Annoyance is not a signal
All software is annoying to somebody. The question is whether the mismatch between your process and the product is costing you money, or merely costing patience.
The signals below are the ones that translate into cost.
Five signals that you are ready
- Spreadsheets exist to work around the system. Every one is a feature the software does not provide and someone is providing manually.
- You are paying for several products and using a fraction of each. Licence cost rising faster than the value you extract.
- The workarounds have workarounds. A macro to fix a problem, a checklist for what the macro misses.
- You cannot answer basic questions without exporting and joining data by hand.
- Your process is your advantage and the software forces you to do it the standard way.
Two or more of these, consistently, is a genuine case. One might be solvable with configuration.
Three that look like signals and are not
- “The interface is dated.” Cosmetic, and rarely worth a five-figure project.
- “It does not do X.” Check whether X is available and unconfigured — it frequently is.
- “Our supplier is unresponsive.” A relationship problem, and building your own does not fix it; it makes you the supplier.
Before commissioning anything, spend an hour with your existing product's documentation and support. A surprising share of “it cannot do that” turns out to be “nobody switched it on”.
The hybrid that usually wins
You rarely need to replace everything. Keep the packaged software for the commodity functions — accounting, email, storage — and build the thin layer that encodes what makes you different.
That gets you differentiation where it matters and someone else's maintenance burden everywhere it does not.
What it commits you to
Custom software is yours forever: maintenance at 15–25% of build annually, an owner, and dependency updates. If that is not affordable, the honest answer is that custom is not affordable, and something smaller is the right project.
Start with the piece that hurts most
Not the whole system. The one process where the mismatch costs the most, built properly, alongside what you already run.
If it works, you have a foundation and evidence. If it does not, you have lost weeks rather than a year.