How to map every integration your business actually needs before replacing your software stack

• An integration map is a written list of every place data enters and leaves your current platform, including manual steps.
• Hidden connections usually live in Zapier accounts, saved spreadsheet exports, email inboxes and one person's browser bookmarks.
• Data silos after a migration happen because a one way sync was rebuilt as a two way sync, or not rebuilt at all.
• Audit by following the data, not the software. Start at a customer record and trace where it travels.
• A custom system removes integrations by removing the systems, but it cannot remove the ones you don't control, like your bank or your carrier.
• You can build the whole map in a spreadsheet with eight columns and a week of asking questions.

Every migration horror story I've heard starts the same way. The new platform got picked, the demo looked great, the data got moved. Then three weeks in, someone notices the shipping notifications stopped going out. Nobody knew the old system was sending them. It had been running since 2019 through a connector an employee set up and never told anyone about.

Software selection feels like the big decision. Vendors encourage that feeling, because the demo is where they win. So teams spend four months comparing platforms and four days thinking about what plugs into them.

That's backwards. The platform is replaceable. The connections are what make your business run day to day, and they're the part nobody can see. A CRM by itself does nothing. A CRM that pulls web form leads, pushes deals to accounting, triggers a welcome email and updates a warehouse spreadsheet is doing real work. Replace the CRM without replacing those four jobs and you've bought an expensive address book.

The other reason this step gets skipped is that it's unglamorous. Nobody gets excited about interviewing the AP clerk about her Tuesday export routine. But she is holding a piece of your operation together with a macro she wrote in 2021, and when you switch platforms, that macro breaks.

Start with the obvious ones. Your accounting package, your email marketing tool, your e-signature service, your phone system. Those show up in a vendor's integrations tab and everyone remembers them.

Then go hunting for the rest. In my experience, the documented integrations are maybe half of what's actually running. The other half hide in four places.

• Automation tools like Zapier, Make or Power Automate, often on a personal account belonging to someone who left.
• Scheduled exports and imports, where a CSV file gets dropped somewhere on a schedule and something else picks it up.
• Email based flows, where the platform emails a report or a notification and a human or a rule acts on it.
• Manual rekeying, which is an integration performed by a person. It counts. It's usually the most expensive one you have.

Pull your credit card statements and look for small recurring software charges. Check the admin panel of your main platform for a list of connected apps and API keys. An API key is a password that lets one program talk to another, and an old unused key is a sign that something was connected once and maybe still is. Ask each department head a simple question. What would stop working tomorrow if this system went dark?

That last question surfaces more than any audit tool. People don't think of their routine as an integration. They think of it as their job.

A silo is just data that lives in one place and never gets to the other places that need it. They form fast after a migration, and they form quietly.

The most common cause is a direction change. Your old setup pushed customer records from the CRM into accounting, one way. The new connector does a two way sync, so accounting now overwrites addresses in the CRM. Sales updates an address, the sync reverts it overnight, and sales stops trusting the CRM. Within a month they're keeping addresses in a private spreadsheet. That's a silo, and it was created by a setting.

The second cause is field mismatch. Your old system had a custom field called Job Type with eleven options. The new one has a Category field with a free text box. The data moves, technically. But now nobody can filter by job type, so the operations report dies, so someone rebuilds it by hand in Excel. Another silo.

Third is timing. A nightly sync and a real time sync behave very differently on a busy Friday afternoon. Teams that were used to instant updates start double checking, then start working around the system entirely.

Follow the data, not the software. Software inventories miss things because they only show what has a login. Data traces catch the human steps too.

1. Pick your three most important records. For most businesses that's a customer, a job or order, and an invoice.
2. Trace the customer from first contact to closed file. Write down every screen it appears on and every person who touches it.
3. At each stop, ask how the data got there. Typed by hand, pulled by an automation, imported from a file, or copied from an email.
4. Note who owns that step and what breaks if it stops. Name an actual person, not a department.
5. Do the same for the job and the invoice. You'll find overlaps, which is good. Overlaps are where the real integrations live.
6. Ask each team what reports they run weekly and where the numbers come from. Reports expose silos better than anything else.
7. Rank every connection as essential, useful or dead. You will find dead ones. Kill them before you migrate, not after.

Give this a week of calendar time and a few hours of real work. It's not a technical exercise. Anyone who understands the business can run it with a notebook.

You don't need a tool for this. Open a spreadsheet and make eight columns. Name them Source system, Destination system, What data moves, Direction, Frequency, How it runs, Owner, and What breaks without it. Add a ninth column called Keep, rebuild or retire.

Fill one row per connection. Manual steps get a row too, with "human" in the How it runs column. When you're done you'll have somewhere between fifteen and sixty rows, and that sheet becomes the actual scope document for your migration. Hand it to any developer or vendor and they can price the work honestly, because they can see the whole surface.

The Keep, rebuild or retire column does the most work. It forces a decision on every line instead of letting things drift into the new system by accident.

Here's the part most people miss. Half your integrations exist only because you bought separate tools that each hold a piece of the same record. Consolidate the tools and the integration disappears. There's nothing to sync when the data lives in one database.

The arithmetic is worth doing on paper. Say you run 45 seats across a project platform at 40 dollars per user per month, a CRM at 75, and a reporting add on at 500 flat. That's about 6,700 a month, or roughly 80,000 a year, before the automation subscriptions and before the two connectors you pay a consultant to maintain. Those numbers rise at renewal. They never fall.

A system your business owns outright has a build cost and then a hosting and maintenance cost, and the seat count stops mattering. Add your twentieth office user and the bill doesn't move. When you write the annual subscription total on one line and the ownership total on the other, the crossover point is usually closer than people expect. Run it with your own real numbers, including the hours your team spends rekeying.

A custom build removes internal integrations. It doesn't remove external ones, and that distinction matters. You still have to talk to your bank, your payment processor, your shipping carriers, your tax filing service and your payroll provider. Those connections stay, they still need maintenance, and their APIs still change without asking you.

Some integrations are also genuinely better left alone. QuickBooks or Xero is a fine place for the general ledger. Rebuilding double entry accounting is a bad use of money. The right move is to keep accounting and feed it clean data, not to absorb it.

There's also the migration cost itself. Twelve years of file attachments in an old platform come out through a scripted export, not a copy and paste. Historical records with custom fields need mapping decisions, and some of those decisions mean accepting a small amount of data loss on fields nobody has used since 2018. Anyone who tells you a migration is clean hasn't done many.

And if your value depends on sharing live views with clients who already work inside the same platform you're leaving, that's a real switching cost. Solvable with a client portal, but not free. Put it on the map like everything else.

How long does an integration audit take?

About a week of elapsed time for a mid sized company, and maybe six to ten hours of actual effort. Most of that is conversations, not analysis. The bottleneck is getting time with the people who know the workarounds.

What if we don't know what our automations do?

Turn one off in a test environment and see who complains. Seriously. If you can't test safely, check the automation tool's run history, which shows what fired and when. A connector that hasn't run in eight months is almost certainly dead.

Should we fix integrations before migrating or after?

Retire the dead ones before. Rebuild the essential ones as part of the migration itself, not as a phase two. Phase two has a way of never arriving, and that's how teams end up running two systems for a year.

Do we need a developer to make the map?

No. An operations manager with a spreadsheet and good questions will produce a better map than a technical audit tool, because the tool can't see the manual steps. Bring in a developer once the map exists and you need the rebuild priced.

Open the spreadsheet. Make the nine columns. Book thirty minutes with each department head and ask what breaks if the system goes dark tomorrow. You'll have a usable draft by Friday, and you'll know more about how your business actually runs than any vendor demo will tell you.

If you'd like a second set of eyes on the finished map, or you want the ownership arithmetic run against your real subscription costs, book a free strategy session and we'll walk through it with you.

Read this next: our breakdown of how to decide between renewing a SaaS platform and building a system you own outright.

Related articles

• What it really costs to migrate off a project management platform
• Client portals without the per user fee
• Why per seat pricing punishes growing companies