StackLift
All playbooks

Playbook

Legacy Data Audit

An audit of the data you already have, run before you migrate or build on it. It lists sources, profiles quality, finds the gaps and ends with a move, fix or archive decision per dataset.

Best reader

CTOs, data owners and operations leads planning a migration, modernization or AI project

Outcome

A data inventory with quality findings, owners and a decision for each dataset.

Use this sequence

1

Inventory every data source, including spreadsheets and exports.

2

Name an owner and a purpose for each dataset.

3

Profile each source for gaps, duplicates and broken references.

4

Map fields to the target model and flag what has no home.

5

Decide per dataset: migrate, clean first, archive or delete.

6

Write down the validation checks you will run after the move.

Find all the data first

Legacy estates hide data in side tables, spreadsheets, reports and exports. The audit starts by finding them.

Databases and file shares

Spreadsheets used as systems

Reports and exports people rely on

Profile before you promise

Profiling shows what the data is really like. Do it before you commit to a migration scope.

Missing required values

Duplicates inside and across systems

Orphaned records

Inconsistent formats

Log every fix

Cleaning data without a record makes it impossible to audit or repeat. Keep a change log.

Rule, reason and date for each change

Counts before and after

A way to rerun on fresh data

Start in 60 seconds

Turn this checklist into a scoped plan.

Answer a few questions and get an AI-assisted, architect-reviewed scope, cost range, and timeline for your software project.

Frequently asked questions

Why audit before migrating?

Because migration moves problems along with data. An audit shows what is worth moving, what needs cleaning and what to leave behind, so scope and effort are realistic.

Does AI work need a data audit too?

Yes. Models and retrieval systems reflect the quality, freshness and permissions of the data they read. An audit reveals gaps before they become wrong answers.

The complete guide

Legacy Data Audit: the complete guide

Updated November 18, 2026 · Published by , Founder, StackLift AI. Written with AI assistance; how we make content.

Short answer

A legacy data audit is a structured review of the data in your old systems before you migrate it or build on it. You inventory sources, name owners, profile quality (gaps, duplicates, broken references, formats), map fields to the new model, and decide per dataset whether to migrate, clean first, archive or delete.

What is a legacy data audit?

A legacy data audit looks at the data you already have: where it lives, who owns it, what condition it is in and whether anyone still needs it. It happens before a migration, a modernization or an AI project, because each of those moves or reads your data and inherits its problems.

The audit is different from the migration itself. It produces facts and decisions, not a new system. Teams that skip it tend to discover the problems mid-project, when changing scope is expensive. It is a natural first step of the legacy modernization roadmap.

What are the steps of a legacy data audit?

  1. Find and list every source. Output: an inventory of databases, files, spreadsheets, reports and exports. Involve IT and each department.
  2. Assign owners and purposes. Output: a named owner and a reason for each dataset. Involve business leads.
  3. Profile each source. Output: a quality report covering gaps, duplicates, broken references and format problems. Involve a data engineer.
  4. Map data to the target. Output: a field-by-field map and a list of data with no home. Involve engineering and the product owner.
  5. Decide per dataset. Output: migrate, clean first, archive or delete, with a reason. Involve the owners and legal for retention.
  6. Define validation. Output: the counts, totals and spot checks you will run after the move. Involve QA and the data owners.

The sections below give you checks for each step. If your goal is an AI feature rather than a migration, also read the enterprise AI readiness checklist.

How do you inventory legacy data sources?

Ask people, not just systems. Business data often lives in places IT does not track: a spreadsheet that runs a department, a report someone exports every Monday, an old database behind a forgotten tool. Walk through each team's week and note every place data is read or written.

Data inventory: one row per dataset
FieldWhat to recordWhy
Name and locationSystem, table or file pathSo it can be found again
Owner and purposeWho is responsible and what it is used forSomeone must decide its fate
Format and sizeType of store and rough volumeShapes the migration method
SensitivityPersonal, financial, health or confidential contentDrives security and retention rules
Update patternHow often it changes and by whomAffects cutover planning
ConsumersReports, systems and people that read itReveals hidden dependencies

Dependencies matter most. A table nobody owns may still feed a monthly report. Record consumers so nothing breaks silently.

How do you profile data quality?

Profiling means measuring what the data is actually like, usually with queries or a profiling tool run against each source. It replaces opinions with evidence. Do it before you agree a migration scope, because the findings change the plan.

Quality checks
CheckWhat it findsTypical fix
CompletenessRequired fields that are empty or half filledBackfill, default or exclude
DuplicatesThe same customer or item more than once, inside or across systemsMatch rules and a merge decision
Referential integrityRecords that point to something that no longer existsRepair or archive the orphan
ValidityValues outside allowed ranges or listsCorrect or flag for review
ConsistencyDates, units and codes written in different waysStandardize in the transform step
FreshnessData nobody has updated in a long timeConfirm relevance or archive
AccuracyValues that disagree with a trusted sourceReconcile with the source of truth

Keep the results in a report anyone can read. A short list of the worst problems, each with an owner and a rough cleaning effort, is more useful than a long dump.

How do you map legacy data to the new system?

Mapping connects each old field to a new one, and shows what does not fit. Write the rules down: which field becomes which, how values are translated, and what happens to fields the new model does not have. Review the map with the people who use the data, because they know what the fields really mean.

  • List fields with no destination and decide: add to the new model, store as an attachment, or drop.
  • Document code tables and value translations, such as old status codes to new ones.
  • Decide how to treat history: full history, a summary or only the current state.
  • Plan keys and identifiers so references survive the move.
  • Record how personal data is handled, including what must be deleted or minimized.
  • Decide on a freeze period or a way to catch changes made during the migration.

Mapping also exposes integration work. The system integration glossary entry and the legacy modernization cost range explain what drives it.

How do you decide what to migrate, clean, archive or delete?

Not everything deserves a place in the new system. Moving all of it makes the migration slower and carries clutter forward. Decide dataset by dataset, with the owner, and record the reason.

Decision guide
DecisionUse whenWatch out for
Migrate as isQuality is good and the data is in useCheck the mapping still holds
Clean, then migrateNeeded, but problems found in profilingLog every change so it can be repeated
ArchiveRarely used but required for records or lawKeep it readable and access controlled
DeleteNo purpose and no retention dutyConfirm with legal and the owner first
Rebuild from sourceDerived data that can be recomputedMake sure the source is still available

Retention rules depend on your industry and location, so check them with counsel instead of guessing. Never delete what you may be required to keep.

How do you clean and validate safely?

Cleaning should be scripted, logged and repeatable. A change log that records the rule, the reason and the counts before and after lets you audit what happened and rerun the process when fresh data arrives. Hand edits in a spreadsheet cannot be repeated or checked.

  • Run validation before, during and after the migration, and keep the results.
  • Compare row counts and totals between source and target for each dataset.
  • Use checksums or sample comparisons to confirm integrity.
  • Have business users check a sample of real records in the new system.
  • Keep the old system readable until sign-off and agree when it is retired.
  • Continue quality checks after go-live, with named owners.

Test the whole move on a copy before the real run, and plan a rollback. The production readiness checklist covers cutover and recovery. To scope the work, use the free estimate or contact us.

Questions people ask

Why audit data before a migration?

Because a migration moves problems along with the data. The audit shows what is worth moving, what needs cleaning and what to leave behind, so scope and effort are realistic before you commit.

How is a data audit different from data profiling?

Profiling measures data quality. An audit is broader: it also inventories sources, assigns owners, maps data to the target and decides what to do with each dataset. Profiling is one step inside it.

Does AI need a data audit too?

Yes. Models and retrieval systems reflect the quality, freshness and permissions of the data they read. Gaps and duplicates in the source show up later as wrong or inconsistent answers.

Who should own the audit?

A named lead with authority across teams, supported by a data engineer and by the business owner of each dataset. Business owners decide what the data means and whether it is still needed.

What if nobody knows who owns a dataset?

Find its consumers and ask them. If no one uses it and no retention rule applies, flag it for archive or deletion with a sign-off from a leader, rather than letting it migrate by default.

Can we clean data after the migration?

You can, but it is harder. Cleaning before the move keeps the new system trustworthy from day one. If time is short, clean the critical datasets first and schedule the rest.