Cloud RFP: How to Compare Three Vendors Without the Deck

We sat in a review session where three cloud proposals lay open: one led with Kubernetes, one with “digital transformation”, one with the lowest monthly fee. The board asked which to choose. Nobody could answer — because each vendor used different units, different SLA definitions, and different assumptions about who patches the ERP database.

RFPs fail when they become beauty contests. They succeed when they force comparable answers. This guide gives mid-market firms a practical matrix — not a forty-page procurement template — to compare managed cloud and infrastructure offers fairly.

Fix the structure before you fix the price

Send every vendor the same outline and page limit:

1. Scope table (in / out / optional)
2. SLA and escalation (severity definitions, on-call coverage)
3. Backup/restore evidence (see our backup checklist article)
4. Admin and tenant model (who owns Azure/AWS tenant, break-glass)
5. Onboarding plan — first 90 days week by week
6. Pricing: monthly, project assumptions, egress/storage model, example at +30% capacity
7. Exit: data export, DNS, credential handover, notice period

Vendors who submit only marketing PDFs fail the first filter. Serious operators fill the table even if some cells say “not included”.

The scoring matrix (example weights)

Use a simple 1–5 score per dimension; committee members score independently, then compare.

Do not let sales weight “strategic partnership” unless it comes with named engineers and runbooks.

Questions that separate operators from resellers

Who is on call for our account at 03:00? Named pool vs anonymous “NOC”. What is excluded from unlimited support? Read the footnotes. How do you handle our internal change freeze in December? Shows operational maturity. What happens if we exceed egress budget? Surprises kill CFO trust.

Common traps in mid-market RFPs

Comparing monthly fee only while one vendor includes migration days and another bills every day. Letting the cheapest bid win without restore drill requirement. Accepting “ISO certified datacentre” as substitute for your application backup. Choosing multi-cloud slideware when you need one landing zone and one on-call number.

After scoring: one pilot, not big-bang

Before full contract signature, agree a paid or fixed-price pilot: migrate one non-critical workload, run one restore test, execute one incident simulation. Score again after 90 days. Contracts signed on faith alone revert to slide comparisons the first time ERP slows down.

How a scoring session actually works

We run vendor comparisons as a 90-minute workshop, not as a spreadsheet emailed around for two weeks. Four people in the room is enough: someone who owns money (CFO or controller), someone who owns operations (IT or plant), someone who will live with tickets (the person who already opens them), and one facilitator who is not selling a platform. Each person scores the seven dimensions independently on paper first. Then you put numbers on a whiteboard. The interesting part is never the average — it is the spread. If IT gives “restore proof” a 2 and sales gives it a 5, you have found a slide that nobody verified.

Typical mid-market mistake: the CEO joins for the last fifteen minutes and picks the vendor who “sounded most strategic.” Counter it by agreeing before the session that any score below 3 on restore proof or admin model is a veto, regardless of monthly fee. We have used that rule three times in the last year. Twice it eliminated the cheapest bid. Once it eliminated the most expensive “transformation partner” who could not name an on-call engineer.

A worked +30% cost example you can copy

Take a 120-person manufacturer with 28 VMs, one ERP, mail already in Microsoft 365, and a customer portal. Vendor A: PLN 18,000 / month “all inclusive”, 12 migration days included, egress “fair use”, restore tests “on request”. Vendor B: PLN 24,500 / month, 20 migration days included, egress priced, quarterly restore drill in the fee, named 24/7 pool. At +30% capacity (eight more VMs, 40% more storage, one extra environment for ERP upgrade):

• Vendor A: monthly jumps via change requests we have seen land at +PLN 6–9k, plus 8–12 extra project days in year one that were “obviously out of scope”.
• Vendor B: +PLN 4.2k in the published price book, restore drill still included, no surprise for the test environment if it was in the +30% example they already calculated.

The cheaper cover page is often more expensive by month nine. Ask both vendors to fill the same +30% sheet with your numbers, not theirs. If they refuse, they are not ready to operate you — they are ready to invoice you.

Clarification call: eight questions, forty minutes

After written proposals, do one joint call per vendor. Record it (with consent) and write answers into the scoring sheet the same day. Questions we actually use:

1. Who is on call for our account at 03:00 on a Sunday — named pool or a ticket queue in another country?
2. Show the last restore report for a client with an ERP like ours. Redact names; keep timestamps and duration.
3. What is excluded from “unlimited support” in the first year? Read the footnote aloud.
4. If Active Directory is compromised, who has break-glass, and where is that documented?
5. How do you handle our December change freeze — and who pays if a security patch cannot wait?
6. Walk through the first 90 days week by week. If week three is “TBD”, they do not have an onboarding machine.
7. What happens if we leave in month 18 — data export format, DNS, credential handover, last invoice items?
8. Which of our systems will you not touch, and who patches those?

Operators answer with names, dates and “not included”. Resellers answer with architecture diagrams and “we can discuss in the next workshop”.

What we decline to bid on

When QData is invited to an RFP, we sometimes walk away. A 12-page slide deck as the only submission format. A requirement for “multi-cloud from day one” with no workload that needs it. A demand for 99.99% on a single-site ERP with no budget for a second site. An invitation to undercut an incumbent by 40% with the same scope — that number is a lie waiting to become a change request. Saying no is part of knowing this market. The firms that stay and invent scope later are the ones CFOs remember unkindly.

Related reading: backup evidence belongs in every cloud RFP; managed cloud for 50–200 people covers the operating model this matrix assumes.

What to do next

Download your last three vendor emails into one folder — if they do not share a common outline, restart the RFP with this structure. Our Cloud Services team helps mid-market firms write neutral scope documents and score responses from an operations perspective — architecture and runbooks, not licence shelfware.

More from this topic

Backup and Restore: What to Demand Before You Sign

Backups on paper are not resilience. A practical checklist for CFOs and IT leads: RPO/RTO in writing, restore…

ERP Migration: When Cloud — and When to Stay On-Prem

“Move ERP to cloud” is not a strategy. Decision criteria for mid-market: downtime tolerance, integration with shopfloor, licensing,…

How to Choose Managed Cloud When You Have 50–200 Employees

You are past “a server in the office” but not ready for a 15-person platform team. A plain-language…