Template
One-page case study
Every design partner's results get published in this format, on one page. It compares a measured 30-day baseline with the same length of time after the switch.
- Every number names its source: a provider invoice, the ROUTEXOR usage export or the partner's own logs.
- Estimates are labeled as estimates. Nothing is extrapolated to a year.
- The partner approves the final page before it is published, and can stay anonymous.
Who
Company or description
Named, or anonymous ("a 12-person developer-tools company"). Partner chooses.
Agents and workloads
What the agents do, how often they run, how many run in parallel.
Providers and models
Which providers and models, before and after.
Before ROUTEXOR (baseline, measured over 30 days)
Monthly model spend
From provider invoices.
Routing or gateway fees
Fees, markups or credit top-up charges paid, from invoices.
Spend controls
How caps worked, and any overruns in the period.
Provider failures
Outages or rejected keys that stopped an agent, with dates.
Setup
Time to first routed call
Minutes from signup.
What was configured
Keys, task profiles, fallback chains, caps.
Results (same 30-day length, after the switch)
Requests and tokens routed
From the ROUTEXOR usage export.
Model spend
From provider invoices, not estimates.
ROUTEXOR fee
From the ROUTEXOR invoice, and as a percent of model spend.
Fee change
Before minus after, in dollars per month.
Caps that blocked a call
Count and dollars avoided, if any.
Fallbacks
Calls that moved to another model after a provider failure.
Median call time
Successful calls, same models before and after where possible.
In their words, and the limits
What changed
Two or three sentences from the partner.
What did not work
Bugs, missing features, anything that got in the way.
How the numbers were checked
Sources for every figure, and the partner's sign-off date.