10decodersRetire SQL Server.
Land on Snowflake in 90 days.
A fixed-fee, twelve-week programme that moves your SQL Server data warehouse onto Snowflake — reconciled row-for-row against the legacy system, handed over to your team, and backed by a written parity guarantee.
The deadline is behind you, not ahead of you
Microsoft ended support for SQL Server 2016 on 14 July 2026. Not mainstream support — all support. No security updates, no bug fixes, no technical assistance. Your instances still run. They just stop being patched.
The “buy another year” option now carries a published price. Extended Security Updates run at roughly 75% of licence cost annually, cover only vulnerabilities rated Critical, are back-billed to the start of the term whenever you enrol, and stop for good in July 2029. For SQL Server 2016 specifically, the free-ESU-on-Azure route that earlier versions enjoyed is closed.
Most mid-market teams are planning around one date. There are three: the database, the operating system underneath it, and the next version already in the estate.
The finding you cannot argue with
ISO 27001, SOC 2, HIPAA and PCI DSS all require in-scope software to be supported and patched. An unsupported database engine is an audit finding, not a debate.
A finding converts a planned migration into an emergency one — and emergency migrations are the ones bought on time and materials, run to twelve months, and end with a reporting layer nobody signed off on.
You are not short of vendors. You are short of a committed end date.
Ask five integrators to quote a SQL Server to Snowflake migration and you get five variations of the same answer: a discovery phase, a range of “three to twelve months depending on complexity”, and a rate card. The range is honest. It is also unbuyable.
What actually blows up the estimate
- Stored procedures carrying two decades of undocumented business rules
- SSIS packages that were never in source control
- SSRS reports finance signs off every month — which do not port to Snowflake at all
- All three found in week seven, renegotiated in week eight, invoiced in week nine
What Cutover 90 does instead
- Count the estate first. A three-week Blueprint inventories every object and scores it
- Price from the count, not from a guess — the fixed fee is quoted off the inventory
- Make parity a contractual test, automated across 100% of in-scope objects
- Put the reporting layer in scope on day one rather than discovering it in week nine
What is true on day 91
Your SQL Server data warehouse is decommission-ready. Every in-scope table, view, stored procedure, SSIS package and report runs on Snowflake. A signed reconciliation report evidences row-for-row parity across the full inventory. Your team holds the credentials, the runbooks and the source code — and has operated the platform for thirty days with us on call.
Parity report
Object-for-object reconciliation evidence an auditor will accept.
24-month cost model
Snowflake consumption modelled against staying on SQL Server plus ESU.
Decommission plan
Shutdown runbook with the licence reclaim already itemised.
The 90 days, phase by phase
Estate Blueprint
- Automated inventory of every database, table, view, stored procedure, function, SSIS package, SQL Agent job, SSRS report, linked server and downstream consumer
- Complexity scoring per object — auto-translatable versus engineer-rewrite, with a conversion-rate forecast
- Target Snowflake architecture: account topology, warehouse sizing, RBAC model, network policy and dynamic masking design
- Snowflake credit forecast and a 24-month total cost model against staying on SQL Server plus ESU
- Wave-sequenced cutover plan with a parallel-run window and named rollback triggers
- A fixed-fee proposal for the full cutover, priced from the inventory
Foundation
- Snowflake account stood up as code across DEV, TEST and PROD — RBAC, object tagging, masking policies, resource monitors and budget alerts
- Ingestion path built and load-tested: change data capture from SQL Server plus historical backfill
- Tooling selected during the Blueprint against your constraints rather than our preference
- Git repository, dbt project scaffold and CI/CD pipeline working end to end with environment promotion
Conversion
- Schema and DDL conversion, including data type, collation and identity or sequence mapping
- T-SQL stored procedures and functions translated to Snowflake SQL and Snowpark — machine-translated first, engineer-reviewed always
- SSIS packages rebuilt as orchestrated ELT using dbt with Snowflake Tasks or Airflow
- SSRS reports rehomed like-for-like onto Power BI or Tableau against Snowflake
- Semantic and business-logic parity layer, documented rule by rule
Reconcile & cut over
- Automated reconciliation across 100% of in-scope objects: row counts, checksums, aggregate and column-level comparison, output as a signed parity report
- Performance benchmark on the top 50 production queries, before and after, with warehouse right-sizing applied
- Parallel run, dual-read validation window and a documented go/no-go gate
- Cutover runbook and rollback runbook, both rehearsed before go-live rather than written after
- FinOps guardrails live: resource monitors, auto-suspend policy and a credit-burn dashboard
Handover
- Three enablement workshops — platform operations, dbt and SQL patterns, cost governance — delivered live and recorded for your team
- Runbooks, architecture decision records and full source handover: infrastructure as code, dbt models, translation harness and reconciliation suite
- SQL Server decommission runbook with licence-reclaim inventory and archive and retention design
- Thirty days of hypercare after go-live, with named engineers and a response SLA
Explicitly out of scope: OLTP application migration (Snowflake does not replace transactional SQL Server workloads), net-new data products, BI redesign beyond like-for-like rehoming, master data management, and remediation of upstream source systems. Each is available as a priced add-on. We list them here because discovering them mid-programme is how fixed fees stop being fixed.
If any in-scope object fails parity, we keep working at our cost.
Reconciliation runs against 100% of the signed inventory — row counts, checksums, aggregate and column-level comparison. It either passes at the cutover gate, or the engagement is not finished. No hourly extension. No change order. No renegotiation.
Five ways this is not your risk
The Parity Guarantee
Reconciliation runs against every object in the signed in-scope inventory. If any of them fails at the cutover gate, we continue at our own cost until it passes.
Fixed fee, locked at Blueprint
The fee is quoted from the counted inventory and does not move for anything inside it. Complexity we misjudged is our problem. Objects you add later are priced from the same published rate card, at your option.
The Blueprint Credit
One hundred per cent of the Estate Blueprint fee is credited against the Cutover 90 fee if you proceed within sixty days. If you do not proceed, you keep the inventory, the architecture, the cost model and the cutover plan — and you are free to hand them to another integrator.
Parallel run and tested rollback
Your SQL Server estate stays live through the entire cutover window. You decide when it gets switched off — we do not need it dark to declare success. The rollback runbook is rehearsed before go-live, not written after.
You own every artefact
Terraform, dbt models, the translation harness, the reconciliation suite, runbooks and documentation all transfer to you. There is no proprietary runtime to license and no reason you need us in month four.
Size your estate, then model the return
Two calculators, no email gate. The first turns an object count into a complexity read and tells you which engagement shape you are. The second models what staying on SQL Server costs you against running the same workload on Snowflake. Both are indicative — the Blueprint is what replaces estimates with counts.
Your estate
Approximate counts are fine. Nothing is sent anywhere — this runs in your browser.
Cutover 90 fits this estate
This is the shape the twelve-week programme was built around — enough stored procedure and SSIS weight to need real engineering, contained enough to finish inside one quarter.
How this is calculated
- Machine-translation rates are weighted by object type. Schema and DDL convert at a high rate; T-SQL procedures at a moderate one; SSIS packages are rebuilt as orchestrated ELT rather than translated; SSRS reports are rehomed to a modern BI tool, not converted.
- The effort index weights each object type by the engineering hours it typically consumes, then adjusts for data volume and source version. Older and mixed estates carry a surcharge for collation, data-type and compatibility-level drift.
- The conversion window covers the build and conversion phases only. It excludes the three-week Blueprint and the two-week reconciliation and cutover window.
- Everything here is an estimate from counts you supplied. The Blueprint replaces it with an automated inventory of the actual estate.
Your return
Your figures, in whatever currency you work in. Data volume is taken from the estate calculator above. Nothing is sent anywhere.
How this is calculated and what it leaves out
- Staying is licence plus infrastructure plus platform headcount, held flat across three years, with Extended Security Updates added at Microsoft's published rate of roughly 75% of licence cost per year when you select enrolment.
- Moving is modelled Snowflake consumption — storage on a compressed volume plus compute credits derived from your data volume and workload profile — plus a reduced platform headcount, since patching, backup and capacity planning stop being your job.
- Migration investment is whatever figure we quote you from the Blueprint. Leave it at zero and the model shows the run-cost difference alone; enter it and the payback field tells you when you are square.
- Not included: the value of decommissioned hardware, reclaimed licences, avoided audit findings, or the analytics you cannot currently run. Those are usually the larger half of the case and none of them are modelled here.
- Snowflake rates move by region, cloud and edition. Treat this as a directional read, not a quote.
- The number most likely to be understated here is Snowflake consumption. Credit burn is driven by query patterns and concurrency, not just data volume, and a badly sized warehouse can multiply it. Modelling it properly needs your actual query profile, which is a Blueprint deliverable rather than something a form can infer.
Both of these are estimates built from numbers you typed in. The three-week Estate Blueprint replaces every one of them with a counted inventory, a modelled credit forecast and a fixed fee. If the numbers above are close to interesting, that is the conversation to have. Book the estate review →
Three ways in
Three engagement shapes. Which one fits is decided by your estate, not by a menu — and the Blueprint is what tells us. Every fee is fixed and quoted from a counted inventory, on a call, against your numbers.
Estate Blueprint
For teams who need a defensible number before they can ask for budget.
- Full object inventory and complexity scoring
- Target Snowflake architecture and RBAC design
- Credit forecast and 24-month TCO model
- Wave-sequenced cutover plan with rollback triggers
- Fixed-fee proposal for the full cutover
- Build and conversion
- Parity Guarantee
- Hypercare
Cutover 90
For one SQL Server warehouse that has to be gone this year.
- Everything in Estate Blueprint, credited in full
- Snowflake platform built as code, all environments
- Schema, T-SQL, SSIS and SSRS conversion
- Parity Guarantee across 100% of in-scope objects
- Performance benchmark and warehouse right-sizing
- Cutover and rollback runbooks, rehearsed
- Three enablement workshops and full source handover
- Thirty days of hypercare
Cutover Enterprise
For a multi-instance, regulated estate you want operated afterwards.
- Everything in Cutover 90
- Additional sources in scope — Oracle, DB2, Synapse
- Twelve months of managed run with an on-call SLA
- Continuous FinOps and quarterly optimisation
- Dedicated Forward Deployed Engineer, embedded
- Semantic model and Snowflake Cortex enablement
- Ninety days of hypercare
Why the middle tier is usually the answer. The Blueprint fee is credited back in full against Cutover 90, so starting with a Blueprint costs you nothing extra and removes the guesswork for both sides. Enterprise earns its premium only if you are carrying multiple source platforms or genuinely want the platform operated afterwards. One SQL Server warehouse and an in-house data team? The middle tier is the whole job. Bring your object counts to the review call and we will tell you which one you are.
The four paths, and what each commits you to
| Path | Commitment model | What you are buying |
|---|---|---|
| Tier-1 SI, time & materials | Time and materials with no ceiling, over six to twelve months | A deep bench and real Snowflake pedigree — plus an open-ended clock, US-only rates and a scope conversation every month. |
| In-house with free tooling | Two to four of your own engineers, off roadmap for six to nine months | Snowflake's SnowConvert is free and converts a large share of code. It does not produce reconciliation evidence, RBAC design, SSIS rebuild, SSRS replacement or cutover risk management — and it does not backfill the six months your data team is not doing their day job. |
| Stay on SQL Server 2016 with ESU | A recurring annual charge for Critical patches only, ending July 2029 | Critical-vulnerability patches only, back-billed to term start. Two to three years of licence value spent to arrive at exactly the same decision, later, under more pressure. |
| Cutover 90 | A fixed fee and a committed end date, both set before work starts | A committed end date, a contractual parity test and full artefact handover, with the fee fixed before work starts — delivered by a blended US and India team, which is why the number is structurally lower rather than discounted. |
Four packs folded into the Cutover 90 package
Work we build anyway on Enterprise engagements, included in the Cutover 90 scope rather than quoted as extras once the programme is already underway.
Snowflake FinOps Guardrail Pack
Resource monitors, budget alerts, a credit-burn dashboard and a 90-day consumption review after go-live.
Legacy Decommission Pack
SQL Server shutdown runbook, licence-reclaim inventory, and archive and retention design for audit continuity.
AI-Readiness Assessment
Semantic-layer and Snowflake Cortex readiness review on the migrated estate, with a prioritised use-case shortlist.
Auditor-Ready Parity Evidence Pack
Signed reconciliation report formatted for ISO 27001, SOC 2 or HIPAA evidence submission.
Built for the mid-market estate the tier-1s are not chasing
Engineers across four offices, on a blended US and India delivery model — the reason the fee is structurally lower rather than discounted.
Clients, with regulated-sector delivery experience across healthcare and financial services already in-house.
Certified, with security and quality processes an enterprise procurement team can evidence rather than take on trust.
- We fix the fee before work starts. Quoted from a counted inventory rather than a guess, and it does not move for anything inside the signed scope.
- We size for the mid-market. Tier-1 warehouse engagements start well above where a mid-market data budget ends.
- We lead with SQL Server. The big partners lead with Teradata, Hadoop and Oracle — the higher-value estates.
- We hand the platform back. Migration and operate is a choice you make, not a dependency you inherit.
Three situations where you should hire someone else
Estates above ~500 TB
Or Teradata, Netezza and Hadoop sources. Specialist Snowflake partners have been doing exactly that for years and have the accelerators and the bench for it. Pay their rate — it is the right rate for that problem.
You hold unused Snowflake PS credits
If professional services credits came bundled into your platform contract, spend those on the assessment before you spend cash with anyone, us included.
Fewer than ten tables
No stored procedures, no SSIS. SnowConvert is free and a managed connector will move the data. Two of your own engineers finish it in a fortnight. Do not hire an integrator for this.
We publish this section deliberately. A migration partner who cannot tell you when to walk away is not giving you an assessment — they are giving you a quote.
What buyers ask before they sign
Does Snowflake replace SQL Server entirely?
No, and any partner who says otherwise is selling you a rewrite you did not ask for. Snowflake replaces the analytical workload — the data warehouse, the reporting layer, the batch transformation. Transactional OLTP applications stay where they are or move to a supported SQL Server version or Azure SQL. Cutover 90 covers the analytical estate; the OLTP path is a separate conversation and we will tell you honestly which parts belong in which.
What happens to our SSRS reports?
They get rehomed, not repointed. Connecting SSRS to Snowflake is a known dead end and is not a viable long-term pattern. In-scope reports are rebuilt like-for-like on Power BI or Tableau against Snowflake, with output compared against the legacy report during reconciliation. This is planned in week one rather than discovered in week nine, which is the single most common reason these programmes overrun.
How can the fee be fixed when you have not seen our estate?
It is not fixed until we have. The three-week Estate Blueprint counts and scores every object, and the fixed fee is quoted from that count. If you engage the Blueprint and proceed within sixty days, the Blueprint fee is credited in full — so the sequence costs you nothing extra and removes the guesswork for both sides.
What does the Parity Guarantee actually commit you to?
Reconciliation runs across 100% of the objects in the signed in-scope inventory — row counts, checksums, aggregate comparison and column-level comparison. If any of those objects fails at the cutover gate, we continue working at our own cost until it passes. It applies to the signed inventory; objects added mid-programme are priced separately and then fall under the same guarantee.
Who owns the code and tooling afterwards?
You do. Terraform, dbt models, the translation harness, the reconciliation suite, runbooks and architecture decision records all transfer to you. There is no proprietary runtime to license and nothing that requires our continued involvement. If you want us to operate the platform afterwards it is because you chose to, not because you cannot leave.
What happens after go-live?
Thirty days of hypercare with named engineers and a response SLA, plus three recorded enablement workshops covering platform operations, dbt and SQL patterns, and cost governance. After that, ongoing managed run and FinOps is available as a separate agreement — it is not bundled into the migration fee, because bundling it is how migration partners become permanent.
How much will Snowflake itself cost us to run?
That depends on data volume, query concurrency, warehouse sizing and edition, and it is modelled explicitly during the Blueprint rather than estimated on a call. The deliverable is a credit forecast and a 24-month total cost comparison against staying on SQL Server with Extended Security Updates. FinOps guardrails — resource monitors, auto-suspend policy and budget alerts — are configured before go-live, not after the first surprising invoice.
A 45-minute estate review, at no cost
Bring your SQL Server version, an approximate object count, and the list of reports finance cannot live without. We will tell you on the call whether this is a three-week Blueprint, a full Cutover, or something you should do yourselves — and if it is the last one, we will tell you how.
- A first-pass read on complexity and likely conversion rate
- The three things in your estate most likely to overrun, named
- An indicative range and a realistic start date — no obligation
Request the estate review
We reply within one business day.