If you're big enough to have millions of dollars to spend on a big ERP (remember: the big names in the ERP industry charge you a % of your business...so if it cost millions, you're making hundreds of millions...), you probably have requirements that aren't exactly trivial to build in house.
If you're, let say, a large international retailer with brick and mortar stores, several factory plants and warehouses, etc... writing the software to handle all the international regulations, the warehouse transfers, handling prepacked product manufactured by third party, etc, will be tens of millions of lines of code. Not exactly something you wipe out in a year.
Of course, if you're just an e-retailer that ships stuff internationally and skip the few countries that make it hard, that's a LOT easier, and a couple of average in-house devs and a good logistic analyst and you're good to go. Bonus point if you outsource your warehousing.
When these big ERP projects fail, its usually companies who have exotic, meaningless processes and refuse to change it, so they have to customize the ERP to hell and beyond. Its normal to have to customize it a bit: everyone is a little different. But there's a point where its the company's fault. I worked somewhere where the concept of SKU. which was used as a unique identifier everywhere (normal, common way to do things) didn't have a 1:1 correlation with a product (ie: they would reuse SKUs for completely different products, and had interns "guess" what it mapped to depending on context in spreadsheets). Thats never gonna work, and things go downhill from there.