What it means
The word tax fits because the cost recurs and is hard to avoid once a stack is in place. It has four parts: the connective software bought to move records between systems; the hours spent keeping those connections working; the reconciliation of several versions of one customer; and, more recently, AI add-ons that each see only the system they were bought for.
None of it is unique to small companies, but it weighs more on them. A large organisation has an integration team; a small one has whoever set up the connectors and still remembers how.
Why it matters when you choose software
Software is usually compared on licence price, the line every vendor negotiates and the one most visible on an invoice. The larger costs appear on no invoice at all. They show up as an operations hire, as a month-end close that runs late, or as a report nobody trusts because two systems disagree.
A tool that looks cheap on its own can be expensive in the stack it joins.
Questions to ask a vendor
- Which of our existing systems will this need to exchange records with, and how?
- Who maintains those connections when either side changes its fields?
- Which system holds the true record of a customer, and what happens when two disagree?
- Can its AI see data held in our other systems, or only its own?
- If we leave, how do we get our data out, and in what shape?
How Cyril approaches it
Cyril’s answer is structural rather than another connector: Sales, Projects, Support, Finance, Marketing, Docs, Mail, Drive and Calendar share one schema, so there is nothing to sync between them. This site publishes an open, assumption-by-assumption model of the cost, in which the integration tax comes to about 70% of the total against roughly 30% for licences. The calculator lets you change every assumption and see your own figure. It is a model with stated assumptions, not a measurement of any customer.