An ERP that doesn't talk to anything is just a very expensive spreadsheet
The whole point of an ERP is to be the single source of truth. But that only works if the systems generating your data — payments, logistics, telephony, e-commerce — actually feed into it. Without integrations, someone ends up re-typing the same order into three systems, and the "single source of truth" quietly becomes three sources of truth that disagree with each other.
The mistake most teams make isn't skipping integrations. It's trying to build all of them at once, in the first phase, before anyone has confirmed the core system actually works.
Rank integrations by cost of manual work, not by how easy they sound
Every integration has a manual-work alternative: someone re-keys the data by hand. Rank your integration wishlist by how expensive that manual alternative actually is.
- Payments — reconciling payment gateway transactions against invoices by hand is slow and error-prone at any real volume. This is usually integration #1.
- Logistics and shipping — if a human is copying tracking numbers between systems, that's hours a week for no analytical benefit.
- Telephony — call logs matter for sales and support teams, but the business doesn't grind to a halt without this one on day one.
- Marketing tools — nice to have a two-way sync, but usually survivable with a manual export for months.
Confirm the core system first
Every integration you add before go-live is one more thing that can break, and one more reason a delay in the core build turns into a delay in three other places. Get the ERP itself stable and adopted, then integrate outward in priority order. It's tempting to want everything connected from day one — resist it.
Ask what happens when the integration goes down
Integrations fail. APIs change, rate limits get hit, a webhook silently stops firing. Before you connect anything, agree on what the fallback is: does data queue and retry, does someone get an alert, is there a manual re-sync process? An integration with no failure plan isn't finished, it's just untested.
Where to start
If you're not sure where to begin, start with whichever manual process currently eats the most staff time each week, not whichever integration is technically the simplest to build. The easy ones can wait.