Committee document
Decision memo: how should we move the warehouse?
A decision memo for the person who approves the migration method. It sets out the four ways to move the logic, the five facts that decide between them, and a format for the recommendation. Fill in the bracketed figures from your own estate.
The decision
What is being decided?
How the warehouse logic moves to [target platform] before [date], and who is accountable for proving the numbers match. The target is assumed to be decided; this memo decides the method.
The options
The four ways to move the logic
| Option | How it works | What you get per object | How it is priced | Fits best |
|---|---|---|---|---|
| SI-led or in-house rewrite | Engineers rewrite and test each object by hand | The tests the team writes | Hours against an estimate | Programs that are mostly re-architecture |
| Platform-native free tool | The target vendor’s converter translates the SQL it covers | Varies by tool; often none | Free or bundled | SQL-heavy estates moving to that one platform |
| AI code assistants | A language model drafts each object; engineers edit it | None by default | Seats plus tokens | A few hundred objects, or logic being redesigned |
| Product-led conversion | Fixed rules convert each object; each is reconciled against the original | A reconciliation report | Fixed price after the estate is read | Large, audited estates with a date |
What decides it
The five facts that decide between them
- Size: how many objects, after dead code is removed. Under about 300, a free tool or a rewrite is usually cheaper.
- Mix: how much is SSIS, scripted control flow and schedules rather than SQL. Free tools are strongest on SQL.
- Proof: whether audit or finance will need evidence per object before cutover.
- Date: whether a renewal, end of support or mandate sets the deadline.
- Price certainty: whether the budget must be fixed before work starts.
What to get first
What to have in hand before deciding
A count of the estate, banded by complexity, with dead code listed separately. Without it every option is priced on a guess. An assessment produces it in a few business days; a free tool’s analyzer gives a partial count at no cost.
The recommendation
Recommendation format
- We recommend [option] for [scope], because [the deciding facts above, with numbers].
- It costs [figure], fixed or estimated, with [credit or remedy].
- It is proven by [evidence per object], signed off by [role].
- The first step is [assessment, pilot or First Domain] on [domain], by [date].
- We rejected [options] because [one line each].
Risks
Risks to state in the memo
| Risk | What to ask for |
|---|---|
| The estimate moves after discovery | A price set after the whole estate is read |
| Converted code is not proven | A reconciliation report per object, on agreed inputs |
| The parallel run has no end date | Cutover by domain, with a threshold agreed in advance |
| The method only works with the vendor present | Native output with no vendor runtime, and ownership from day one |
Below the document
Where to read next
Print or save this page as the document. Every figure in it is one the reader can check on the linked pages.
FAQ
Frequently asked questions
How do we decide between an SI, a free tool and a conversion product?
Count the estate first. Then decide on five facts: size after dead code, the share that is SSIS and control flow, whether audit needs evidence per object, the date, and whether the price must be fixed before work starts. Large, audited estates with a date usually favor product-led conversion; small SQL-heavy ones a free tool.
Can we combine the options?
Yes, and many teams do. A free tool takes the SQL it converts well, a conversion product takes the rest with a report per object, and an SI runs the program around both.
Last reviewed:
Find out what your logic debt is costing you.
Every object counted, lineage, dead code and a fixed price in three business days. $3,000, credited.
Not sure which door? The scorecard takes ten minutes and tells you.