Eight months of no-code glue, then one schema change cost me a week

Before the rebuild question, what does your subscriber agreement say about data accuracy? Most platform terms disclaim accuracy, and if yours doesn't, four days of wrong fields is a bigger exposure than six days of your time. Have a software attorney look at your actual terms, since what's enforceable in that kind of disclaimer varies by state. That's separate from whether you rebuild, but I'd fix it this week either way.

The part that would keep me up is that nobody complained. You've got 26 people paying for something and apparently a chunk of them aren't looking at it. That's a churn problem sitting there waiting.

@marlow beat me to it. harrow, the schema break is a symptom. If half your seats didn't notice garbage in the owner name field, your renewal risk at month 12 is higher than your infrastructure risk. I'd spend the eight weeks finding out why they're not using it before I'd spend it on validation code.

You don't need a rebuild to catch this. You need a check that fails loudly when a column disappears or when 90% of a field suddenly matches an address pattern. That's a day of work, not eight weeks.

@verity is right and that's the cheap answer, though it only catches the failures you thought of. Every unit I own has taught me a failure I didn't predict. Same principle here. Validation on known shapes plus a weekly eyeball on a sample is probably where I'd sit at 26 seats.

$14,000 at 26 seats. What's your monthly revenue? If it's $5,000 you're spending three months of gross on plumbing that doesn't add a single feature a buyer would pay for. I hit my ceiling by doing exactly that kind of thing, building the version of the business I wanted to have instead of the one I had.

The contractor number is the one I'd interrogate. $14,000 for an ingestion rebuild that only covers the counties you have today, and then each new county is a change order. Ask for the quote to include what a new source costs to add, in writing. That figure usually reframes the whole decision.

Naive question that I think matters here. If the no-code chain has run since March and this is the first break, is one break in eight months actually bad? Or is the standard higher because people are paying?