Migration

What a Migration Assessment Report Should Include

Rakesh Mamidala·Founder & Lead Engineer··9 min read

The Report Is the Decision

A migration assessment report is the artifact that turns "we should move off Oracle" into a plan you can staff, schedule, and defend in a budget meeting. Done well, it's the most valuable document in the whole project — it sets the timeline, the team size, and the risk posture before anyone writes a line of conversion code.

Done badly, it's a glossy PDF full of adjectives — "complex," "significant effort," "some manual work required" — that commits you to nothing and protects no one. The difference is specificity. A good assessment report is numeric, reproducible, and honest about the hard 5%. Here are the ten sections it should contain.

The Ten Sections

#SectionWhat it must answer
1Executive summaryOne page: the verdict, the headline numbers (objects, data size, % auto-convertible), the risk level, and the rough timeline. A CFO/CIO should grasp the scope in 60 seconds.
2Object inventoryCounts per schema and object type — tables, views, packages, triggers, sequences, types, jobs. The table-to-package ratio tells you if this is a data move or a code project.
3Data profileTotal size, largest tables, LOB volume, partitioned tables. Drives bulk-load runtime and whether CDC is needed to keep the cutover window short.
4PL/SQL surface + automation rateTotal lines of PL/SQL and the measured percentage that transpiles automatically (not a guess). The single best predictor of conversion effort.
5Hard-construct catalog (the 5%)Every construct that needs a human decision — packages-with-state, autonomous transactions, CONNECT BY, XMLTYPE, AQ, DB links — counted and located. This is the section that earns trust.
6Dependency & load-order mapCircular foreign keys and cross-schema references, so the load order is deterministic and constraints can be deferred safely.
7Risk scorecardEach axis (data volume, code volume, hard-construct density, dependencies, platform features) scored low/medium/high, rolled into one readiness signal.
8Effort & timeline estimateA phase-by-phase estimate derived from the numbers above — not from database size alone — with an explicit contingency factor.
9Recommended approachThe concrete plan: logical vs CDC cutover, phasing, which tables go first, validation strategy. An assessment without a recommendation is just data.
10Evidence appendixThe raw queries and outputs behind every number, so the report is reproducible and auditable rather than “trust me”.

Every Claim Needs a Number

The test for any line in an assessment report: could two engineers read it and arrive at the same plan? "Significant PL/SQL" fails. "62,000 lines of PL/SQL across 40 packages, ~95% auto-convertible, 18 constructs flagged for review" passes — it tells you exactly how many engineer-days of review to budget. Replace every adjective with a measurement:

  • "Large database" → 35 TB; largest table 18.1 GB; 4 partitioned tables
  • "Some complex code" → 18 hard constructs (3 autonomous txns, 5 CONNECT BY, 2 packages-with-state…)
  • "Manual effort required" → ~3,000 lines for human review ≈ 2 engineer-weeks
  • "Low risk" → risk score 6/15; one orange item (no CDC plan for the 18 GB fact table)

Red Flags of a Useless Report

  • No hard-construct list. If it doesn't name the 5% that needs human work, it hasn't actually looked.
  • A single automation percentage with no basis. "90% automated" with no per-object breakdown is marketing, not assessment.
  • An estimate anchored on data size. Timeline driven by gigabytes — ignoring code and application-layer SQL — is the classic under-estimate.
  • No recommendation. Data without a plan pushes the hard decisions back onto you.
  • Not reproducible. If you can't see the queries behind the numbers, you can't trust them.

Generating It Automatically

Most of this report is mechanical — it's reading the data dictionary and the PL/SQL source and counting. DBMigrateAIPro's assessment produces all ten sections in about a minute: the object inventory, data profile, PL/SQL surface with a measured automation rate, the flagged hard-construct catalog, the dependency map, and a Migration Advisor risk score — with the raw queries saved as the evidence appendix. You start from a complete, reproducible report and spend your time on the one section a tool can't fully write for you: the judgement calls in the recommendation.

The Honest Verdict

A migration assessment report isn't a formality you produce to start the project — it isthe project's first and most important deliverable. Insist on the ten sections, demand a number behind every claim, and make sure it names the hard 5%. A report that does those three things turns a risky migration into a planned one.

Get a real assessment report in a minute

DBMigrateAIPro's assessment is read-only, runs in about a minute, and produces all ten sections with the evidence behind them. Free for Year 1 — no signup, no license key.