate, a cheaper tier, or no move at all for some workloads. Independent advisors, paid only by the buyer, have no consumption to protect, which is the position we built our practice around, 500+ engagements, 20+ years of combined Oracle experience, and 24/7/365 operations behind the quotes we do write. The same logic applies to the licensing dimension of any quote: entitlement questions belong with an independent licensing firm, not with whoever profits from selling replacement licenses.
An honest quote will rarely be the cheapest page in the pile. It will be the one whose total survives contact with the project, and over the life of a migration that is the only cheapness that counts. Read for structure first, price second, and make the missing lines speak before you sign.
Part of a series
This guide is part of OCI Migration — our complete pillar guide on the topic.
Moving Oracle workloads to OCI, or already running on OCI and not sure the architecture or the spend is right? Most teams bring in a specialist before they commit to a region, a shape, or a Universal Credits number. OCISpecialists.com plans the landing zone, runs the migration, and manages the estate after go live, on a fixed project fee, a managed monthly retainer, or a cost optimization fee paid only on verified savings.