Do not rebuild a familiar workflow to avoid a subscription
A team dislikes its software bill and proposes an internal replacement. Before estimating screens, list everything the subscription currently supplies: account management, permissions, updates, operational support, integrations, and the routine work nobody noticed until it stopped. A smaller feature list does not automatically mean a cheaper responsibility.
Choose SaaS when a standard workflow fits and its limitations are acceptable. Build when a distinctive operational requirement cannot be handled reliably through configuration or a narrow integration. There is no universal seat count at which the answer flips. The relevant comparison is the complete workflow and the capability needed to keep it running.
Start with the process, not the product list
Observe how the work is actually completed. Identify where people wait, copy data, request exceptions, or move into spreadsheets. Then ask whether those detours are caused by the software or by a process nobody has agreed to change.
A custom tool can encode a clear operational advantage. It can also preserve a confusing process in a more expensive form. Automating an undocumented exception does not make the underlying policy coherent.
Before rejecting a SaaS product, test its relevant configuration and integration options against representative work. A missing field is different from an incompatible permission model. The former may be easy to address; the latter can undermine the entire design.
Count ownership on both sides
A subscription transfers some implementation and infrastructure responsibilities to a provider, not all accountability. Microsoft's shared-responsibility guidance keeps customer data, identities, and access management within the customer's responsibilities across cloud models. Exact contractual responsibilities still depend on the selected service and agreement.
For SaaS, inspect administration, access removal, exports, integration limits, and the consequences of a plan change. For a custom tool, assign responsibility for security updates, backups, recovery, incident handling, and future development.
The comparison should include staff time spent maintaining workarounds as well as software expenditure. Avoid invented break-even calculations. Use actual quotes, existing invoices, and observed operating work, then state the assumptions that could change the result.
Build only the part that is genuinely different
The useful answer may be a narrow custom interface around an existing system. Keep commodity capabilities where they already work and create the missing workflow without duplicating every feature of the underlying product.
This introduces integration responsibilities: decide which system owns each record, how errors are reconciled, and what happens when an external interface changes. A thin layer still needs a support plan. It should not become an invisible second database of inconsistent business facts.
Clodron's application and SaaS engineering includes building digital products, not selling a universal replacement platform. The decision starts with the work people need to complete, much like the alternatives in when a chatbot is the wrong answer. Sometimes a better form or an existing workflow is enough.
Compare a working slice this week
Choose a recurring task with a clear owner and a visible outcome. Run it through a configured SaaS candidate and a small prototype of the proposed custom path. Use representative, non-sensitive data.
Record completion obstacles, permission gaps, manual handoffs, and maintenance assumptions. Include the person who will administer the solution after launch, not only the person sponsoring its purchase.
Buy when the standard workflow works and the remaining constraints are acceptable. Build when a demonstrated operational difference justifies sustained ownership. Combine them when a small custom layer can close the gap without duplicating the reliable parts. If no one will own the custom system after delivery, that is a reason to stop the build decision.
