New: Odoo and Dynamics 365 cost calculators.Estimate your costs

ERP migration that matches the freeze

ERP migration: move masters, open items, and history so the freeze trial balance, aging, and inventory match. Checks, steps, and October 2026 costs.

Both platformsImplementation & Go-Live11 min readUpdated 16 sources

Key takeaways

  • Prefer a month-end or quarter-end freeze so you reconcile a full period.
  • Write down whether the on-hand file includes units already on open receipts.
  • Accuracy: the address, cost, and tax identifier match the real customer, vendor, or item.
  • Book the freeze only after a full-volume rehearsal has a measured duration and a signed tie-out.

What ERP migration and ERP data conversion mean

ERP data migration is the controlled transfer of master data, open transactions, and selected history into Dynamics 365 or Odoo so the day-one trial balance, AR and AP aging, and inventory match the legacy freeze. Searchers use ERP migration, ERP data migration, and ERP data conversion for that same job. Conversion is the reformatting inside the job: chart of accounts rollups, units of measure, tax codes, and posting groups. Migration is the program around it: scope, cleanse, map, trial load, validate, cutover, and hypercare.

A file copy is a different task. IBM describes database schema conversion as its own step: when the target database's data-definition language is incompatible, the schema has to be converted before the data moves, and integrity is checked after the transfer. That guidance is about storage and databases. A controller still proves open invoices, on-hand quantity, and the general ledger against the freeze.

General results for this search often explain what an ERP is, or how to move a database. The decision here is which records go live, on which freeze date, until the trial balance, aging, and inventory tie. Business Central's version 14 reimplementation tool moves master data, opening balances, and setup, while a Dynamics GP cloud migration can also carry transactional history and vendor 1099 data. Odoo 19 loads the files you prepare into contacts, products, journal entries, and orders, so historical invoices stay open documents only when those open items are in the file.

Panorama Consulting wrote in June 2019 that most ERP vendors do not take responsibility for data migration or cleansing because the work is costly, and that 41% of respondents in its 2019 ERP Report said data issues caused the project to run over schedule. Treat that 41% as a 2019 survey result, a warning that dirty masters slip dates, and a dated figure. It is a planning input from that report, and it is separate from any 2026 benchmark.

The SME sequence from audit to hypercare

Run the work in this order: name data owners, audit and cleanse, map (including tax and posting groups), trial-load in a sandbox, reconcile, fix, reload deltas, freeze, cut over, reconcile again, then hypercare. Loop the middle steps. A load that reports success is the start of reconciliation, and the freeze balances still have to be checked.

Microsoft's Dynamics GP to Business Central cloud migration is the concrete product sequence to copy in spirit. It runs assessment, preparation (including getting data clean), cloud migration setup with no data moved yet, company configuration, data replication (repeatable), data upgrade, validation, then completion. The on-premises system stays the system of record until migration is completed. After a successful data upgrade you cannot replicate again from an earlier version without risking mixed upgraded and non-upgraded records.

Microsoft does not state how many hours the first Business Central replication takes. Duration depends on volume, SQL configuration, and connection speed. The first replication moves all selected data. Later runs move changes only, so they are faster. That is why you rehearse replication before the one-time data upgrade, and why you measure a full-volume pass before you book the freeze weekend.

  1. 1. Name owners and audit

    Assign a business owner for customers, vendors, items, the chart of accounts, open AR, open AP, and inventory. Profile duplicates, blank required fields, and posting groups before anyone builds a map. Panorama's buyer question still applies: is our data clean? If the answer is no, the schedule risk sits here, which is the pattern in that 2019 report where 41% of respondents tied data issues to an overrun.

  2. 2. Cleanse in the legacy system

    Merge duplicate customers and vendors to one surviving record and repoint open documents to that record. Fill tax identifiers, payment terms, units of measure, and posting groups. Stop using the new ERP as the cleaning bench. Odoo imports are permanent and cannot be undone. Business Central replication overwrites data already in the online target.

  3. 3. Map fields, tax, and posting

    Write a field map with a business sign-off per entity. Include tax codes, posting groups, and currency. A US company mapping GP vendor 1099 boxes is doing a different job from a Canadian company mapping GST, HST, or provincial tax codes into the target posting setup. Lock the map before the first trial load so later deltas use the same rules.

  4. 4. Trial-load in a sandbox

    Load reference data, then masters, then open documents, then opening balances. For GP, repeat replication into a non-production company. For Odoo, upload, map columns, run Test, and only then Import. For a non-SQL source into Business Central, Microsoft's path is configuration packages, then the cloud migration tool if the target is online. Keep the trial tenant off the company people already use for business.

  5. 5. Reconcile to the freeze image

    Compare debits to credits, the trial balance to the legacy freeze, AR and AP aging to the control accounts, and inventory quantity times cost to the inventory account. Those aging and inventory-value checks are controller tie-outs to the freeze. On a GP migration, Microsoft's validation page is a separate check: it compares G/L accounts, statistical accounts, bank accounts, customers, items, purchase orders, vendors, and vendor 1099, with failed-record counts and a drill-down of the expected GP value versus the Business Central value. It does not show aging or an inventory-value tie-out. Record every break with an owner.

  6. 6. Fix and reload deltas

    Correct the source or the map, then load the change. In Business Central, later replications move changes only, until you run data upgrade. In Odoo, an External ID lets the same file update a record. After a GP data upgrade, do not start another replication from an earlier version. Split large Odoo files into smaller batches so a timeout does not leave a partial, permanent load.

  7. 7. Freeze, cut over, and hypercare

    Pick a month-end or quarter-end when you can. Finish bank reconciliations, payment runs, and journals, then export the freeze trial balance. Load the final set, reconcile again, and open the new system only after finance signs the controls. Keep the legacy system available for questions. On Business Central, on-premises stays live until migration is completed, and the online tenant is read-only for migrated data while cloud migration is connected.

What to load, and what to leave in the legacy system

Use two menus. The minimum go-live set is the chart of accounts, customers, vendors, items, open AR, open AP, inventory on-hand, open orders, and general-ledger opening balances. The optional set is closed history. Keep closed history in the legacy system or a read-only archive when the new system cannot recreate it as open documents. Panorama's 2019 advice still matches the product limits: do not plan to move every legacy record and expect the new software to standardize it.

Business Central's version 14 reimplementation tool follows the minimum menu. Microsoft describes it as a move of essential master data, opening balances, and setup, and it does not require converting customizations to extensions because only that essential data moves. A GP cloud migration can go further: system and company setup, master data for accounts, customers, vendors, and items, transactional data, classes, vendor 1099, and historical data. Which company data moves is configurable.

Odoo 19 imports into business objects from Excel (.xlsx) or CSV, including contacts, products, bank statements, journal entries, and orders. That is a file load into those objects. It is a weak stand-in for a subledger rebuild. Plan opening balances as a balanced journal, and plan open AR and AP as the documents you still need to collect or pay. If a closed invoice only needs to be looked up, leave it where auditors can already read it.

Worked scope, with the assumptions stated. Assume one US company, a 30 September freeze, and a target of Business Central online from GP. Minimum load: chart, customers, vendors, items, open invoices and bills, open orders, on-hand inventory, and a general journal that posts the freeze trial balance. Add GP history and vendor 1099 only if reporting or 1099 filing needs them inside Business Central, and only after you check the capacity section. Assume the same company moving to Odoo: same minimum documents, opening balances as a journal whose debits equal credits, and closed years left in the legacy database.

  • Prefer a month-end or quarter-end freeze so you reconcile a full period.
  • Load open AR and AP as documents when collectors and payables clerks still apply cash and payments to invoices.
  • State history depth per object (general ledger, AR, AP, inventory, attachments) in the scope list.
  • On GP, decide company by company which data the cloud migration includes before the first replication.
Minimum go-live data versus optional history for an SME ERP migration.
Data classExamplesWhere it lives at go-liveWhy
Master dataCustomers, vendors, items, chart of accountsLive in the new ERPTransactions need a single surviving record and valid posting setup
Reference setupPayment terms, tax codes, units of measure, posting groups, currenciesLive, loaded before mastersDocuments fail or post to the wrong account when this layer is missing
Open transactionsOpen AR, open AP, open sales and purchase orders, on-hand inventoryLive, as of the freezeCollections, payments, fulfillment, and stock depend on document-level balances
Opening balancesFreeze-date trial balance and subledger control totalsLive, mandatoryThe first close has to match the legacy freeze to the cent
US 1099 vendor detailGP vendor 1099 amounts on a GP to Business Central migrationLive only when the US filing needs it in Business CentralMicrosoft's GP validation compares vendor 1099; that test is a US form, and it is separate from Canadian tax
Closed historyPosted invoices, old journals, retired items, attachmentsLegacy system or read-only archiveHistory consumes capacity and time. Load it when a report cannot be answered any other way
Lot and serial on handLot or serial IDs for units still in stock or on open jobsLive for open stockPicks and traceability fail if controlled units arrive as plain quantities

Validation checks a controller can sign

Validation proves the new ledger matches the freeze. An import with zero error rows can still be wrong. Odoo states that a journal entry is balanced only when total debits equal total credits, and that assets equal liabilities plus equity. Bank reconciliation is finished when the bank account balance in Odoo matches the statement balance. Use those same equalities on Business Central.

Worked tie-out, assumptions stated. Assume a single company and a freeze after the legacy close. You export the legacy trial balance, AR aging, AP aging, inventory valuation, and the bank statement balance on that date. The new system passes when the opening journal debits equal credits, assets equal liabilities plus equity, the sum of open customer documents equals the AR control account, the sum of open vendor documents equals the AP control account, the sum of on-hand quantity times cost equals the inventory account, and the bank account equals that statement balance. Any difference stops go-live for that company.

For a GP to Business Central migration, Microsoft's validation compares source and target for G/L accounts, statistical accounts, bank accounts, customers, items, purchase orders, vendors, and vendor 1099. Results show failed-record counts and, on drill-down, the expected GP value versus the Business Central value. Tests can be ignored, which means a green board is a choice. Finance should refuse to ignore a failed G/L, bank, item, vendor, or 1099 validation, and should also refuse go-live when the separate freeze tie-out for trial balance, aging, or inventory value fails.

For Odoo, run Test before Import. Then post the opening journal and reconcile the bank to the statement. The import itself will not roll back. Created-on or last-modified filters are the documented way to find what an import changed. Finance and operations sign the pack; the migration lead signs row counts and the load log.

When the freeze trial balance, aging, or inventory value is what decides go-live, book a free discovery call. Bring the source system, a recent trial balance, and the history you expect to load. Flectic will order those tie-outs for a company of 20 to 100 people in the US or Canada. The written proposal sets a fixed price for each phase before work starts, and the delivery model is designed to deliver up to 3x faster.

Cutover checks for an SME ERP data migration.
CheckOwnerPass criteriaIf it fails
Debits equal credits, and assets equal liabilities plus equityControllerOpening journal balances and the accounting equation holdsStop cutover and fix the opening journal or the account map
Trial balance to the legacy freezeControllerAccount totals match the freeze-date legacy trial balanceHold go-live. Recheck chart mapping and opening entries
Open AR agingAR leadOpen customer documents sum to the AR control accountKeep collections on the legacy system until the aging ties
Open AP agingAP leadOpen vendor documents sum to the AP control accountHold vendor payments from the new system
Inventory quantity and valueWarehouse and costingQuantity times cost ties to the inventory accountFreeze shipping and receiving until the valuation ties
Bank to statementControllerBook balance matches the statement balance used at freezeFinish reconciliation before opening the period
GP validation entities, including vendor 1099 for US companiesMigration lead and controllerFailed-record count is zero on the entities you refuse to ignoreDrill into expected GP value versus Business Central value and reload

Inventory, lots, and open documents at cutover

Open subledgers and controlled inventory cause more go-live pain than the choice of import tool. On-hand conversion is quantity and value as of the freeze, by location, on the costing method the new system will use. If quantity times cost does not equal the inventory account, the ledger can still show equal debits and credits for the wrong reason: the difference is sitting in a conversion account nobody owns.

Agree one rule for receipts that are in the warehouse and still uninvoiced. If those units are inside the on-hand quantity and also inside an open purchase order receipt, the first invoice match after go-live can add them again. The same double count hits open jobs when material already issued to the job is also loaded as free stock. Close what you can before the freeze, and convert the jobs you must finish in the new system.

Lot and serial control applies to units still on hand or still on an open job. Load the identifier with the quantity. Closed lot movements can stay in the archive unless a recall or a lot-level cost method needs them live. If the new system turns lot or serial control on for the first time, warehouse labels have to match the file before the first pick. That labeling time belongs on the cutover runbook.

Open AR and AP should land as invoices and bills when the team still collects, pays, and ages at document level. A single balance-forward journal keeps the control account and drops the aging. Open sales and purchase orders need the item, customer or vendor, tax code, and warehouse already loaded. In Odoo, sales-order lines are one-to-many records: one spreadsheet row per line, and the customer and product have to exist first.

  • Write down whether the on-hand file includes units already on open receipts.
  • Match the cost method before the trial load. Standard cost can load quantity against the standard. Average or actual cost needs the receipt-layer cost from the legacy system.
  • Put first-time lot or serial labeling on the runbook with an owner and a clock time.
Inventory and open-item decisions to lock before the freeze.
ObjectMinimum for go-liveWhat breaksDecision
On-hand by locationQuantity and cost aligned to the freeze valuationAvailable stock and the inventory account disagreeAlways load. Cycle-count before the freeze when the warehouse can
Lot or serial on handIdentifier on every controlled unitPicks fail and traceability stopsRequired when the item is controlled. Label the warehouse in parallel
Received, still uninvoicedA single owner for the quantity and the accrualThe first vendor bill doubles inventory or sticks the accrualCo-load with open purchase orders and name the clearing account
Open jobs and WIPIssued material kept apart from free stockStock is counted twice and the floor works the wrong binConvert open jobs. Close finished jobs before the freeze
Open AR and APInvoice-level open itemsAging and cash application lose the documentUse documents when clerks apply cash to invoices
Closed lot historyUsually the archiveA larger cutover with little operational useKeep open lots live. Archive closed movements

How to judge whether the data is clean enough

Ask the question Panorama puts in front of buyers: is our data clean? Score each master file before mapping. The six dimensions defined by the Data Management Association UK, and used by the UK Government Data Quality Hub, are a practical checklist: accuracy, completeness, consistency, validity, uniqueness, and timeliness. A customer with three spellings, a blank tax identifier, and a payment term that disagrees with the credit file will fail more than one of them.

The 2019 Panorama figure belongs next to this review. In that year's ERP Report, 41% of respondents said data issues caused the project to run over schedule. Cleaning after the first trial load is how that overrun shows up: the same file is mapped, loaded, and reconciled twice. Do the scoring on the legacy extract, assign each failed row to a steward, and reload only the repaired file.

For this conversion, completeness means required target fields are filled: posting group, tax code, unit of measure, and currency. Uniqueness means one surviving customer, vendor, and item. Validity means dates and amounts will pass the loader. Odoo can misread a CSV date such as 01-03-2016 and tells you to use ISO 8601 (YYYY-MM-DD). Parentheses mean a negative number, and a currency symbol on a negative amount must sit inside the parentheses. Timeliness means open balances match the latest subledger close, which is the freeze you will reconcile to.

  • Accuracy: the address, cost, and tax identifier match the real customer, vendor, or item.
  • Completeness: required fields for posting are filled before the trial load.
  • Consistency: the same customer looks the same in the sales file and the ledger.
  • Validity: formats match the target, including ISO dates and signed amounts in Odoo files.
  • Uniqueness: duplicates are merged to one record and open documents point at it.
  • Timeliness: open AR, AP, and inventory match the freeze close, and they match a current extract.

Who owns the data during the migration

Name owners before the first extract. In the DAMA Data Management Body of Knowledge, a data owner is the senior business leader accountable for a domain, and a data steward is the specialist who runs the rules day to day. For a 20 to 100 person company, that is usually the controller for the chart, tax, and 1099 or tax codes, an AR and AP lead for open documents, and a warehouse lead for items and on-hand quantities. One person, the migration lead, keeps the map and the load log.

Master data management, in Gartner's definition, is the discipline of keeping shared masters uniform, accurate, and owned. An SME rarely needs a separate MDM product for this project. You still need the rule: one surviving record, written merge rules (which address, tax identifier, payment terms, and credit limit win), and open documents repointed before the trial load. Three vendor records with three bank accounts is how a payment goes to the wrong account in week one.

Freeze new custom fields once mapping starts. Each late field is cleaned in the legacy file and mapped again in the target. On Business Central, custom tables and fields move only if the customization is an extension installed on both the on-premises database and the online tenant, or if table mappings are defined. Agreeing a field in a workshop does not put it on the migration.

Business Central paths, GP rules, and finance and operations

Microsoft documents cloud migration to Business Central online from Business Central on-premises (version 25 and later directly; versions 15 to 24 only after an on-premises upgrade to at least version 25; version 14 either by that full upgrade or by a reimplementation tool that moves only master data, opening balances, and setup), from Dynamics NAV only after an upgrade to Business Central on-premises, from Dynamics GP 2015 and later, from Dynamics SL 2015 and later, and from other SQL sources with custom migrations. The source must be SQL Server 2016 or later with database compatibility level 130 or higher.

For a non-SQL source, the documented path is configuration packages into Business Central, then the cloud migration tool if the target is online. QuickBooks customers are pointed to Microsoft apps rather than the SQL pipeline. The default configuration package contains 27 tables, including customers, vendors, items, G/L accounts, posting groups, sales and purchase headers and lines, general journal lines, and item journal lines. Microsoft says to leave worksheet columns where they are: move, change, or delete a column and the file will not import. Blob fields cannot be moved through Excel.

Cloud migration copies data with Azure Data Factory. A local SQL Server source uses a self-hosted integration runtime so the copy does not require opening inbound firewall ports. Replication is one way, from on-premises to the online tenant. Running replication overwrites existing data in the online target. Microsoft warns against pointing cloud migration at a production environment already used for business, including when the target is a different company, because some data is shared across companies.

While a cloud migration is connected, users without SUPER are placed on the Intelligent Cloud permission set and the online tenant is treated as read-only for migrated data. On-premises users and permission sets do not replicate. Online users must be Microsoft Entra accounts created and permissioned in the tenant. Standard cloud migration moves only tables that exist on both sides. The version 14 reimplementation path moves essential data and does not require converting customizations to extensions.

GP companies can include history and vendor 1099, and the on-premises system stays the system of record until completion. Replicate more than once, upgrade once, then validate. A second replication after data upgrade risks mixed records. If online storage exceeds 80 GB, some administrative tasks are disabled. Microsoft recommends keeping each migration run under 30 GB, migrating fewer companies per run, and deleting log and archive tables before the run. The first run moves all selected data; later runs move changes only.

Dynamics 365 finance and operations is a different product with a different tool. In the Data management workspace, each entity is staged before it is applied so the file can be checked, cleaned, or converted. Entities in one job run in parallel inside an execution unit when they share a level, and in sequence within a level. Required fields are marked and must be mapped. Finance and operations automatically deletes data-management job history and related staging data older than 90 days. The newer system cleanup job, on by default from version 10.0.20, runs hourly and keeps seven days of execution history, archiving the purged package for 7 days by default and up to 90 days. Those retention rules are finance and operations behavior. They are separate from Business Central cloud migration.

Odoo 19 imports, External IDs, and what will not rebuild itself

Odoo 19 imports into any business object from Excel (.xlsx) or CSV, including contacts, products, bank statements, journal entries, and orders. The documented sequence is upload, map columns, Test, then Import. Imports are permanent and cannot be undone. If a trial file is wrong, you find the rows with created-on or last-modified filters and you correct them with a new import or with manual edits. Test on a copy of the database before you import into the company that will go live.

Odoo matches related records by name, Database ID, or External ID, and says to use only one method per field. External ID lets the same file be imported again to update a record. Many-to-many tags are comma-separated with no spaces. One-to-many lines, such as sales order lines, need one spreadsheet row per line. Related records must be imported before the records that point at them: countries and payment terms, then customers and vendors, then products, then orders and journal lines.

Large imports can time out or fail to process. Split them into smaller batches. For CSV, prefer ISO 8601 dates (YYYY-MM-DD). Parentheses mark a negative number, and a currency symbol on a negative amount sits inside the parentheses. A journal entry is balanced only when total debits equal total credits. Use that test on the opening balance file before Import.

Odoo describes a successful implementation as analysis, configuration, training and coaching of key users, import of data, and customization. Data import assistance shows up on Success Packs starting at the Standard pack (50 hours), which is covered in the cost section. The Starter pack (4 hours) and the Basic pack (25 hours) do not list that row. Assistance hours help you run the import. They are a block of time, and they are separate from a fixed-scope conversion of your ledger.

For an Odoo Community major-version upgrade of an existing database, teams use OCA OpenUpgrade, which moves one major version at a time. That upgrade path is for a database you already run on Odoo Community. A company leaving QuickBooks, Sage, GP, or spreadsheets is on the import path above, with External IDs and a tested opening journal.

Freeze, cutover, and the first weeks after go-live

Most SMEs on one company can cut over on a single freeze date once two full trial loads have tied the controls. A phased cutover, by company or by warehouse, fits when you cannot freeze every site on the same weekend. Each phase still needs its own freeze trial balance. A parallel period, with people posting in both systems, doubles the close work. Use it only for a named window, such as the first period, and only if the business can staff the double entry.

Write the runbook before the dress rehearsal: who freezes which process, the clock time of the final extract, the reconciliation queries, the pass or fail numbers, and the person who can send everyone back to the legacy system. Process bank reconciliations, payment runs, and journals before the hard freeze. Export the freeze trial balance and keep it as the file finance will sign against.

Hypercare is part of the migration. On Business Central, on-premises stays the system of record until migration is completed. While cloud migration is connected, the online tenant is read-only for migrated data and users without SUPER sit on the Intelligent Cloud permission set. Rebuild online users as Microsoft Entra accounts. They do not come across with the SQL copy. On Odoo, a bad import cannot be rolled back, so hypercare fixes are new imports using External IDs or direct corrections, found through created-on and last-modified filters.

Cutover choices for a 20 to 100 person company.
ChoiceWhen it fitsData consequenceWatch-out
Single freezeOne company, rehearsals already tie, month-end is workableOne opening balance set and one agingThe weekend is only as long as the last full-volume rehearsal
Phased by company or siteSeveral companies or warehouses with different readinessA freeze file and a tie-out per phaseOpen orders that cross sites need a written home
Short parallel closeLeadership wants one period of comparison and can staff itDeltas must be identified or people will post twiceThe second close is real work, and it needs an end date

How long the migration takes, and how to shrink the file

Microsoft does not publish an hour count for the initial Business Central replication. The FAQ says duration depends on data volume, SQL configuration, and connection speed, and it asks the same question buyers ask: how do I reduce the volume? Delete log and archive tables, migrate fewer companies per run, and keep each run under 30 GB. The first run moves everything you selected. Later runs move changes only and finish faster. Use that pattern so the freeze weekend is a delta, measured once at full volume.

There is a published product sequence even without an hour count. On GP, plan time for assessment, cleaning, migration setup, company configuration, more than one replication, a single data upgrade, validation, and completion. Do the upgrade once, after the rehearsal replications look right. A replication started after that upgrade can mix upgraded and non-upgraded records.

Odoo time sits in file preparation more than in the click on Import. Related records go first, Test runs before Import, and large files are split so they do not time out. Success Pack hours are the published assistance block: data import assistance starts at 50 hours on the Standard pack, as of October 2026. That clock also covers configuration and training on that pack. It is assistance time, and your cleanse and reconciliation sit around it.

The written scope sets the first phase's timeline and fixed price before work starts. Data audit, mapping, and at least two reconciled trial loads have to fit inside the phase you are buying. What stretches a phase is a late decision to bring closed history, owners named after the first failed trial balance, and a cutover slot booked from a tiny extract. What keeps it intact is a locked entity list, External IDs or repeatable GP replication, and a validation pack with names on it.

  • Book the freeze only after a full-volume rehearsal has a measured duration and a signed tie-out.
  • Shrink Business Central runs: drop logs and archives, fewer companies, under 30 GB per run.
  • Shrink Odoo runs: smaller batches, ISO dates, External IDs, Test before Import.
  • Put the first-phase timeline from the written scope on the same calendar as cleansing, or it slips for data reasons of the kind Panorama reported in 2019.

What Microsoft and Odoo publish, as of October 2026

Separate three money questions. Microsoft does not sell an ERP migration SKU. History consumes Business Central database capacity. Odoo sells assistance hours that include data import help from a defined pack upward, and those hours are a time block. Public partner labor for a scoped conversion was a separate quote. Use the figures below as list prices and quotas, then price the human work as a scoped phase.

Default Business Central database capacity, as of October 2026, is 80 GB plus 3 GB per Premium user, 2 GB per Essentials user, and 1 GB per Device license, shared across all environments in the tenant. Each extra production environment adds 4 GB of tenant-wide database capacity and three sandbox environments. Extra capacity is bought as 1 GB, 100 GB, or overage add-ons through a CSP partner. Going over the quota blocks new environments and copies. It does not stop transactions, and Microsoft does not auto-bill the overage. If online storage exceeds 80 GB, some administrative tasks are disabled, which is a second reason to keep migration runs under the 30 GB guidance.

Worked capacity, assumptions stated. Assume 12 Essentials users, no Device licenses, and no extra production environment. Included database capacity is 80 GB plus 2 GB times 12, which is 104 GB, shared by every environment in the tenant. A first replication that Microsoft would rather see under 30 GB still has to land inside that tenant quota along with configuration data and later years of posting. Adding a second production environment adds 4 GB, to 108 GB on these assumptions, plus three sandboxes. If closed history pushes the tenant over the quota, extra capacity is bought as a 1 GB, 100 GB, or overage add-on through a CSP partner.

Odoo Success Pack prices below are USD list figures as of October 2026, with the new-customer price and the returning-customer price. The first pack is 15% off for a new customer. Hours expire after one year. Data Import Assistance is listed from the Standard pack up. Starter and Basic do not list that row. The estimator on the pack page also says 80% of projects in the illustrated scope go to production in 200 hours or less. That 80% is Odoo's statement about the illustrated scope. It is a pack-page claim, and it is separate from a promise about your ledger.

Odoo.sh hosting changes the history decision again. As of October 2026 the pricing configurator lists a worker at USD 72 per month (USD 57.60 discounted), storage at USD 0.25 per GB per month (USD 0.20 discounted), and a staging environment at USD 18 per month (USD 14.40 discounted). Contact is required above 16 workers or 512 GB. Worked storage, assumptions stated: 200 GB of retained files at the list rate is 200 times USD 0.25, which is USD 50 per month, or USD 40 per month at the discounted USD 0.20 rate. Staging is where you Test an import before it becomes permanent in production.

Odoo Success Pack prices in USD as of October 2026. Data import assistance starts at Standard.
PackHoursNew customer (USD)Returning (USD)Data import assistance
Starter4493580Not listed
Basic2530603600Not listed
Standard5059507000Listed, with configuration and training
Custom1001062512500Listed, plus customizations
Pro2002125025000Listed, plus customizations

Scope the conversion on a discovery call

Read the capacity example and the Success Pack table as list quotas and assistance hours. Microsoft does not sell a migration SKU. As of October 2026, the Standard pack at USD 5,950 for a new customer is 50 hours that include data import assistance along with configuration and training. The cleanse, the trial loads, and the signed freeze tie-out are the work to put on a phase plan.

Book a free discovery call with that table in front of you. Flectic will say which history stays in the legacy system, which open documents must tie to the freeze, and what belongs in the first usable phase. Flectic stays for hypercare while the first close is produced from the new ledger.

What changes between a US company and a Canadian company

The tie-out is the same in both countries: debits equal credits, the trial balance matches the freeze, AR and AP aging match the control accounts, and inventory quantity times cost matches the inventory account. The master-data map is where the countries diverge. Tax codes, posting groups, and currency are mapping work for every company. A US vendor file often carries 1099 boxes. A Canadian vendor file carries GST, HST, or provincial tax setup instead.

Microsoft's GP validation explicitly compares vendor 1099. That is a US information-return form. A pass on that test supports a US 1099 file inside Business Central. Turn the 1099 test on for US companies that will file from Business Central. For a Canadian company, spend the same review time on tax codes, posting groups, and whether amounts are in CAD or in a foreign currency that must convert at the freeze.

Odoo Success Pack prices in the cost section are the USD amounts on Odoo's pricing page. Use those USD figures when you compare assistance hours. Currency conversion for your own books is a mapping choice (which currency is company currency, which open invoices stay in the document currency), and it is independent of the currency printed on the pack page.

Opening documents still have to post under the target tax setup. Load tax codes and posting groups before customers, vendors, and items. Then load open invoices. A trial load that ignores tax will tie a pretax aging and fail the first real invoice. Flectic implements this map for companies of 20 to 100 people in the US and Canada as part of the phase, then stays for hypercare while the first close is produced from the new ledger.

Where ERP data migrations go wrong

The failures that reach a controller are repeated often enough to plan for them. Dirty masters get loaded, history scope grows after the design is signed, and the freeze weekend is estimated from a small file. The product rules add a second list: a GP replication after data upgrade, an Odoo import run in the live database without Test, and a Business Central cloud migration pointed at a tenant where people are already posting.

Panorama's June 2019 article warned that teams expect the new ERP to standardize whatever they dump into it, and the 2019 ERP Report found 41% of respondents linking data issues to a schedule overrun. The current product pages say the same thing in operational language. Business Central will overwrite the target, will leave users behind, and will move custom fields only under the extension or mapping rules. Odoo will keep a bad import. Capacity and the 30 GB run guidance punish a decision to lift every closed year.

Avoid the expensive variants. Cleanse before the first paid load. Lock the minimum go-live menu and treat history as optional. Reconcile the freeze trial balance, aging, inventory value, and bank on every trial. For GP, replicate, upgrade once, then validate, and leave a failed G/L, bank, item, vendor, or 1099 validation active. Hold go-live when the separate freeze tie-out fails. For Odoo, Test, then Import, with External IDs and batches. Rebuild permissions on purpose. Keep the legacy system readable through hypercare.

  • Loading duplicates and blank posting groups, then discovering them when the first invoice fails.
  • Adding years of closed history after the trial load was sized, and blowing the 30 GB run guidance or the tenant quota.
  • Posting open AR and AP as one journal when clerks still need invoice-level aging.
  • Loading inventory quantities without cost layers, locations, or lot and serial IDs.
  • Pointing Business Central replication at a production company that already has transactions, or replicating again after data upgrade.
  • Importing into live Odoo without Test, then discovering the load cannot be undone.
  • Expecting on-premises users and permissions to appear in the Business Central tenant.
  • Buying an Odoo Starter or Basic pack and assuming data import assistance is included. That row starts at Standard, 50 hours, USD 5,950 new or USD 7,000 returning as of October 2026.
  • Treating the GP vendor 1099 validation as proof that a Canadian tax setup migrated.
  • Signing off because row counts match, while the freeze trial balance, aging, or inventory value does not.

How Flectic can help

Flectic moves master data, open documents and a chosen slice of history into Odoo or Dynamics 365 for companies of 20 to 100 people in the US and Canada, and checks migrated balances before anyone relies on them. Our Odoo industry packages run 8 to 28 weeks, and a first phase outside the packages gets its own timeline in the written scope. Review our ERP implementation service, then book a free discovery call with Flectic.

Sources

  1. As of October 2026, Microsoft documents cloud migration to Business Central online from Business Central on-premises (version 25 and later directly; versions 15 to 24 only after upgrade to at least version 25; version 14 by that upgrade or by a reimplementation tool that moves only master data, opening balances, and setup), from NAV only after upgrade to Business Central on-premises, from GP 2015 and later, from SL 2015 and later, and from other SQL sources with custom migrations. Source must be SQL Server 2016 or later with compatibility level 130 or higher. Non-SQL sources use configuration packages, then cloud migration if the target is online. QuickBooks is pointed to Microsoft apps rather than the SQL pipeline.

    learn.microsoft.comChecked Microsoft Learn migrate-data, seen 2026-10.
  2. Business Central cloud migration copies data with Azure Data Factory. A local SQL Server uses a self-hosted integration runtime so inbound firewall ports stay closed. Replication is one way. While connected, users without SUPER use the Intelligent Cloud permission set and migrated data is read-only online. On-premises users and permission sets do not replicate; online users must be Microsoft Entra accounts. Standard migration moves tables that exist on both sides; custom tables and fields require an extension on both sides or table mappings. The BC14 reimplementation path moves essential data and does not require converting customizations to extensions. The first replication moves all selected data and later runs move changes only. If online storage exceeds 80 GB, some administrative tasks are disabled. Microsoft recommends each run under 30 GB, fewer companies per run, and deleting log and archive tables. Duration depends on volume, SQL configuration, and connection speed.

    learn.microsoft.comChecked Microsoft Learn cloud migration FAQ, seen 2026-10.
  3. On GP to Business Central, the tool can move system and company setup, masters for accounts, customers, vendors, and items, transactional data, classes, vendor 1099, and historical data. Which company data moves is configurable. On-premises stays the system of record until migration is completed. Sequence: assessment, preparation including cleaning, cloud migration setup with no data moved yet, company configuration, repeatable data replication, data upgrade, validation, completion. After a successful data upgrade, replicating again from an earlier version risks mixed upgraded and non-upgraded records. Replication overwrites the online target. Do not point cloud migration at a production environment already used for business, including a different company, because some data is shared across companies.

    learn.microsoft.comChecked Microsoft Learn GP migration overview, seen 2026-10.
  4. GP to Business Central validation compares G/L accounts, statistical accounts, bank accounts, customers, items, purchase orders, vendors, and vendor 1099. Results show failed-record counts and, on drill-down, the expected GP value versus the Business Central value. Tests can be ignored.

    learn.microsoft.comChecked Microsoft Learn GP migration validation, seen 2026-10.
  5. Default Business Central database capacity is 80 GB plus 3 GB per Premium user, 2 GB per Essentials user, and 1 GB per Device license, shared across all environments. Each extra production environment adds 4 GB of tenant-wide database capacity and three sandbox environments. Extra capacity is bought as 1 GB, 100 GB, or overage add-ons through a CSP partner. Going over the quota blocks new environments and copies, does not stop transactions, and Microsoft does not auto-bill the overage.

    learn.microsoft.comChecked Microsoft Learn capacity page, seen 2026-10.
  6. The default Business Central configuration package contains 27 tables, including customers, vendors, items, G/L accounts, posting groups, sales and purchase headers and lines, general journal lines, and item journal lines. Moving, changing, or deleting worksheet columns stops the import. Blob fields cannot be moved through Excel.

    learn.microsoft.comChecked Microsoft Learn configuration packages, seen 2026-10.
  7. In Dynamics 365 finance and operations, the Data management workspace stages each entity before it is applied. Entities in one job run in parallel inside an execution unit when they share a level, and in sequence within a level. Required fields are marked and must be mapped. Job history and related staging data older than 90 days are deleted automatically. The system cleanup job, on by default from version 10.0.20, runs hourly, keeps seven days of execution history, and archives the purged package for 7 days by default and up to 90 days.

    learn.microsoft.comChecked Microsoft Learn finance and operations data import and export jobs, seen 2026-10.
  8. Odoo 19 imports into business objects from Excel (.xlsx) or CSV, including contacts, products, bank statements, journal entries, and orders. Sequence: upload, map columns, Test, then Import. Imports are permanent and cannot be undone. Created-on or last-modified filters find what an import changed. Related records match by name, Database ID, or External ID, one method per field. External ID supports update on re-import. Many-to-many tags are comma-separated with no spaces. One-to-many lines need one spreadsheet row per line. Related records are imported first. Large imports should be split. CSV dates should be ISO 8601 (YYYY-MM-DD). Parentheses mean a negative number, and a currency symbol on a negative amount sits inside the parentheses.

    odoo.comChecked Odoo 19.0 export and import data, seen 2026-10.
  9. Odoo states that a journal entry is balanced only when total debits equal total credits, that assets equal liabilities plus equity, and that bank reconciliation is finished when the bank account balance in Odoo matches the statement balance.

    odoo.comChecked Odoo 19.0 accounting cheat sheet, seen 2026-10.
  10. As of October 2026, Odoo describes a successful implementation as analysis, configuration, training and coaching of key users, import of data, and customization. Data Import Assistance starts at the Standard Success Pack (50 hours), and is not listed on Starter (4 hours) or Basic (25 hours). USD prices: Starter 493 new / 580 returning; Basic 3060 new / 3600 returning; Standard 5950 new / 7000 returning; Custom (100 hours) 10625 new / 12500 returning; Pro (200 hours) 21250 new / 25000 returning. First pack is 15% off for new customers. Hours expire after one year. The estimator says 80% of projects in the illustrated scope go to production in 200 hours or less. Pack prices are assistance hours for that scope.

    odoo.comChecked Odoo Success Packs pricing, seen 2026-10.
  11. As of October 2026, Odoo.sh lists a worker at USD 72 per month list and USD 57.60 discounted, storage at USD 0.25 per GB per month list and USD 0.20 discounted, and a staging environment at USD 18 per month list and USD 14.40 discounted. Contact is required above 16 workers or 512 GB.

    odoo.comChecked Odoo pricing configurator, seen 2026-10.
  12. In June 2019 Panorama Consulting wrote that most ERP vendors do not take responsibility for data migration or cleansing because it is costly, and that 41% of respondents in its 2019 ERP Report said data issues caused the project to run over schedule. The article advises against moving every legacy record and expecting the new software to standardize it.

    panorama-consulting.comChecked Panorama Consulting, ERP Data Migration and Cleansing Tips, June 2019, reviewed 2026-10.
  13. IBM describes database schema conversion as its own step: if the target data-definition language is incompatible, the schema must be converted before the data moves, and integrity should be checked after the transfer. The page addresses infrastructure transfer.

    ibm.comChecked IBM Think, What is data migration?, seen 2026-10.
  14. The six data quality dimensions (accuracy, completeness, consistency, validity, uniqueness, timeliness) are defined by DAMA UK and used by the UK Government Data Quality Hub.

    gov.ukChecked Existing guide source is the UK Government Data Quality Hub page that cites DAMA UK for the six data quality dimensions.
  15. DAMA-DMBOK distinguishes a data owner (senior business leader accountable for a domain) from a data steward (specialist executing governance). Gartner defines master data management as a technology-enabled discipline for uniformity, accuracy, stewardship, governance, semantic consistency, and accountability of shared master data.

    dama.orgChecked Existing guide source is the DAMA-DMBOK page used for the data owner and data steward roles, with the Gartner master data management definition cited alongside it in the prior guide.
  16. OCA OpenUpgrade is the primary open-source tool for Odoo Community database upgrades and requires version-by-version migration. This applies to upgrading an existing Odoo database, which is a different job from importing legacy ERP files.

    github.comChecked Existing guide source is the OCA OpenUpgrade repository, which documents the version-by-version path for upgrading an existing Odoo Community database.

Frequently asked questions

What is ERP data migration?

ERP data migration is the controlled load of master data, open transactions, and a chosen slice of history into the new ERP so the freeze-date trial balance, AR aging, AP aging, and inventory match the legacy system. People also search this as ERP migration or ERP data conversion. Conversion is the reformatting of accounts, units, tax, and posting groups inside that program. The sign-off is those freeze balances, which a database copy or a hosting move leaves untested.

What is the difference between ERP data migration and ERP data conversion?

ERP data conversion is the transform: source fields are rewritten so they fit the target chart, tax codes, units of measure, and posting groups. ERP data migration is the full sequence around that transform, from the audit and the freeze through trial loads, validation, cutover, and hypercare. A search for ERP data conversion is asking whether those day-one balances match the legacy freeze. This playbook is the plan for that search.

Which products and versions can move to Business Central online?

Microsoft documents Business Central on-premises version 25 and later directly, versions 15 to 24 after an on-premises upgrade to at least version 25, and version 14 by that upgrade or by a reimplementation tool that moves master data, opening balances, and setup. Dynamics NAV moves only after an upgrade to Business Central on-premises. Dynamics GP 2015 and later and Dynamics SL 2015 and later are supported. The source must be SQL Server 2016 or later with database compatibility level 130 or higher. Other SQL sources need a custom migration, and non-SQL sources use configuration packages.

How is on-premises data migrated to Business Central online?

Cloud migration copies data with Azure Data Factory. A local SQL Server uses a self-hosted integration runtime so you do not open inbound firewall ports. Replication is one way, from on-premises to the online tenant, and it overwrites data already in the target. On GP the sequence is assessment, preparation, setup with no data moved yet, company configuration, repeatable replication, one data upgrade, validation, then completion. The on-premises system stays the system of record until migration is completed.

Are there limits on how much data can be migrated?

Standard Business Central cloud migration moves tables that exist on both sides. Custom tables and fields need an extension installed on both sides or a defined table mapping. Microsoft recommends each migration run stay under 30 GB, with fewer companies per run and log and archive tables deleted first. If online storage exceeds 80 GB, some administrative tasks are disabled. Included database capacity starts at 80 GB plus per-user gigabytes, shared across the tenant. Going over the quota blocks new environments and copies and does not stop transactions.

Will custom tables, users, and permissions migrate?

Custom tables and fields migrate when the customization is an extension on both the on-premises database and the online tenant, or when table mappings are defined. The version 14 reimplementation path moves essential data only, so it does not require converting customizations to extensions. On-premises users and permission sets do not replicate. Online users must be Microsoft Entra accounts created in the tenant. While migration is connected, users without SUPER are placed on the Intelligent Cloud permission set and migrated data is read-only online.

How long does the initial migration take, and how do we reduce volume?

Microsoft says the duration depends on volume, SQL configuration, and connection speed, and it does not publish a fixed hour count. The first replication moves all selected data and later runs move changes only, so they are faster. Reduce volume by deleting log and archive tables, migrating fewer companies per run, and keeping each run under 30 GB. Leave closed history in the legacy system when you do not need it as open documents. Flectic plans these rehearsals into the first phase, and the written scope sets that phase's timeline before work starts.

What do Odoo Success Packs include for data import?

As of October 2026, Data Import Assistance is listed from the Standard pack (50 hours) upward, at USD 5,950 for a new customer and USD 7,000 for a returning customer. Starter (4 hours, USD 493 new or USD 580 returning) and Basic (25 hours, USD 3,060 new or USD 3,600 returning) do not list that row. Custom is 100 hours at USD 10,625 new or USD 12,500 returning, and Pro is 200 hours at USD 21,250 new or USD 25,000 returning. These are assistance hours, including configuration and training on Standard, and they expire after one year. They are a time block, and your trial balance is a scoped project around that block.

Is our data clean enough to migrate?

Score customers, vendors, items, and the chart for accuracy, completeness, consistency, validity, uniqueness, and timeliness before you map. Duplicates, blank posting groups, and tax codes that will not post are the usual blockers. In Panorama's 2019 ERP Report, 41% of respondents said data issues caused the project to run over schedule. Cleanse in the legacy extract, then trial-load. Odoo cannot undo an import, and Business Central replication overwrites the online target, so the first load into a live company is a poor place to discover dirty rows.

Should we migrate all historical transactions?

Load the minimum go-live set: chart, customers, vendors, items, open AR, open AP, on-hand inventory, open orders, and opening balances. Add closed history only when a report or a filing needs it inside the new system. GP cloud migration can include historical data and vendor 1099, and that 1099 test is a US form. Business Central version 14 reimplementation moves essential data and opening balances. Odoo's published import loads the files you prepare, including journal entries and orders. Closed years that you only need to look up can stay in the legacy system or a read-only archive.

More on the blog: Implementation and go-live