Most shortlists of BizTalk Server alternatives are built from feature grids, and feature grids hide the differences that decide a migration. Three things matter more than connector counts. First, how many of your schemas, maps and pipelines carry forward without modification. Second, whether the platform preserves BizTalk routing and orchestration behaviour or requires you to re-implement it. Third, whether your operators can locate and recover a failed message without calling a developer.
Every integration platform on the market can transform and move a file. Vendor demos show us almost nothing about how our estates will move.
BizTalk estates are different. Artifacts built 20 years ago contain flat file schemas built for a single partner, maps with embedded C#, subscriptions that have not been reviewed since they were created, and orchestrations that wait three days for a reply. The alternative you choose determines what you move forward and what you rebuild. That is why the artifacts and their impact drive the direction and scope of this series.
01 / The clock
The deadline is fixed. Your shortlist is not.
There will be no new BizTalk release. Microsoft has confirmed that BizTalk Server 2020 is the final version, with mainstream support ending April 11, 2028 and extended support ending April 9, 2030. Microsoft has also announced that it expects sales of BizTalk Server 2020 and Host Integration Server 2020 to end on March 31, 2027. Teams still on BizTalk 2016 face an earlier date, with extended support ending January 11, 2027.
The dates do not make any one platform better. They shorten the window in which you can migrate in orderly waves, which means each criterion below should be judged partly on how much rework it adds.
Time pressure rewards fidelity, not breadth
When the window is short, the best alternative is the one that moves the most interfaces with the least rework and retraining, not the one with the longest connector list.
02 / Artifacts
Your schemas and maps are the migration budget
This is where you find out how much of your estate is a migration and how much is a rewrite.
Schemas
XSD schemas are the most portable asset you own. A credible alternative should load them unchanged, including the flat file annotations used to parse positional and delimited files. Ask the vendor to load three of your hardest flat file schemas, not a sample.
Maps
BizTalk maps compile to XSLT, but many contain logic that does not move cleanly: scripting functoids with inline C#, database lookup functoids and custom functoid assemblies.
Pipelines and rules
Custom pipeline components are written against BizTalk interfaces and will not run anywhere else as is. Business Rules Engine policies need a new home.
Microsoft’s own guidance for migrating to Azure Logic Apps notes that it is not a lift-and-shift migration: keep the required business practices, and modify the implementation. The amount of rework you face varies greatly between systems, so ask each vendor to quantify it.
| BizTalk artifact | What to ask | Red flag |
|---|---|---|
| XSD schemas | Will our flat file schemas load unchanged? | “We’ll regenerate them from samples” |
| Maps (.btm) | Native XSLT? What happens to inline C#? | Visual remapping with no XSLT import |
| Orchestrations | Which patterns map to what? | “Rebuild them as workflows” with no pattern mapping |
| Pipelines | Is there an extension point for custom components? | All custom logic moves to a separate service |
| BRE policies | Where do rules live after migration? | No clear answer |
| Bindings | Can port and endpoint settings be imported? | Manual re-entry of every endpoint |
Ask these questions of every vendor in the same order, and record the answer verbatim. A vague answer in this table usually becomes a line item in the statement of work.
Scripting functoids set the cost of your maps
Count the maps that contain inline C# before any vendor call. That count, not the total number of maps, predicts how much map review effort you will pay for.
03 / Routing
BizTalk routes by subscription, and most estates forgot that
BizTalk is a publish/subscribe engine. Receive ports publish messages to the MessageBox, and send ports and orchestrations pick them up through filters on promoted properties. One inbound order can reach an ERP, a warehouse system and an archive simply because three subscriptions match it.
Many estates depend on this without anyone remembering. A point-to-point workflow tool can reproduce it, but only if every fan-out is rebuilt explicitly and maintained by hand.
Does the platform natively support content-based subscriptions?
If not, each fan-out becomes a branch in the publishing flow.
What happens if we add a fourth subscriber?
This should be a configuration change, not a change to the publishing flow.
04 / Orchestration
The hard patterns live in orchestrations
Orchestrations hold the behaviour that is hardest to move. Ask each vendor to show these patterns running, not described on a slide.
Show these running
Live, on a real interface
- Correlation sets that match a reply to the right running instance, such as an order confirmation arriving hours later
- Sequential and parallel convoys that collect related messages before processing
- Long-running transactions and compensation that undo earlier steps when a later step fails
- Persistence and dehydration, so a process waiting days for a reply holds no compute
- Ordered delivery combined with content-based routing. Many platforms support each separately. Fewer support both in the same flow. Our comparison of integration platforms for Azure found that MuleSoft makes you choose one or the other on a single destination.
Correlation is native
The platform matches replies to instances by message content, with state persisted automatically.
Correlation is a table you build
You design, index and maintain a correlation store for the life of every integration that uses it.
Correlation is polling
Instances check for replies on a timer, adding latency and cost that grow with volume.
Correlation is manual
Operators match late replies by hand when the automation misses them.
05 / Operations and B2B
The 2 a.m. test matters more than the demo
This criterion is weighed least by buyers and regretted most. In BizTalk, a failed message leads to a suspended instance. Microsoft’s documentation describes resumable and non-resumable instances, and resuming a messaging instance sends the message on to its send port. Your support team is built around this behaviour.
What operators need
Score these
- To find a failed message by business key, such as a PO number, not only by run ID
- To resume one message or a batch without a developer or a redeploy
- Retry policies they can set per endpoint
- Tracking that is on by default, with its cost known in advance
What a demo usually shows
Do not score these
- A happy-path run
- A dashboard with green tiles
- An error message on a test flow
Microsoft states that BizTalk message body tracking is not a substitute for legally binding auditing. If your compliance team treats BizTalk tracking as the audit trail, settle that before you score replacements.
“Supports X12” is not an answer. Check the following:
06 / Platform fit
Where it runs, how it changes, what it meters
| Question | Why it matters | What to ask |
|---|---|---|
| Whose tenant runs the runtime? | Data residency, compliance scope, network design | Our Azure tenant, the vendor’s cloud, or shared multi-tenant? |
| Where does any AI run? | Message data may leave your boundary | Does map conversion or assistance send our data outside our tenant? |
| How does it reach on-premises systems? | SQL Server, file shares, SAP and MQ rarely move first | Is any inbound firewall change needed? |
| How do changes go live? | Speed of fixing a partner issue | Does a change take effect on save, or need a build and release? |
| What does it meter? | Cost follows the meter your estate drives hardest | Per call, per connection, per flow, per vCore, or reserved capacity? |
| How would we leave? | Avoiding the next end-of-life trap | Do schemas stay XSD and maps stay XSLT? |
The last row is the one most often skipped. A platform that stores your logic in a proprietary format turns today’s migration into the next one.
Metering differs more than pricing pages suggest. Azure Logic Apps Standard, for example, bills reserved hosting capacity whether workflows run or not. Managed connector calls are metered per call, and stateful workflows add storage transactions. Connection-licensed platforms scale with endpoints instead. The scorecard article in this series shows how to model cost on each meter at your volume.
A lighter platform is sometimes right
If your estate is mostly SaaS-to-SaaS API calls, with few orchestrations, no EDI and no ordered delivery, a general iPaaS or Logic Apps alone may serve you well. The criteria above matter most where those patterns exist.
07 / Before vendor calls
Two counts to take before the first demo
Interfaces that use orchestration patterns
Correlation, convoys, long-running scopes and compensation. This count tells you how much of your estate depends on criterion 04, and how hard to push vendors on it.
Number of trading partners
A platform that costs per partner or per agreement can be cheap at five partners and expensive at fifty. It also tracks your partner coordination work during cutover.
Those two values turn the preceding criteria into weighted priorities, not a checklist. In the next piece in this series, How to Build a BizTalk Alternative Scorecard From Your Own Workloads, we show how to turn these workloads into weights.
Want this mapped against your BizTalk estate?
We have run integration in production for twenty years, and Cerebrum City is a Microsoft V-TSP recognized for BizTalk Server. Our courtesy assessment inventories your ports, orchestrations, maps and adapters, flags the interfaces most likely to break, and returns a staged Azure roadmap. No obligation.
Call (678) 203-0800.
Frequently asked questions
What is the most important thing to look for in a BizTalk alternative?
Artifact fidelity: how many of your schemas, maps, pipelines and bindings carry forward without a rewrite. It sets the size of your migration more than any other single factor, because every artifact that cannot move becomes design, build and test work. Start with flat file schemas and maps that contain scripting functoids, since they are the most common sources of hidden effort.
Can a BizTalk alternative reuse my existing maps?
Often in part. BizTalk maps compile to XSLT, so platforms that run XSLT natively can reuse much of the transformation logic. Scripting functoids with inline C#, database lookup functoids and custom functoid assemblies usually need conversion and review. Ask every vendor to convert three of your real maps and show the diff, rather than accepting a general percentage.
Is Azure Logic Apps a BizTalk alternative?
It is Microsoft’s recommended destination. Microsoft’s guidance suggests using native Logic Apps Standard capabilities first, then adding other Azure Integration Services as requirements justify. Whether it fits depends on your orchestration patterns, B2B depth and operations needs. Estates heavy in correlation, EDI and suspend-and-resume operations should test those areas specifically.
Does BizTalk stop working when support ends?
No. BizTalk keeps running after support ends, but security updates stop when extended support finishes on April 9, 2030 for BizTalk Server 2020. Microsoft also expects sales of BizTalk Server 2020 to end on March 31, 2027. Running unsupported adds security and compliance risk each year, which is why most teams plan to finish migrating before that date.
How should AI-assisted conversion affect my evaluation?
Score the output, not the promise. AI can shorten map and component conversion, but Microsoft’s own guidance for its Migration Agent says generated artifacts do not replace architectural review or testing. Ask where the AI runs and whether message data leaves your tenant. Then measure how much reviewer time the converted artifacts actually need.
How many platforms belong on a shortlist?
Two or three. Each serious candidate needs a proof of concept on a real interface, and running more than three properly costs more time than most migration windows allow. Use knockout requirements, such as tenancy, healthcare protocols or on-premises reach, to cut the field before scoring, rather than scoring every platform on the market.
What should operators check during an evaluation?
Ask them to recover a failed message on each platform themselves. They should locate it by business key, inspect the error and resume it without a developer. Time each step. Operators also spot gaps in retry settings, alerting and tracking that architects miss in demos. Their verdict on day-to-day recovery should carry real weight in the final score.
WMC · Cerebrum City
Cerebrum City is a Microsoft V‑TSP–recognized BizTalk specialist in Atlanta, Georgia, with two decades running BizTalk Server in enterprise production for clients including Humana, Deloitte, JetBlue, and U‑Haul.
Want this mapped against your BizTalk estate?
We inventory your ports, orchestrations, maps, and adapters, flag the interfaces most likely to break, and hand back a staged Azure roadmap. No obligation.

