Phase 0

Decide & Plan

Lock the decisions that scope every downstream phase — Kinetic vs on-prem, vertical, history depth, cutover anchor.

Confirm the Epicor product, vertical, and deployment model under contract. Document whether the target is Kinetic (public cloud SaaS), Epicor 10 / Epicor for Manufacturing on-premise, Prophet 21 (distribution vertical), BisTrack (building products), or iScala (international). Each vertical ships a different Business Object catalog and a different DMT template set.
Risk if skipped: Building a mapping spreadsheet against the wrong vertical surfaces missing tables and templates only after staging begins.
Confirm the exact Epicor release line under contract. Capture the major.minor.patch (for example Kinetic 2025.2 or Epicor 10.2.700). Customizations and DMT template versions are pinned to the release.
Risk if skipped: A DMT template built for one release silently rejects columns added in another, producing partial loads that pass row counts but fail field-level validation.
Define migration scope by module and business object. Document every in-scope object explicitly: Customer, Vendor, Part, BOM, Routing, GLAccount, GLJrnHed, APInvHed, ARInvHed, OrderHed/OrderDtl, POHeader/PODetail, Project, Job, FAAsset, EmpBasic, and any UD tables. Mark each as full-history, open-only, or out-of-scope.
Risk if skipped: Undocumented scope around open jobs and historical AP/AR is the single biggest cause of cutover slippage.
Decide how many fiscal years of transactional history migrate. Pick a cutoff in writing. Reddit and Epicor User Forum threads document 20-year historical loads that bloated LaborDtl past 2.3M rows and pushed BAQs to time out. Common cuts: open documents plus 1–3 closed fiscal years; anything older stays in an external archive or UD-table read-only store.
Risk if skipped: Importing decades of closed-period documents inflates the database, slows BAQs, and forces 2 TB+ backup windows that nobody costed in the implementation plan.
Decide the cutover date relative to the fiscal calendar. Anchor go-live to a period boundary — typically Period 1 of a new fiscal year. The opening-balance Journal Entry is simpler when it lands on a clean period and the fiscal calendar in Company Configuration has no partial periods.
Risk if skipped: A mid-period cutover forces partial-period journals and double-counts revenue and expense across the legacy and Epicor ledgers.
Decide the multi-company architecture. Pick the company structure in Epicor: single company, multi-company with consolidations, or multi-site within a single company. The choice fixes GLAccount segment design, intercompany trading, and how DMT keys parent/child records.
Risk if skipped: Switching from single to multi-company after master data loads requires rekeying Company on every downstream record.
Identify GDPR, SOC2, HIPAA, and retention constraints. Capture residency rules that govern which Epicor cloud region hosts the tenant (US, EU, APAC) and how long backups are retained. For EU subsidiaries on a US-hosted tenant, document the cross-border transfer mechanism before any personal data leaves the source.
Assign a single accountable owner for the migration. Name one DRI plus named leads for finance, distribution, manufacturing, projects, IT, and end-user enablement. Every cross-team decision routes through the DRI.
Document explicit success criteria for go-live. Write down what must be true to declare cutover complete — trial balance ties to source within a defined tolerance, AR and AP aging match by customer and vendor, inventory quantity-on-hand reconciles by part and warehouse, open sales and purchase orders all loaded, BOMs and routings approved.
List out-of-scope items in writing. Examples: historical email threads, legacy custom SSRS reports, terminated employee records past retention, closed jobs older than the cutoff, change-log entries pre-cutover. Circulate the list to all stakeholders.
Risk if skipped: Items not explicitly out-of-scope reappear as last-minute demands during sandbox testing.
Set and communicate the source-system freeze window. Pick the date and time after which no new transactions post in the legacy system. Broadcast it to every team that writes to the legacy.
Risk if skipped: Late writes to the source after final export create a delta no one reconciles until users notice missing transactions post-cutover.

Phase 1

Pre-Migration Prep

Set up the destination tenant, cleanse the source, and prepare the people side before any byte moves.

1a. Source-system audit and cleansing

Run dedup analysis on source customer and vendor masters. Match on tax ID, legal name + country, and primary contact email. Resolve flagged duplicates in the source before export — merging in Epicor after the fact requires reversing AR and AP history.
Risk if skipped: Duplicate customers loaded into CustGroup or Customer tables fragment AR aging and break credit-limit logic.
Standardize part-number formatting in the source. Force a single case (commonly upper), strip leading or trailing whitespace, and replace separators that conflict with Epicor's PartNum constraints. Part-number normalization done in the source is one-pass; done in Epicor after load it cascades across PartBin, Job, Quote, and Order tables.
Risk if skipped: Mixed-case part numbers create phantom duplicates in inventory and break BOM resolution.
Normalize phone, country code, and address formatting. Pick one phone format (E.164 recommended), one country-code standard (ISO 3166-1 alpha-2), and one address-line convention. Apply across Customer, Vendor, Contact, and EmpBasic source records.
Identify and resolve records with missing required fields. Epicor business objects enforce non-null on a defined set of fields per release. Run a profile pass against the source for blank Company, CurrencyCode, SalesTerritory, ClassID (parts), and similar mandatories.
Risk if skipped: DMT silently skips rows missing required fields and reports them only in the job log, so partial loads can pass row-count checks but fail business validation.
Resolve picklist value mismatches against Epicor user codes. Source values for terms, ship-via, FOB, currency, country, sales tax, and similar must map to entries you will load into UserCodes. Build the mapping now, not at transform time.
Re-encode all source extracts to UTF-8 without BOM. Latin-1 and Windows-1252 source files corrupt diacritics in customer and vendor names on import. Force every extracted file to UTF-8 before any transformation.
Risk if skipped: Mojibake in customer and contact names is hard to spot until a user reports an email going to the wrong person.
Export current record counts per object as a baseline. Persist a dated baseline CSV with one row per source object and its row count. Use this for Phase 6 reconciliation deltas.

1b. Destination tenant setup

Provision production and sandbox/pilot environments. Confirm pilot is a clean clone of the intended production configuration. Lock the version of pilot to the version production will go live on; do not let Epicor's cloud auto-upgrade pilot mid-project.
Risk if skipped: A pilot one minor release ahead of production silently diverges DMT templates and Business Object signatures.
Configure Company Configuration before any data lands. Set base currency, fiscal calendar, period structure, tax liabilities, default plant, and country-of-incorporation. These drive validation on every downstream import.
Risk if skipped: Changing base currency after master data loads requires a full reload — Epicor stores monetary values pre-converted on each row.
Build the chart of accounts and segment structure first. Define the GL Control segment plus department, division, project, or other COA segments via Chart of Accounts Maintenance. The segment structure is foundational and rebuilding it after journal data lands is extremely costly.
Risk if skipped: Changing segment count or width post-load forces a database-level rewrite of every posted GL row.
Define the fiscal calendar and open the relevant periods. Set fiscal-year start, period count, and period close cadence in Company Configuration. Open the periods you intend to post into; closed periods reject DMT loads.
Load currencies and set up daily exchange-rate import. Create every currency the business transacts in, then configure the exchange-rate maintenance pattern (manual entry or BAQ-driven) before any multi-currency document is loaded.
Risk if skipped: Loading foreign-currency invoices without an exchange rate for the document date posts them at zero rate and corrupts the GL.
Create all UD fields and UD tables before any DMT load. Use User Defined Column Maintenance and Solution Workbench to add UD columns to every business object that needs them. DMT will only populate columns that exist when the template is generated.
Risk if skipped: DMT silently drops data in columns that do not exist in the destination, so missing UD fields cause invisible loss.
Create UserCodes for every source picklist mapped in 1a. Load reason codes, terms, ship-via, FOB, sales territory, and similar via UserCodes Maintenance or the DMT UserCode template before any document import references them.
Provision named users with appropriate security groups. Create every user in the User Account Security screen — or via the DMT UserFile template — with the correct security groups, company access, and site access before cutover.
Stage Solution Workbench solutions for BPMs, BAQs, dashboards, and customizations. Build BPMs that enforce business rules and personalizations as Solutions in pilot. Export the solutions so they ship to production as a single artifact at cutover.
Pause all outbound integrations on the destination. Disable scheduled Service Connect workflows, EDI triggers, BPM email notifications, and any external integration that fires on master-data or document creation during load.
Risk if skipped: Integration noise during DMT loads sends spurious notifications to customers and vendors and pollutes downstream systems with not-yet-real data.

1c. People prep

Communicate the freeze window to every source-system user. Calendar invite, internal wiki page, and stand-up announcement. Confirm acknowledgment from each functional lead.
Schedule role-based training on Kinetic before cutover. Per the 2026.1 release notes the Classic smart client is being retired; train only on the browser-based Kinetic UI. Segment training by role: finance, sales, purchasing, inventory, production.
Risk if skipped: Training users on the Classic shell creates double change-management debt when the shell is removed in a later release.
Designate first responders for the first 72 hours post-cutover. Name one person per functional area who triages user issues and routes data discrepancies to the migration DRI.

Phase 2

Source Export

Get every byte out of the source in a form you can transform.

Export GL chart of accounts with full segment values. One row per account; include every segment (control, department, division, project, etc.). Preserve account type and posting flag.
Export the GL trial balance as of cutover for every period in scope. One file per period. Debits and credits per account-segment combination. This is the source of truth for the opening-balance journal in Phase 3.
Risk if skipped: Without a frozen trial balance snapshot you cannot prove the destination ties to the source in Phase 6.
Export customer and vendor masters with all standard and UD fields. Include billing and shipping addresses, terms, credit limit, tax ID, default currency, and any extension fields. Export contacts separately keyed to the parent record.
Export the Part master with UoM Class, default plant, and class. Capture PartNum, base UoM, UoM Class, costing method, product group, class, and the inventory and revenue GL controls.
Risk if skipped: Missing UoM Class on the Part master breaks every transaction that converts between purchase, stock, and sales UoMs.
Export BOMs and routings, parent-then-child. Multi-level BOMs export as a flat list of parent PartNum + child PartNum + quantity + UoM rows; routings as Operation + Resource Group + standard times. Maintain the sequence numbers exactly.
Export inventory on-hand, on-order, allocated, by Part, Plant, Warehouse, and Bin. One row per Part-Plant-Warehouse-Bin-LotNum-SerialNum combination. Lot and serial-controlled parts need every dimension.
Risk if skipped: Aggregating quantities above the bin or lot level loses the granularity Epicor needs to reconcile cycle counts.
Export open AP and AR invoices with aging buckets and remaining balance. Header plus line for each open document. Capture original date, due date, terms, currency, original amount, and current open amount.
Export open sales orders, purchase orders, jobs, and projects. Header and line for each. Open quantity and dollar amounts only; shipped or received quantities are out of scope unless explicitly in.
Export payment, cash-receipt, and bank-account configuration. Bank accounts, AR payment terms, payment methods. These configure cash-application behavior on day one.
Export attachments and document metadata from the source DMS. For each attachment, capture the binary file, the source record key (e.g. invoice number), the file name, MIME type, and original upload date. Note Epicor Kinetic Cloud attachments cap at approximately 175 MB per file per the user forum; flag any source file larger than the cap.
Risk if skipped: Files larger than the destination cap need a Phase 3 decision (split, store externally, or skip); discovering the cap at cutover breaks the attachment migration.
Capture configuration exports for every report, dashboard, and automation. SSRS reports, BAQs, dashboards, scheduled jobs, workflow rules, and external integrations are rebuilt manually in the destination. Screenshot or XML-export each one for the Phase 7 rebuild.
Timestamp every export file and persist the manifest. One CSV manifest row per exported file with filename, source-table name, row count, and ISO-8601 export timestamp. This is the audit record of what was in scope at cutover.

Phase 3

Transform & Map

Turn raw source exports into Epicor-shaped DMT input files.

3a. Mapping spreadsheet

Build a master mapping spreadsheet keyed by Epicor business object. One tab per target object (Customer, Vendor, Part, BOM, Routing, GLAccount, OrderHed, OrderDtl, APInvHed, ARInvHed, etc.). For each source field, document the destination column, the transformation rule, nullability, and whether it is required by the DMT template.
Risk if skipped: Without a master mapping, multiple analysts make conflicting decisions about the same field and the contradictions surface only in sandbox.
Decide the upsert key per object. Typical keys: CustNum or CustID for Customer, VendorNum or VendorID for Vendor, PartNum for Part, plus Company on every row. Persist the key choice in the mapping sheet.
Decide handling for source fields with no destination equivalent. Per field, pick one: drop, concatenate into a Comment or Notes column, or create a UD field. Implement UD-field creation in Phase 1b before this transform completes.
Validate every transformed value against the destination type catalog. Epicor types include nvarchar with hard length limits, decimal(28,8) for monetary, datetime, and bit. Truncation or rounding rules must be explicit in the mapping sheet.
Risk if skipped: Implicit truncation on long Description fields silently loses information that auditors will eventually ask about.

3b. Data transformation

Convert all dates to ISO-8601 with explicit timezone. Use yyyy-MM-ddTHH:mm:ssZ. DMT accepts shorter forms but ambiguity around DD/MM versus MM/DD parses silently wrong on locale-different rows.
Risk if skipped: Date-format drift on transactional dates pushes documents into the wrong fiscal period.
Normalize decimal separator and precision per Epicor type. Use dot decimals everywhere. Round monetary values to the precision the destination column accepts (commonly 4 or 8 places). Truncate, do not round, on quantity fields where the source carries spurious extra digits.
Truncate or split source fields that exceed Epicor column limits. For each source field longer than the destination, decide: truncate with logged loss, split into multiple fields, or overflow into a Comment Text record. Document the choice in the mapping sheet.
Re-encode every transformed file to UTF-8 without BOM. DMT accepts UTF-8 cleanly; a leading byte-order mark on column-1 of row-1 makes the first column look unrecognized.
Risk if skipped: A stray BOM is a one-character bug that fails the entire file with a misleading column-name error.
Escape commas, quotes, and newlines in free-text fields. DMT consumes CSV and XLSX. For CSV, quote every field containing a comma, double inner quotes, and replace embedded newlines with a literal escape if the destination type does not accept multi-line content.
Generate external-ID columns on every object for upsert idempotency. Carry the source primary key into a UD column (e.g. UD_LegacyID) on every object. This is the rerun-safe handle if a Phase 5 load step fails partway.
Risk if skipped: Without external IDs, re-running a failed load creates duplicates or silently overwrites unrelated rows.

3c. Relationships, audit trail, and attachments

Decide and implement strategy for original document dates. Source create-dates on AP, AR, and sales orders are valuable for aging. Decide per object: load the source date into the destination Date field (best for aging accuracy) or capture it in a UD_OriginalDate field while the create-stamp reflects load date.
Decide and implement record-ownership strategy. Either map source owner to the matching Epicor user by email, or assign every migrated row to a placeholder data-migration user and reassign in Phase 7.
Decide and implement the attachment strategy. Two paths: re-upload binaries into Epicor's attachment store (subject to the per-file size cap), or keep binaries on an external store and load only the URL into a UD field. Decide per attachment class.
Enable change-log data directives only after historical load completes. Change-log capture writes a row to the change-log tables on every update. Leaving directives enabled during DMT loads multiplies database growth by the number of migrated rows.
Risk if skipped: Loading 1M+ rows with change-log directives enabled balloons the change-log table and degrades load throughput.

3d. ERP-specific transforms

Build the GLAccount file with full segment values per the destination COA. One row per account combination. Validate every segment value against the segment masters loaded in Phase 1b. Mark account type (Asset, Liability, Equity, Revenue, Expense) and posting flag.
Risk if skipped: Loading GL transactions before GLAccount completes pushes every line to a suspense account and reconciliation has to be re-done.
Prepare the GL opening-balance Journal Entry as a Period 1 BegBal journal. Build one Journal Entry whose lines mirror the Phase 2 trial balance snapshot. Post it to Period 1 of the cutover fiscal year. Exclude current-year P&L lines if you also intend to import current-year posted invoices, or you will double-count income.
Risk if skipped: Either an opening journal that includes YTD income or current-year invoices loaded on top of an opening journal that already contains them double-counts income.
Verify debits equal credits on the opening-balance journal before staging. A balanced journal is required to post. Run the sum check before saving the DMT input file.
Build the Part master file with UoM Class and default plant. One row per Part. Reference the UoM Class loaded in 1b; reference the Plant loaded in 1b. Set base UoM, costing method, product group, class, inventory GL control, and revenue GL control.
Build the UoM Class, UoM, and UoM Conversion files before the Part master loads. Epicor's Advanced UoM module supports dual UoMs and attribute-driven conversions. Decide per Part class whether Advanced UoM applies. Stage UoM Class first, UoM second, conversions third, and Part fourth.
Risk if skipped: Loading Parts before the UoM Class they reference fails the row; loading them after a partial UoM Class load corrupts the link with no hard error.
Build the BOM file parent-then-child, level by level. Multi-level BOMs cannot load bottom-up — DMT validates that the child PartNum exists when the parent row references it. Sort the file by BOM level ascending. Stage Routings in parallel keyed to the same Part.
Risk if skipped: An out-of-order BOM file rejects the child row before the parent and produces an incomplete bill of materials.
Build inventory-on-hand files at Part-Plant-Warehouse-Bin-LotNum-SerialNum granularity. Match the source export granularity from Phase 2. Lot-controlled parts need every LotNum row; serial-controlled parts need every SerialNum row. Quantities post via the Quantity Adjustment business object, not by writing PartBin directly.
Risk if skipped: Direct writes to PartBin bypass costing logic and break the inventory-to-GL tie at the next inventory revaluation.
Stage daily exchange rates for every active currency. Load enough history of daily rates to cover the dates on every multi-currency document being imported. The CurrencyExRate business object accepts DMT uploads.
Risk if skipped: Missing a rate for a document date causes the document to post at the wrong converted amount, silently corrupting the multi-currency GL.
Prepare open AP invoices for DMT into APInvHed and APInvDtl. Header plus lines per open invoice. Document date drives aging; due date drives the cash forecast. Validate that every Vendor and PurPoint exists from Phase 1b.
Prepare open AR invoices for DMT into ARInvHed and ARInvDtl. Header plus lines per open invoice. Validate that every Customer, ShipTo, and salesperson exists. Carry tax amounts in the source currency, not pre-converted.
Build the AR and AP open-payment files keyed to invoices. Partial payments need to attach to the invoice they applied against. Stage Cash Receipts and Payment Headers keyed by invoice GroupID.
Build fiscal-calendar and period-close configuration before the first transactional load. Confirm fiscal years and periods are open in Fiscal Calendar Maintenance for every period any document references. Closed periods reject DMT row-by-row.

Phase 4

Sandbox Test Migration

Run every DMT template in pilot before touching production.

Load a 1–5% representative sample of each object into pilot. Sample by date range (e.g. one month of orders), customer tier (top 10 plus bottom 10), and part class (one row per ProdGroup). Use DMT's standard templates.
Validate field-by-field on 10 random sandbox rows per object. Open the record in the Kinetic UI and compare every field against the source extract. Spot edge cases: long descriptions, special characters, NULL handling, multi-currency values.
Risk if skipped: Row-count checks pass while field-level corruption hides until a user opens the record in production.
Run a full-volume DMT load against pilot for the largest object. Per the DMT documentation, large jobs are limited by application-server throughput. Time the load and extrapolate to production volume.
Risk if skipped: Loading 500K rows in pilot at the wrong concurrency setting wedges the application server and stretches cutover past the planned window.
Test multi-level BOM and Routing import in dependency order. Confirm the BOM file sorted by level ascending loads cleanly. Confirm Routings load against the same Part and that resource groups resolve.
Risk if skipped: An out-of-order BOM load rejects child rows and leaves an incomplete bill of materials that production-control will catch only at the first job.
Test the opening-balance Journal Entry posts cleanly to Period 1. Stage the BegBal journal via DMT, post it, run a trial balance, and tie the trial balance back to the source.
Risk if skipped: An unbalanced or wrong-period opening journal taints every downstream financial report from go-live forward.
Test inventory quantity adjustments to PartBin. Load on-hand via QuantityAdjustment, then run a Stock Status Report. Confirm Part-Plant-Warehouse-Bin-Lot-Serial totals match source on-hand.
Test multi-currency invoice import for a non-base currency. Load one foreign-currency AP and one foreign-currency AR invoice. Confirm both the document amount and the converted amount post correctly using the staged daily rate.
Risk if skipped: A misconfigured exchange-rate type produces invoices that look right on the document and wrong on the GL.
Test attachment upload via DMT XFileAttch. Upload a representative sample of attachments, including the largest file in scope, to confirm none of them hit the per-file cap (approximately 175 MB on Kinetic Cloud).
Risk if skipped: Discovering the file-size cap mid-cutover forces an emergency external-storage plan.
Run a full-cycle business test in pilot: create order → ship → invoice → cash. Pull a real customer, create a sales order with migrated parts, ship from a migrated bin, invoice in a migrated period, apply cash. Confirm GL postings are correct.
Time the end-to-end sandbox load and extrapolate to production. Record wall-clock per object. Multiply by the production-to-pilot row ratio and compare to the cutover-window budget.
Risk if skipped: An end-to-end load that runs longer than the cutover window in pilot will not magically run faster in production.
Have finance, operations, and a power user sign off pilot results. Written sign-off naming the artifacts reviewed and the issues outstanding. Block Phase 5 on sign-off.

Phase 5

Production Cutover

A tightly-sequenced execution window — load order matters, idempotency matters more.

Confirm source freeze and verify no new writes in the last hour. Query the source for the latest update timestamp on every in-scope table. Confirm it is older than the freeze cutoff.
Risk if skipped: A single late source write produces a reconciliation delta nobody hunts down for weeks.
Take the final source export (delta from Phase 2 baseline). Re-export every in-scope object. Diff against the Phase 2 export and merge the delta into the transformed files.
Confirm BPMs, data directives, and integrations are paused on the destination. Disable in-transaction and post-processing data directives, BPM email actions, and Service Connect schedules. Re-enable only after Phase 5 completes.
Risk if skipped: A live BPM emailing every Customer record-creation event spams thousands of contacts during cutover.
Load masters first: UserCodes, Plants, Warehouses, Bins, COA segments. These are the foundation. Every subsequent load depends on these existing.
Load users, then Customer and Vendor masters. UserFile via DMT, then Customer via DMT, then Vendor. Each by its DMT template.
Load Part master, UoM Class, UoM, BOM, and Routing in dependency order. UoM Class → UoM → UoM Conversion → Part → BOM (level ascending) → Routing. Each step must complete before the next begins.
Risk if skipped: Skipping the level-ascending order on BOM leaves orphan child rows that production-control surfaces at first job creation.
Post the GL opening-balance Journal Entry to Period 1. Load via DMT, then post via the Journal Entry business object. Run a trial balance immediately and compare against the Phase 2 source snapshot.
Risk if skipped: Posting before debits=credits is checked corrupts every downstream financial report.
Load inventory on-hand via Quantity Adjustment. Run the QuantityAdjustment DMT template. Run a Stock Status Report and tie back to the source.
Load open AP invoices into APInvHed/APInvDtl, then open AR into ARInvHed/ARInvDtl. Header file before line file for each. Load partial payments after both invoice loads complete.
Load open sales orders, purchase orders, jobs, and projects. Header before lines on every object. Open quantities and balances only; do not load already-shipped or already-received history unless explicitly in scope.
Load attachments last, keyed to the records that own them. Attachments depend on the parent record existing. Run XFileAttch after parent loads succeed.
Capture import job IDs, row counts, and error logs for every step. Persist the DMT log file in a dated cutover-evidence folder. These are the audit record.
Risk if skipped: Without logs you cannot reproduce a failed load idempotently or prove a number to an auditor later.
Re-enable BPMs, data directives, change-log capture, and integrations. Only after every load step has signed-off row counts. Smoke-test each integration with a single record before unpausing the rest.

Phase 6

Validate

Prove the migration is correct before any end user logs into production.

6a. Reconciliation

Compare per-object row counts: source vs destination. Build a BAQ per object that returns the destination row count and place it next to the Phase 2 baseline. Investigate every non-zero delta.
Risk if skipped: A 1% row-count delta sounds small until it lands on AR invoices and the customer with the missing row notices first.
Run the Epicor Trial Balance and tie to the source trial balance. Per the 10.2.400 Reconciliation Reporting reference, run GL Trial Balance, GL Opening Balance, AP Aging, AR Aging, Open POs, and Material Transactions; tie each back to the source.
Risk if skipped: An unreconciled trial balance at go-live becomes a year-end audit finding.
Reconcile AP aging by vendor. Run AP Aging in Epicor and compare per-vendor balances to the source aging snapshot. Differences should be zero or explainable.
Reconcile AR aging by customer. Run AR Aging and tie per-customer balances to the source. Confirm bucket cutoffs match the source aging policy.
Reconcile inventory on-hand by Part, Plant, Warehouse, Bin. Run a Stock Status Report or BAQ that aggregates PartBin and compare to the Phase 2 source export.
Risk if skipped: Inventory rebuilds done after go-live break the GL-to-inventory tie because cost layers were never recreated.
Cross-check 25–50 random records per object field-by-field. Sample customers, vendors, parts, invoices, orders, BOMs, and attachments. Compare every populated field in source vs destination.

6b. Relationship validation

Verify parent-child relationships on a sample. Order header → order lines, invoice header → invoice lines, BOM parent → BOM children, Job → JobAsmbl → JobOper → JobMtl.
Verify ship-to and bill-to associations on customers. Each migrated customer should resolve to the right CustShipTo and CustBillTo. Spot-check 20 random customers.
Verify attachment-to-record links resolve and the file opens. Open the attached file from the parent record in the Kinetic UI for a random sample of 25 attachments. Confirm the right file opens.
Verify UD-field values arrived intact on every object that has UD fields. DMT silently drops data in non-existent columns; this is the last chance to catch a UD field that was renamed between mapping and load.

6c. Audit and compliance

Confirm change-log data directives are now enabled in production. Per the audit-trail discussion in the Epicor User Forum, change-log capture is enabled per table via data directive on the columns you care about. Activate the agreed-upon set in production.
Run the GDPR data-subject lookup against a known person. If the destination tenant hosts EU personal data, confirm the data-subject access workflow returns all the fields it should. Validate the data-residency commitment matches the tenant region.
Risk if skipped: A GDPR mismatch found in audit costs more than the entire migration project.
Snapshot the Phase 5 evidence pack (DMT logs, row counts, sign-offs) to long-term storage. SOC2 and HIPAA evidence requirements typically demand the migration log be retained alongside the migrated data.

6d. User-acceptance check

Have power users from each function run their day-1 workflow on real migrated data. Sales creates an order from a migrated customer. Purchasing creates a PO against a migrated vendor. Inventory issues material to a migrated job. Finance posts a payment against a migrated invoice.
Document data issues surfaced by UAT and decide fix-now vs fix-later. Triage every issue: blocking (fix before go-live), cosmetic (fix in Phase 7), or acceptable. Persist the triage decision.

6e. Sign-off

Get written sign-off from the migration sponsor before opening production. Name the artifacts: trial-balance match, AR/AP aging match, inventory match, UAT triage decisions, evidence pack. No production login until sign-off.

Phase 7

Post-Migration Cleanup

Wrap up so the team can move on.

Announce go-live with login instructions and a where-to-get-help contact. Include the Kinetic URL or on-prem launch path, named first responders, and a known-issues list.
Run role-based training sessions on the Kinetic browser UI. Segment by function. Confirm every named user has logged in once within the first 5 business days.
Set the source system to read-only and announce the read-only date. Revoke write permissions in the source. Keep read access for the planned retention window.
Rebuild reports and dashboards in Epicor via BAQ + SSRS or Kinetic dashboards. Source reports do not transfer. Use the Phase 2 configuration captures to rebuild the equivalent in BAQs, BAQ Reports, and SSRS.
Risk if skipped: Missing reports at month-end forces finance back to spreadsheets and erodes confidence in the new system.
Promote BPMs, personalizations, and customizations from pilot via Solution Workbench. Export the pilot solution package and import into production. This is the upgrade-safe path documented in the Epicor User Forum.
Risk if skipped: Hand-recreating BPMs in production diverges them from pilot and breaks the next upgrade.
Re-establish integrations to external systems. EDI, banking, CRM, payroll, calendar sync, BI tools. Smoke-test each one before unpausing scheduled jobs.
Reassign ownership from the data-migration placeholder user to real users. If you used a placeholder owner in Phase 3, run a BAQ-driven update to set CreatedBy and ChangedBy to the real owners now.
Schedule the 30-day retrospective. Capture what went well, what didn't, and the residual issue list from Phase 6 UAT triage.
Decide and document the source-system decommission date. Typically 90 days after cutover. Confirm the retention obligation is independently satisfied by the archived exports.
Archive all source exports, mapping spreadsheets, DMT input files, and DMT logs. Long-term cold storage labeled by cutover date. This is the forensic record if a discrepancy surfaces months later.
Risk if skipped: Without archived exports, a year-end audit question about a migrated transaction has no source of truth to reference.
Close out the migration project formally with a final status report. Final task counts, residual issues with owners, lessons learned, and the link to the 30-day retrospective.

Watch list

Risks to track throughout

These risks live across multiple phases — keep an eye on them from kickoff through cutover.

DMT silently drops data into columns that do not exist

The Data Management Tool will load a row whose mapping references a UD column or table that was never created in the destination — the row succeeds in DMT but the missing-column values vanish. Audit the destination column list once in Phase 1b, again immediately before each Phase 4 sandbox load, and again before each Phase 5 production load. Treat any rename of a UD field as a blocking change that requires re-running every dependent DMT template.
Opening-balance journal and current-year postings double-count income

If the opening-balance journal includes current-year P&L lines and you also import the current-year invoices, revenue and expense are counted twice. Decide in Phase 0 which approach you are using — snapshot opening balances at fiscal-year-end with current-year documents loaded fresh, or skip current-year P&L lines in the opening journal — and verify the trial balance ties in Phase 4 sandbox and again in Phase 6 production.
Customizations and integrations do not lift-and-shift to the cloud

On-premise BPMs that hit the database directly, Service Connect workflows that reach external services on the LAN, and reports tied to Progress OpenEdge tables do not run on Kinetic Cloud. Inventory every customization and integration in Phase 2, decide rebuild-via-Epicor-Functions vs accept-standard vs accept-the-gap in Phase 3, and stage the rebuilds in pilot during Phase 4 — not after cutover.