A8gent

Planner

Data Migration Planner

Select your source and target systems for a data migration complexity score, an estimated timeline, and the approach least likely to lose records.

Interactive planner

Plan a data migration between systems

Select your source and target to get complexity rating, time estimate, and a step-by-step checklist.

Free

Guided action plan

1Audit source data: count records and identify duplicates
2Map fields from source schema to target schema
3Clean data: fix formats, remove orphaned records
4Export source data (CSV or API)
5Run test import with 10-20 records
6Validate test: check field mapping accuracy
7Full migration: import all records
8Post-migration audit: verify counts and integrity
9Set up ongoing sync if bidirectional needed

Migration readiness

73

Needs prep

Complexity 3/10. Estimated time: 3-6 hours.

Complexity

3/10

Low complexity - straightforward migration.

Estimated time

3-6 hours

With automation tools. Manual migration takes 2-3x longer.

Recommended course

Auto guided

No-Code AI Implementation Course

Learn how to plan and execute data migrations using no-code tools, including field mapping, testing, and validation.

Open course path
Contents
  1. Fast answer
  2. Quick facts
  3. What this covers
  4. Why it matters
  5. Where to go next
  6. Outcomes
  7. Step by step
  8. What your complexity score is really measuring
  9. Cost and timeline benchmarks for planning
  10. The tooling landscape in 2026: pick by volume and shape
  11. How to stay out of the overrun statistics
  12. The methodology: what the planner actually looks up
  13. The unglamorous middle: field mapping, duplicates and dates
  14. A worked example, and the systems the planner does not know
  15. What the planner ignores, and when to route elsewhere
  16. At a glance
  17. Mistakes to avoid
  18. FAQ
TL;DRAnswer-first

Fast answer

The data migration planner gives you a complexity score, estimated time, recommended approach, and specific tools for automating a transfer once you select your source and target systems. The complexity score from 1 to 10 factors in schema compatibility between source and target, the number of field types requiring transformation such as dates, picklists, and relationships, volume sensitivity, and whether linked records like contacts-to-companies keep their associations. It covers the 10 most common business systems with specific paths for common pairings like HubSpot to Salesforce or Sheets to Airtable, and provides a general framework for pairs outside its database. For scores 1 to 4, self-service is feasible in a few hours; scores 5 to 7 suggest a freelancer or migration service; scores 8 to 10 warrant a specialist because data-loss risk rises with complexity. To avoid data loss, keep source data read-only for 30 to 90 days, test with a small batch first, validate record counts, spot-check individual records, and migrate during low-activity hours.

Facts last verified 2026-08

Quick facts

Migration overrun rate
80%+ of projects run over time or budget, Bloor Group research
Average overruns
30% on cost, 41% on schedule, Bloor Group
Cost of poor data quality
$12.9M average per organization per year, Gartner (2020 research, still cited)
Data believed inaccurate
25% on average at US companies, Experian, Nov 2024
Typical CRM migration
4-8 weeks under 100K records; 8-14 weeks with custom objects, 2026 guides
Basic migration cost
$1,500-$7,500 for contacts and deals; $30K-$90K+ for complex moves

On this page

What this page covers

A tool visitor should leave with a decision, not just a number: build now, prepare first, choose another workflow, or follow a course path.

  1. 01Systems
  2. 02Complexity
  3. 03Plan
  4. 04Tools

Why does this matter now?

Select your source and target systems. Get a migration complexity score, estimated time, recommended approach, and specific tools for automating the transfer.

Internal path

Where to go next from this page

These links are part of the A8gent learning and conversion path. Use them to move from concept, to diagnosis, to workflow build, to course.

Start with readiness

What you should be able to do after this

  • Complexity score
  • Time estimate
  • Migration approach
  • Tool recommendations

How do you do it, step by step?

1. Count what you have before you move it

Record counts per object, duplicates by match key, empty required fields, and the oldest record you would actually look for. This is the baseline you reconcile against afterwards.

2. Write the field map as a document

Source field, target field, transformation rule, and the person who decides when the answer is ambiguous. Free-text statuses into picklists is where this earns its keep.

3. Decide the match key and the merge rule

Email for people, domain for companies, normalised phone where email is patchy. Dedupe in the source, where you still have the context to judge.

4. Decide what stays behind

Stale contacts, ancient lost deals and activity logs older than your reporting horizon go to an archive. Migrating them costs money and dilutes every report and every agent you later build.

5. Test with 10 to 20 records, then reconcile counts

Check the test rows by eye, not by count. Then run the full import, compare counts per object, and spot-check twenty records at random before anyone starts working in the new system.

What your complexity score is really measuring

The planner scores complexity on record volume, schema mismatch between source and target, custom objects, and integration count, because those are the variables that historically blow up migration budgets. Bloor Group research, the most-cited body of work on migration outcomes, found more than 80 percent of data migration projects run over time or budget, with cost overruns averaging 30 percent and schedule overruns averaging 41 percent. A related figure claiming 83 percent of migrations fail circulates widely with a Gartner attribution, but no public primary source exists for it, so treat the well-documented Bloor overrun numbers as your planning baseline.

A high score does not mean do not migrate; it means the default timeline is wrong. Migration guides for CRM moves in 2026 put a basic contact and company migration at 1 to 3 weeks, a typical assisted migration at 4 to 8 weeks for under 100,000 records, and complex migrations with custom objects at 8 to 14 weeks. If your score lands in the top tier, plan for the longer end plus the average 41 percent schedule overrun rather than the vendor's happy-path estimate.

The score also proxies for data quality exposure, which is the silent budget item. Experian research from November 2024 found US companies believe an average of 25 percent of their data is inaccurate, and Gartner has put the average annual cost of poor data quality at $12.9 million per organization, a figure from its 2020 data quality research that the firm still promotes. Every inaccurate record you migrate becomes an inaccurate record you now trust in a new system, which is why cleansing belongs before the move, not after.

Cost and timeline benchmarks for planning

Published 2026 pricing for CRM migrations gives usable brackets. A straightforward contact, company, and deal migration runs roughly $1,500 to $7,500 when handled by a partner; professional services engagements for mid-complexity moves run $5,000 to $20,000; and migrations involving custom objects, years of activity history, and multiple integrated systems run $30,000 to $90,000 or more. Doing it in-house is cheaper on invoices and more expensive in attention, and the overrun statistics above were mostly generated by in-house teams.

Realistic time budgeting matters more than cost budgeting because the overrun risk is larger on the schedule side. A 60 to 90 day plan for a meaningful CRM migration is the norm in current guides, split roughly into audit and cleansing, field mapping and test migration, and cutover with validation. The single most effective de-risking step in that sequence is the test migration of a representative sample, because it surfaces the mapping surprises while they are still cheap to fix.

Budget explicitly for the post-migration tail. Validation, deduplication, retraining the team on the new system, and running the old system read-only for a transition period all consume real hours after go-live. Teams that declare victory at cutover consistently rediscover the project two months later as a data quality complaint, which is the expensive way to learn that migrations end at verified adoption, not at data transfer.

The tooling landscape in 2026: pick by volume and shape

For one-time moves between common SaaS systems, native importers and vendor migration services are usually the cheapest correct answer. For recurring or high-volume pipelines, the ETL platforms take over. Fivetran prices on Monthly Active Rows with a free tier of 500,000 MAR per month, and its 2026 pricing update added a $5 base charge per connection and made deletes count toward paid MAR, so re-quote if you costed it before the change. Airbyte offers an open-source self-hosted option with 600+ connectors, cloud plans starting at $10 per month, and published volume pricing around $15 per million rows for API sources.

Automation platforms occupy the middle ground for operators who already run them. Zapier's Transfer feature moves existing records in bulk, with each transferred record consuming one task on your plan, which is fine for hundreds of records and ruinous for millions. n8n and Make.com can do the same style of batched record movement with execution-based pricing that scales better, and they double as the place where post-migration sync workflows live, which is why many teams route small and medium migrations through the automation platform they already pay for.

AI agents changed the labor profile of migration in 2025 and 2026 rather than the fundamentals. LLM-assisted field mapping, deduplication, and data classification cut the manual grind of the audit and mapping phases, and agents built on Claude or similar models can normalize free-text fields at a scale that used to require interns and weekends. What they do not change is accountability: the test migration, the validation queries, and the sign-off remain human jobs, and the overrun statistics argue for keeping them that way.

How to stay out of the overrun statistics

Scope ruthlessly on what not to migrate. Every migration guide and most post-mortems agree that dragging seven years of stale activity history into a new system inflates cost, extends timelines, and imports garbage. Archive the long tail somewhere queryable and migrate the working set. Cutting migrated volume is the only lever that simultaneously reduces cost, schedule, and data quality risk.

Freeze the source schema during the migration window and assign a single owner for mapping decisions. The 41 percent average schedule overrun that Bloor documented is substantially a coordination failure: fields keep changing, two teams map the same object differently, and the test migration invalidates itself. One owner, one mapping document, and a change freeze convert an open-ended project into a bounded one.

Finally, define success as reconciled counts and spot-checked records, not as the absence of error messages. Run row counts and checksum comparisons between source and target, sample records per object type for field-level verification, and get written sign-off from the people who will live in the new system. The Experian finding that companies already believe a quarter of their data is inaccurate is your warning: without explicit verification, a migration silently launders old errors into new confidence.

The methodology: what the planner actually looks up

The planner is a lookup, not a calculator, and saying so is the fastest way to make it useful. You pick a source and a target from ten systems: Google Sheets, HubSpot, Salesforce, Airtable, Notion, Excel or CSV, Monday.com, Pipedrive, Mailchimp and QuickBooks. That is 90 ordered pairs. The source has twelve hand-written entries covering Sheets, HubSpot and Salesforce as the source system, each with a complexity score out of ten, a time range, a written approach and a tool list. Every other combination falls back to a per-source default, and any source outside those three falls back to a global default of 4 out of 10 and 3 to 6 hours. So if you select Pipedrive to Airtable you get the same answer as Mailchimp to Notion, because neither pairing exists in the table.

The scores that are written down encode a real and useful asymmetry. Sheets to Airtable is 2 and 1 to 2 hours. Sheets to HubSpot is 3 and 2 to 4 hours. Sheets to Salesforce is 5 and 4 to 8 hours, because of picklists and lookup relationships. Salesforce to Sheets is only 3 and 1 to 2 hours, because pulling data out is nearly always easier than pushing it in. HubSpot to Salesforce and Salesforce to HubSpot are both 7 and 8 to 16 hours, the highest in the table, because a CRM-to-CRM move has to preserve associations between contacts, companies and deals rather than just rows. The banding is fixed: 2 or less reads LOW, 3 to 4 MEDIUM, 5 to 6 HIGH, 7 and above VERY HIGH.

Two output tiles deserve less trust than they look. The estimated time is described as 'with automation' and reflects hands-on configuration and testing for a modest record count, not the calendar time a migration occupies once approvals, a freeze window and a validation pass are included. The 'tools needed' tile is simply the length of the recommended tools array, which is three or four for every pairing in the table, so it carries no information about your situation at all. The nine-item checklist at the bottom is likewise identical for every combination. Read it as a competent generic order of operations, which it is, rather than as advice tailored to the pair you picked.

The unglamorous middle: field mapping, duplicates and dates

Field mapping is where migrations are won and lost, and it is nearly always underestimated because it looks like a spreadsheet exercise. It is not. Mapping is a series of small decisions about meaning: your source has one Name field and the target wants first and last; your source has a free-text Status that four people have filled in nine different ways and the target wants a picklist with five options; your source stores phone numbers as whatever the person typed and the target validates E.164. Every one of those is a rule someone has to write down, and the rules that get skipped are the ones that produce a target system full of records nobody trusts. Do the mapping as a document before you touch an importer, and give every ambiguous field a named owner who decides.

Duplicates are the second reliable disaster, and the important insight is that a migration does not create them, it reveals them and then multiplies them. Two records for the same customer in the source become two in the target, plus a third if you re-run a partial import after a failure. Decide the match key before you export: email is the usual choice for people, domain for companies, and a normalised phone number for the trades and clinics where email is often missing. Decide the merge rule too, meaning which record wins when the two disagree, and decide it in favour of the most recently updated rather than the oldest. Then run the dedupe in the source, where you still have context, rather than in the target, where you have lost it. A target system's import wizard will happily create both records and tell you it succeeded.

The third thing to settle in advance is what not to migrate at all. Historical records that no one has opened in three years, contacts that have never opened an email and have no phone number, deals lost more than two years ago, and any activity log older than your reporting horizon are all candidates for an archive rather than an import. There are three good reasons. You will pay for them in storage or record tiers, they will pollute every list and report in the new system, and they will dilute any AI agent you later point at the CRM, because an agent grounded in stale records gives confidently stale answers. Export the archive to cold storage, note where it lives, and migrate the live business. The bar we suggest: if you cannot name the query that would send you looking for a record, it does not need to be in the live system.

A worked example, and the systems the planner does not know

Take a two-office dental practice moving marketing and recall out of spreadsheets and into HubSpot. Selecting Sheets to HubSpot returns complexity 3, MEDIUM, 2 to 4 hours, and an approach that says export CSV, map columns to HubSpot properties, use the import wizard and clean data types such as dates and phones first. That is accurate as far as it goes, and for 4,000 patient contacts with six clean columns it really is an afternoon. The part it does not model is that the practice management system, Dentrix or Open Dental, holds the appointment history that makes recall work, and that system is not one of the ten. The spreadsheet is a copy of a copy, and migrating it faithfully migrates a stale snapshot.

That gap is the honest limitation of this tool for most small service businesses. The systems that actually run these companies are absent from the list: Jobber, Housecall Pro and ServiceTitan in the trades, Dentrix and Open Dental in dentistry, Clio, MyCase and Smokeball in law, AppFolio, Buildium and Yardi in property management, Mindbody, Vagaro, Boulevard and Phorest in salons, ezyVet, AVImark and Cornerstone in veterinary practice, Toast and Square in restaurants, Shopify with Klaviyo and Gorgias in ecommerce, Follow Up Boss and kvCORE in real estate, and QuickBooks, Xero and Karbon in accounting. For any of those, the planner returns its global default of 4 out of 10 and 3 to 6 hours, which is a placeholder rather than an estimate. What transfers is the method: audit, map, dedupe, decide what to leave behind, test small, validate counts.

Finish the dental example properly and the sequence looks like this. Pull the patient list from the practice management system rather than the spreadsheet, because that is the system of record. Normalise phone numbers, because recall runs on SMS. Dedupe on phone rather than email, because families share addresses and half the older records have no email at all. Exclude anyone marked inactive or deceased, which is the single most important exclusion in a healthcare migration and the one most often missed. Import 20 records, check them by eye, then import the rest and reconcile counts. Only then connect anything that sends messages. Our recall automation guide covers what happens after the data lands, and the appointment reminders guide covers the message layer.

What the planner ignores, and when to route elsewhere

There is no record count anywhere in the tool, which means volume does not affect the score or the time. That is defensible up to a point, because a well-mapped import of 50,000 rows is not much slower than 5,000, but it breaks at the boundaries: API rate limits, import row caps and the cost of a re-run after a failure all scale. There is also no notion of ongoing sync versus one-off move, no allowance for attachments and files, no handling of custom objects, and no treatment of the freeze window, meaning the period where staff must stop editing the source. In practice the freeze is the part that annoys people most and it belongs in your plan even though it is absent from the tool's.

A migration is also the moment your data quality becomes visible to everyone, which is a management problem before it is a technical one. Counts that do not reconcile, records that arrive with blank owners, a pipeline where a third of deals sit in a stage nobody uses: none of this is created by the move, but all of it surfaces during the move and gets blamed on it. Say so in advance, publish the pre-migration audit numbers, and the conversation afterwards is about cleanup rather than about whether the project failed.

Use a different tool when the question is not about moving records. If you want to know whether two systems can talk to each other on an ongoing basis, use the integration checker. If the real goal is to put an agent on top of the destination once the data lands, the CRM AI agent checklist covers the data hygiene floor, and the onboarding planner covers the fourteen days after go-live. If a vendor is quoting you for the migration, our automation build cost and AI automation cost pages give you the range to check the quote against, and the free automation audit will tell you whether the move is worth doing at all. Where a competitor wins outright: for a CRM-to-CRM move with custom objects, a dedicated migration service or the target vendor's professional services team will beat a do-it-yourself pipeline on both time and risk, and the 8 to 16 hour estimate the tool returns for HubSpot to Salesforce is the optimistic end of that job.

At a glance

Data migration cost and timeline benchmarks (published 2026 guides)

ScopeTypical costTypical timelineBest-fit tooling
Contacts and companies only$1,500-$7,5001-3 weeksNative importers, vendor migration service
Full CRM under 100K records$5,000-$20,0004-8 weeksVendor tools plus partner services
Custom objects, deep history$30,000-$90,000+8-14 weeksProfessional services, ETL platform
Recurring sync pipelineUsage-basedOngoingFivetran (MAR pricing), Airbyte ($10/mo+ cloud or self-hosted)
Small bulk moves, few thousand recordsPlatform plan tasksDaysZapier Transfer (1 task per record), n8n, Make.com

What mistakes should you avoid?

  • Treating the complexity score as specific when your pairing is not in the table and you are reading the global default of 4 out of 10.
  • Mapping fields in the importer instead of in a document, so nobody can say later what rule was applied.
  • Deduplicating in the target system, after the context that would let you judge which record is right has been lost.
  • Migrating everything because it is easier than deciding what to leave behind.
  • Skipping the freeze window, so records edited during the move exist in two versions and neither is correct.
  • Letting the old system stay writable after go-live, which guarantees a slow split-brain rather than a clean cutover.
  • Connecting anything that sends messages to customers before the counts have been reconciled.

FAQ

How is the complexity score calculated?
It is not calculated, it is looked up. The tool holds twelve hand-written pairings covering Google Sheets, HubSpot and Salesforce as the source system, each with a fixed score, time range, approach and tool list. Anything else falls back to a per-source default, and any source outside those three falls back to a global default of 4 out of 10 with a 3 to 6 hour estimate. So an unusual pairing returns a placeholder rather than an assessment. Check whether your exact pair is one of the twelve before you plan around the number.
Why is Salesforce to Sheets easier than Sheets to Salesforce?
Because export is nearly always easier than import, and the tool encodes that: Salesforce to Sheets scores 3 and 1 to 2 hours, while Sheets to Salesforce scores 5 and 4 to 8. Pulling data out is a report and a CSV. Pushing it in means satisfying validation rules, matching picklist values exactly, resolving lookup relationships to record IDs and respecting required fields. The asymmetry is real and it is the single most useful thing the lookup table knows.
Does record count affect the estimate?
No. The tool never asks how many records you have, so 500 rows and 500,000 rows return the same score and the same hours. Up to a point that is fair, because the mapping work dominates. It breaks at the edges: API rate limits, import row caps, timeouts on large files, and the cost of re-running a failed batch all scale with volume. Above roughly 100,000 records, or wherever your target's importer starts refusing files, treat the estimate as a floor.
What should I deliberately not migrate?
Contacts with no activity and no reachable channel, deals lost more than two years ago, activity history older than your reporting horizon, and anything marked inactive, deceased or do-not-contact. Archive them somewhere retrievable and note where. Three reasons: you often pay per record or per storage tier, stale records pollute every list and report in the new system, and any AI agent you later ground in that CRM will answer from whatever it can see. A smaller, truer database is worth more than a complete one.
How do I handle duplicates?
Decide the match key and the merge rule before you export, then dedupe in the source system where you still have the context to judge which record is right. Email works for people, web domain for companies, and a normalised phone number for trades, clinics and salons where email coverage is patchy. When two records disagree, prefer the most recently updated rather than the oldest. Importers do not merge by default, and a partial re-run after a failure will cheerfully create a second copy of everything that already landed.
Will migrating make our data quality look worse?
It will make it visible, which feels the same to everyone watching. Counts that do not reconcile, blank owners, statuses nobody recognises and a pipeline stage used by one person: none of that is created by the move, but all of it surfaces during the move. Manage it by publishing the pre-migration audit numbers first, so the discussion afterwards is about a cleanup backlog you already knew about rather than about whether the migration broke something.
Does the planner cover industry systems like Jobber, Clio or Dentrix?
No. The ten systems it knows are Google Sheets, HubSpot, Salesforce, Airtable, Notion, Excel or CSV, Monday.com, Pipedrive, Mailchimp and QuickBooks. The systems that actually run most service businesses, including Jobber, Housecall Pro, ServiceTitan, Dentrix, Open Dental, Clio, AppFolio, Mindbody, ezyVet, Toast and Follow Up Boss, are not in the list, so those selections are impossible and the nearest equivalent returns the generic default. The method still transfers: audit, map, dedupe, decide what to leave, test small, reconcile counts.
Should we do it ourselves or hire someone?
Self-service is reasonable for the low-complexity pairings the tool scores at 2 or 3, which are broadly spreadsheet-shaped moves into a system with a decent import wizard. For a CRM-to-CRM move with custom objects, preserved associations and pipeline history, the honest answer is that a migration specialist or the target vendor's professional services team will usually beat a first attempt on both time and risk. The 8 to 16 hours the tool quotes for HubSpot to Salesforce is the optimistic end of that job and assumes nothing custom.

Sources & further reading

Was this page helpful?