Three buckets, not one price

Setup is the one-time work to design, build, and test the system. Usage is what you pay while it runs, such as model calls, telephony minutes, or per-event platform fees. Support is the ongoing work to fix, tune, and change the system after launch.

Mixing these into one figure makes comparison hard and creates surprises later. Example: two quotes both say monthly. One includes design work spread over the first quarter. The other charges a build fee and a smaller running fee. Split the numbers and the two quotes stop looking equal.

  • Separate one-time setup from recurring costs
  • List every recurring fee and who charges it
  • Ask what support includes

What drives setup cost

Setup depends on how many processes you cover, how messy the current data is, and how many systems must be reached. A single, well-defined workflow with clean input costs far less than a broad project across several teams.

Access is a hidden driver. Some systems allow a connection and some do not, and a system without a usable interface means manual steps or extra work. Confirm access during scoping, before the build starts, because it changes the shape of the project and not just the price.

  • Count the processes in scope
  • Check how clean the input data is
  • Confirm access to each system
  • Agree what a finished version includes

What drives ongoing usage cost

Usage follows volume. More calls, messages, or records mean more usage. Some fees are per conversation, some per minute, and some per seat on a platform. These are usually paid monthly and can change when your activity changes.

Example: a retailer's order questions spike before a holiday. If the agent is charged per conversation, that month costs more than a quiet one. That is not a fault in the system; it is how usage works. Budget for a variable line, not a flat one.

  • Estimate a normal month and a busy month
  • Find out what is billed per unit
  • Identify fees you pay directly
  • Review usage after the first full month

What support really covers

Support is the work that keeps a system matching reality. It includes fixing a broken connection, adjusting rules, and responding when a platform changes. It is not the same as setup, and it does not stop when the build is delivered.

Handling personal data adds obligations that affect support. Under Canadian privacy law, organisations must follow fair information principles such as limiting collection and keeping data accurate. That can shape what a system logs and how long it keeps it, which is support work rather than a build detail.

  • Write down what support includes
  • Decide who responds when a run fails
  • Review data handling and retention
  • Plan for platform and rule changes

Plan for change, not just launch

A system that works on day one is not finished. Hours change, prices change, staff change, and a connected tool updates. Treat launch as the start of a maintenance period rather than the end of the project.

Example: a booking agent works well until the clinic changes its hours for summer. If nobody owns that update, the agent gives wrong answers. A named owner and a simple change path cost little and prevent the failure. This is a support question, not a setup one.

  • Name who owns updates
  • Set how changes get requested
  • Review the system on a schedule
  • Keep a record of what changed

A simple way to compare options

Put each option into the three buckets on one page. Add a setup column, a monthly usage column for a normal and a busy period, and a support column. Then compare like with like, and note anything a vendor leaves blank.

Example: one option has low setup and higher usage, another has higher setup and lower usage. Which is better depends on your volume and how long you expect to run it. Without splitting the buckets, that choice is a guess.

  • Put every option in the same three buckets
  • Compare normal and busy usage
  • Mark unclear or missing numbers
  • Decide what happens if you stop

Take this with you

Cost has three parts: setup once, usage while it runs, and support after launch. Usage tracks volume, so plan for busy periods. Support keeps the system matching reality. Compare options bucket by bucket, and treat any single all-in figure with care.

Source / Further reading