Skip to content
DT Nova

Data migration cost: how to estimate it and what drives it

Data migration cost depends less on how much data you have than on what state it is in. Ten million clean, consistent records can be cheaper to move than fifty thousand customer records spread over three systems, full of duplicates and free-text fields. This guide explains what drives the cost of a data migration, how to estimate it, the hidden costs that make migrations overrun, the pricing models, and a checklist that makes quotes comparable.

It does not quote our own prices. It explains how a migration budget is built, so any quote you receive, from us or from anyone else, makes sense. For how we run migrations, see data migration services; for moving whole systems to the cloud, see cloud migration cost.

The short version

  1. Assessment: list sources, profile the data and map old fields to new ones.
  2. Build: extraction, transformation and loading scripts, plus cleansing rules.
  3. Testing: trial runs, reconciliation of counts and totals, and checks by the people who use the data.
  4. Cut-over: the final run, a short freeze on the old system, and support afterwards.

What drives data migration cost

DriverWhy it adds workHow to keep it in check
Number of sourcesEach system has its own format, access method and quirksRetire minor sources by hand where it is quicker
Record typesCustomers, items, orders, invoices and documents each need their own mappingMove the types the new system needs on day one
Model differencesOld and new systems structure the same data differentlyMap early and agree rules for every gap
Data qualityDuplicates, gaps and inconsistent values must be fixed or decided onProfile a sample in the first weeks
HistoryEvery extra year of transactions adds mapping and checkingMove open items and recent history; archive the rest
AttachmentsDocuments and images need linking to the right recordsMove links and store files in a document system
DowntimeLittle or no downtime needs incremental loads and synchronisationPlan a short freeze if the business can accept it
Personal dataAccess control, logging and agreements with every processorMove only what is needed and lawful to keep

How to estimate a data migration

  1. Inventory. Every source, record type and rough count.
  2. Profile. Run checks on real data: duplicates, empty fields, invalid values, orphans.
  3. Map. Each target field gets a source and a rule, or a decision to leave it empty.
  4. Estimate per record type. Extraction, transformation, cleansing, loading and reconciliation.
  5. Add test runs. At least two full trial runs before the final one is common practice.
  6. Add cut-over and support. The final run, checks, and fixing what users find in the first weeks.

A three-point estimate (optimistic, most likely, pessimistic) per record type shows where the risk sits, and the spread is usually widest for the oldest and least documented source. Our guide to custom software development cost explains the method.

People's time

Migration is labour, and labour costs follow location. Eurostat's release of 31 March 2026 puts average hourly labour costs across the whole economy in 2025 at €34.9 in the EU and €38.2 in the euro area, from €12.0 in Bulgaria to €56.8 in Luxembourg (read 12 November 2026). Your own staff time is often the larger hidden part: deciding which duplicate is right, checking trial loads and signing off the result.

Hidden costs that make migrations overrun

  • Problems found late. Data checked only at the first trial run leaves no time to fix it properly.
  • Undocumented rules. Old systems hide logic in reports and habits, not in the data.
  • Access to the source. Export limits, missing interfaces or a vendor that charges for extracts.
  • Parallel running. Keeping two systems in step during a long cut-over.
  • Rework after go-live when users find records that did not arrive as expected.

Getting data out is becoming easier in the EU. Under the Data Act, contracts for data processing services must specify all categories of exportable data, and from 12 January 2027 providers may not charge switching fees for the switching process (Regulation (EU) 2023/2854, Articles 25 and 29, read 12 November 2026). See our guide to the EU Data Act.

Personal data and the cost of compliance

A migration is processing of personal data. A supplier doing it is a processor, and the work must be governed by a contract that meets GDPR Article 28; see our guide to data processing agreements. The GDPR also requires that personal data are kept no longer than necessary for their purpose (Article 5(1)(e)); both read 12 November 2026. A migration is a natural moment to apply retention rules, which also reduces the volume to move. See our guide to data retention policy.

Pricing models

ModelHow it worksWho carries the risk
Fixed-price assessmentInventory, profiling and mapping at an agreed priceSmall, known commitment; produces the estimate for the rest
Fixed price per record typeEach type priced after profilingShared; prices rest on evidence
Time and materialsHours billed as workedThe buyer carries overrun risk
Included in a system projectPart of a CRM, ERP or cloud project priceCheck how much data and cleansing is really included

How to compare data migration quotes

  • Which sources, record types and years of history are included?
  • Is data profiling done on real data before the price is fixed?
  • Who cleanses the data, and who decides on unclear records?
  • How many trial runs, and how is the result reconciled?
  • How much downtime is planned at cut-over?
  • Are scripts, mappings and reconciliation reports delivered to your repository?
  • Is there a data processing agreement, and where is data handled?
  • What is excluded, in writing?

Data migration is often part of a larger change; see CRM development cost and ERP implementation cost. All our services are on the services page, and every guide is in the guides library.

Questions people ask

How much does data migration cost?

It depends less on the amount of data than on its condition and complexity: how many source systems, how many kinds of records, how well the old and new data models match, how much cleansing is needed, how much history moves and how much downtime is allowed. Any figure given before a sample of the data has been examined is a guess.

How do you estimate data migration cost?

List the sources and record types, profile a sample of the data to find duplicates, gaps and inconsistencies, map each field from old to new, then estimate the work per record type for extraction, transformation, cleansing, loading and checking. Add test runs and the cut-over. A short paid assessment with real data gives a far better estimate than a guess from a description.

Why does data migration take longer than expected?

Because data problems are found late. Duplicates, missing values, free-text fields used for several purposes and undocumented rules in the old system each create work. Profiling the data in the first weeks finds them while there is still time to decide what to do.

Who should clean the data, us or the supplier?

Usually both. The supplier can find problems, fix those that follow clear rules and report the rest. Decisions about which customer record is correct or which items are obsolete need people from the business. Agree who decides what before the work starts.

Is it cheaper to migrate less data?

Usually yes. Moving open items and the history people actually use, and archiving the rest in a readable format, reduces mapping, cleansing and testing. It also helps meet the GDPR principle that personal data should be kept no longer than necessary.

Sources

Your data migration scope checklist

  • Every source system, its owner and how data can be exported.
  • Record types and rough counts per type.
  • Years of history needed in the new system.
  • Known data problems: duplicates, gaps, free-text fields.
  • Attachments and documents to move.
  • Downtime the business can accept at cut-over.
  • Personal data, and the retention rules that apply.
  • Who from the business checks and signs off the result.
  • What happens to the old system afterwards: archive or switch off.
  • Ownership: scripts, mappings and reports in your company's repository.