SaaS Data Migration for Customer Onboarding: Build or Partner?
SaaS data migration for customer onboarding: what building data import in-house really involves, what it costs to maintain, and when to partner instead.
Published 13 min read

On this page
- What is the build vs buy decision for SaaS data migration?
- What does building a help desk data migration tool involve?
- How long does it take to build a migration tool, and who does it?
- What does an in-house importer cost to keep running?
- When does building your own importer make sense?
- Build vs partner: how do the costs compare?
- Is there a middle path between building and buying?
- How does a migration partnership work for software vendors?
- Build vs buy FAQ
- Changelog
SaaS data migration is the step in customer onboarding where a new customer's records move from their old platform into yours. For a help desk, ITSM or CRM software vendor, build vs buy is one question: do you build your own help desk data migration tool for customers switching in, or partner with a migration platform? Build it when one or two source platforms bring most of your switchers, the data is simple, and you can fund a team to maintain each connector for years. Partner when switchers arrive from many platforms, deals need full history with attachments, or onboarding is already the bottleneck.
This guide is for product leaders, onboarding and professional services heads, and partnership managers at software vendors. If you are a buyer deciding how to move your own service desk, our guide to data migration companies vs in-house is the one for you.
What is the build vs buy decision for SaaS data migration?
For a software vendor, the build vs buy decision is whether your engineers write and maintain the importers that move a new customer's records out of a competitor's platform, or a migration platform does it, often under your name. It decides who owns every connector, every failed record and every source API change.
The question usually shows up in one of three ways:
- Sales: a prospect asks, late in the deal, "how do we get five years of tickets out of our current tool?" and nobody has a confident answer. A field-mapping framework for sales directors helps answer it early.
- Onboarding: data onboarding is the slowest part of SaaS customer onboarding, and your implementation team is writing one-off scripts per customer, and each one takes longer than the last.
- Product: someone proposes "a proper importer" as a roadmap item, and the scope keeps growing in planning.
Each of these is the same decision seen from a different seat. SaaStr founder Jason Lemkin argues vendors should invest in migration because a switching customer's lifetime value pays for it: "Even if it costs you $20K or more per deal, it's worth it." He also says that even with a partner, the vendor should "own the whole process yourself, in house." Most vendors agree on the first point. The build vs buy question is about the second: owning the process does not have to mean writing every connector.
What does building a help desk data migration tool involve?
Building a help desk data migration tool means writing one connector per source platform, and each connector has to read and rebuild a whole object model: users, organizations, groups, tickets, public replies and internal notes, attachments, custom fields, SLA data, knowledge base articles and history. A CSV upload screen covers a small part of that.
Here is what a production-grade importer has to handle for each source:
| Component | What it has to do | Where it usually breaks |
|---|---|---|
| Authentication and access | Connect to the customer's source with their credentials or an admin token, with scopes their security team accepts | Tokens expire mid-run; customers will not grant admin access |
| Extraction | Page through every object inside the source's rate limits and pagination rules | Paging stops early; archived or deleted records are skipped |
| Users, organizations and groups | Create people first, including departed agents, so every ticket and comment can resolve its author | History credited to one integration user |
| Tickets and their fields | Map status, priority, type and every custom field, including picklist values that do not exist in your product | Unmapped values left blank |
| Comments | Keep public replies public and internal notes private, in original order, with original author and time | Internal notes exposed to requesters |
| Attachments and inline images | Download, re-upload and re-link every file; rewrite inline image URLs that point at the old platform | Images break when the old account closes |
| SLAs and metrics | Decide where historic SLA outcomes live, since timers should not restart on closed tickets | Reports show years of breached SLAs |
| Knowledge base | Move articles, sections, images and translations, and fix internal links | Dead links inside articles |
| Links between records | Keep problem, change, parent and CC links, plus CMDB items for ITSM sources | Links stored as plain text |
| Automations in your product | Mute triggers and notifications during the load | Requesters emailed about years-old tickets |
| Delta sync | Pick up records that changed after the bulk load, so agents keep working until cutover | A freeze of days instead of hours |
| Validation | Reconcile counts, diff fields and check files, then hand the customer evidence | Nobody can prove the move worked |
Multiply that table by every platform your customers leave. A vendor competing with Zendesk, Freshdesk, ServiceNow, Jira Service Management and Salesforce Service Cloud needs five connectors, each with its own object model and limits.
Which platform limits shape the build?
Rate limits and import rules decide how long a migration takes and which data can be rebuilt faithfully. They differ by platform and by the plan the customer is on. These are the documented limits on two common source and target platforms, checked on 11 Nov 2026:
| Platform | Documented limit or rule | What it means for your importer |
|---|---|---|
| Zendesk Support API | 200 requests per minute on Team, 400 on Professional, 700 on Enterprise, 2,500 with the High Volume API add-on (Zendesk API docs) | Extraction speed depends on the plan your prospect pays for, not on your code |
| Zendesk incremental exports | 10 requests per minute, or 30 with the High Volume add-on (Zendesk API docs) | Large accounts take hours to read even with a perfect connector |
| Zendesk offset pagination | Limited to the first 100 pages and 10,000 resources; past that the API returns a 400 error (Zendesk pagination docs) | A connector built on offset paging silently tops out at 10,000 records |
| Zendesk Ticket Import API | Up to 100 tickets per bulk request; "metrics and SLAs are not supported for imported tickets"; triggers do not run on imported tickets until they are next updated (Zendesk Ticket Import API docs) | Historic SLA data needs its own home, and your load plan needs batches |
| Freshservice API | 100 to 500 calls per minute by plan, with lower sublimits per operation such as Create Ticket (Freshservice API docs) | Writes into Freshservice are throttled separately from reads |
| Freshservice ticket attachments | Total size of a ticket's attachments created through the API "cannot exceed 40 MB" (Freshservice API docs) | Large files need a rule agreed before the run |
Every source platform has a list like this, and it changes. For a ServiceNow example, see ServiceNow export limits. Our guide on how to export tickets from Zendesk shows how deep one source goes on its own.
How long does it take to build a migration tool, and who does it?
There is no neutral public benchmark for how long a help desk importer takes to build, so be wary of any figure quoted without a source. What you can plan is the staffing: every connector needs an engineer who knows the source API, someone to test against real customer data, an onboarding lead to run migrations, and a security review.
| Role | What they own | Is it a one-off? |
|---|---|---|
| Integration engineer, per source platform | Extraction, mapping, load, delta sync, fixes after each run | No: the connector needs an owner for as long as you support that source |
| QA or data analyst | Test loads on messy real data, reconciliation, sample diffs | No: every major change needs a re-test |
| Onboarding or professional services lead | Scoping with the customer, running the migration, cutover, sign-off | No: one per migration |
| Security and legal | Data handling, access scopes, data processing terms for customer records passing through your systems | Partly: the review repeats when scope or regulations change |
| Product manager | Deciding which source to support next, and which edge cases to fix | No: the request queue never empties |
The useful question for your team is not "how many weeks to build?" but "who owns this connector in two years, and what do they stop doing to own it?"
What does an in-house importer cost to keep running?
The build is the visible cost. Upkeep is the larger one, and it never ends: source platforms change their APIs, each new customer brings data shapes you have not seen, sales asks for the next platform, and the onboarding team becomes the support desk for the importer. Budget for these lines before you start.
- Source API changes. You do not control your competitors' APIs. Zendesk, for example, published new limits for offset-based pagination: its docs now cap offset paging at the first 100 pages and 10,000 resources, and recommend cursor pagination. A connector written before a change like that has to be reworked, or it stops short of the full data set.
- Edge cases per customer. Departed agents, merged tickets, very old records, inline images hosted on the old domain, picklists with hundreds of values. Each new customer adds a few.
- The next platform. Win rates against one competitor improve, and sales asks for the next source. Each one is a new connector, not a configuration change.
- Support load. When a migration fails at 2 a.m. before a go-live, your engineers are the only people who can fix it.
- Security scope. Your systems now process customer records from another vendor's platform, including personal and sometimes regulated data, which widens what your security reviews have to cover.
- Opportunity cost. Every sprint spent on importers is a sprint not spent on the product your customers are buying.
Our post on why automated migration scripts fail covers the technical failure modes in more depth.
When does building your own importer make sense?
Building your own migration tooling makes sense when a single source platform dominates your switchers, the data model is simple, and migration is part of what you sell, not a step before it. Partnering makes more sense when switchers arrive from many platforms, history and attachments matter, or your engineering roadmap is already full.
| Build in-house if | Partner with a migration platform if |
|---|---|
| One or two source platforms bring most of your switchers | Switchers come from five, ten or more platforms, including legacy tools |
| Customers move small, flat data sets: a few fields, little history | Customers need years of history, comments, attachments and custom fields |
| Self-serve onboarding is your growth model, and import is part of signup | Mid-market and enterprise deals stall on "how do we migrate?" |
| You can fund a named owner for each connector for years | Your engineers are committed to product work |
| Migration is a feature you want to sell and differentiate on | Migration is a step customers must get through to start using your product |
| You already handle the customer data categories involved | Health, payment or regulated data needs specialist handling and sign-off evidence |
A quick decision checklist
Answer each question for your last 20 closed deals. Three or more "yes" answers usually point to partnering.
- Did switchers come from more than three different source platforms?
- Did any deal include a legacy or on-premise tool you have no connector for?
- Did any customer need more than a year of ticket history moved with comments and attachments?
- Did migration questions appear in security review or procurement?
- Did an engineer get pulled off roadmap work to write or fix a one-off script?
- Did any go-live slip because of the data move?
Build vs partner: how do the costs compare?
Without a neutral benchmark, the honest comparison is structural: who pays for what, when, and for how long. Building turns migration into a fixed engineering cost per connector plus open-ended upkeep. Partnering turns it into a cost per migration, which you can pass to the customer, fund yourself, or split.
| Build in-house | Partner with a migration platform | |
|---|---|---|
| Upfront cost | Engineering time for each connector before the first customer uses it | None before the first deal; scoped per migration |
| Cost of a new source platform | A new connector, owned for its lifetime | The partner's job; MigrateX builds new connectors in 2 to 3 days |
| Ongoing cost | API changes, edge cases, re-tests and support, forever | Included in the partner's service |
| Time to first migration | After the connector is built and tested | As soon as the deal needs it |
| Who owns a failed record | Your engineers | The partner, inside the agreed scope |
| Coverage of the long tail | Only the platforms you chose to build | Whatever the partner supports, and new ones on request |
| Proof for the customer | Whatever reporting you build | A reconciliation report the customer signs |
| Brand | Yours | Yours too, if the partner works white-label |
| Security scope | Your systems process another vendor's customer data | The partner's controls and terms, which you review once |
| Control over the experience | Full | Shared: agree scope, reporting and escalation up front |
The cost that decides most cases is not on either column: the deals that stall or slip while migration is unresolved. Our post on how migration fear stalls enterprise deals covers that side.
Is there a middle path between building and buying?
Yes, and most vendors end up there. Keep a simple self-serve import (CSV or a single API connector) for small accounts and the one competitor most switchers leave, and use a migration partner for everything else: other platforms, enterprise accounts, regulated data and full history. Then let real migration data decide what is worth building.
Three patterns work well:
- Self-serve for small, partner for complex. Your in-product import handles a few thousand flat records. Anything with history, attachments or a cutover window goes to the partner.
- Partner first, build later. Route every switcher through a partner for a few quarters. If one source platform turns out to bring most of your volume, you now know its edge cases and can decide whether to build that one connector.
- Partner for the long tail. You maintain the connector for your biggest competitor; the partner covers legacy tools, smaller platforms and anything new sales brings in.
Whichever pattern you choose, keep ownership of the customer experience, as Lemkin advises: the scoping call, the timeline and the sign-off stay with your team.
How does a migration partnership work for software vendors?
A migration partnership turns data onboarding into a fixed step of SaaS customer onboarding: your sales and onboarding teams can answer "how do we migrate?" on the first call. The partner scopes the move for your prospect and runs a free sample on their real data before they sign. Then it delivers the full migration and hands over a reconciliation report, under your brand or alongside your team.
Here is how it works with MigrateX:
- Scoping during the deal. We count records by type, check attachments, custom fields and history in scope, and give your prospect a free 100-record sample on their own data, then a fixed quote within 24 hours. Your team can show the sample in the evaluation.
- Your brand or ours. We can work white-label behind your team, with your brand on the deliverables, or join the deal as your migration partner. See how we work with partners.
- Who pays. Some vendors fund the migration to win the deal; others pass the cost to the customer. We can set it up either way.
- Every connector live. Our connectors for help desk, ITSM and CRM platforms are all live (see the connector list). If a prospect runs a platform we have not connected yet, we build the connector in 2 to 3 days.
- Experts on every run. ITSM experts review every step, from field mapping to cutover (see how it works). The full run keeps your customer's team working, with a final sync of last changes at cutover, and every production cutover has a rollback plan.
- Proof at sign-off. We count every record, source to target, and hand over a reconciliation report with every exception closed before the customer signs off. Our guide to data migration validation shows what that report checks.
- Security your customers can review. MigrateX is SOC 2 Type II, GDPR compliant and HIPAA compliant, with a DPA and BAA available (see security).
One example of a partner-led project: for a publicly listed energy group in Southeast Asia, we moved 19,970 records from ServiceNow to Freshservice through an IT services partner. Fields were mapped and a demo shown within 12 hours, and the full run took 136 minutes (customer story).
What should you ask a migration partner before you sign?
Ask these as a vendor, not as a buyer. They test whether the partner fits your sales and onboarding motion.
- Can you scope a prospect's migration before they sign, on their real data?
- Can you work under our brand, and who talks to the customer during the run?
- Which source platforms do you support today, and how fast can you add one?
- Who owns a failed record, and how do you report exceptions to us and to the customer?
- What security and compliance terms can you sign with our customers directly?
Want to see it on a live deal? Talk to us about partnering, or send us a prospect's record counts for a free 100-record sample.
Facts checked November 11, 2026 from Zendesk rate limits, Zendesk Ticket Import API, Zendesk pagination docs, Freshservice API docs and SaaStr. No vendor paid for this article. How we research
Build vs buy FAQ
What is a build vs buy analysis?
A build vs buy analysis compares the full cost and risk of creating a capability in-house with buying it from a specialist. For migration tooling, it weighs connector build time, years of upkeep, support load and security scope against a per-migration cost, the partner's platform coverage and the control you keep over the customer experience.
When should a software vendor build its own migration tool?
When one or two source platforms bring most of your switchers, the data is simple, and you can fund a named owner for each connector for years. If switchers come from many platforms, or deals need full history with attachments and a sign-off report, partnering usually costs less than the upkeep you would carry.
How much does it cost to build a data migration tool in-house?
No neutral public benchmark exists, and published figures mostly come from vendors selling importers. Price it yourself: engineer time per connector, testing on real data, security review, then yearly upkeep for API changes, edge cases and support. The upkeep, repeated for every source platform, is usually larger than the first build.
Should we build an importer only for our biggest competitor?
It can be the right call if that competitor brings most of your switchers and its data model is well documented. Plan for the rest anyway: the deals from other platforms still need an answer, and a migration partner can cover them while you learn which second connector, if any, is worth building.
Who should pay for a customer's migration, the vendor or the customer?
Both models work. Funding the migration removes a reason to delay signing; SaaStr founder Jason Lemkin argues it is worth it even at $20K or more per deal because of the customer's lifetime value. Passing the cost on works when the quote is clear and scoped early, so it never surprises the buyer late in the deal.
Can a migration partner work under our brand?
Yes, if the partner offers white-label delivery. MigrateX can work behind your team with your brand on the deliverables, or join the deal as your named migration partner. Either way, agree up front who talks to the customer during the run and how exceptions are reported.
What happens when a source platform changes its API?
Someone has to update the connector, re-test it on real data and re-run anything that failed. With an in-house importer that is your team, for every source you support. Zendesk, for example, limits offset pagination to the first 10,000 resources, so a connector that pages that way has to move to cursor pagination.
Changelog
- 11 Nov 2026: First published.
Written by
Ayush BothraFounding Member, MigrateX
Ayush writes about ITSM platforms, pricing and what it takes to move a service desk to a new platform without losing its history.


