How investigation cost breaks down
Based on industry data and observations across B2B SaaS support teams, the cost decomposition per technical ticket looks like this:
| Phase | Time | Cost at $50/hr | % of total |
|---|---|---|---|
| Triage and routing | 1-2 min | $0.83-1.67 | 3-5% |
| Investigation | 20-45 min | $16.67-37.50 | 75-85% |
| Response drafting | 3-5 min | $2.50-4.17 | 10-12% |
| Follow-up | 2-3 min | $1.67-2.50 | 5-8% |
The $50/hour fully-loaded cost is conservative for a support engineer at a B2B SaaS company - when you factor in salary, benefits, tooling, and management overhead, the real number is often $60-80/hour. At $70/hour, investigation costs $23-52 per ticket.
Why existing tools don't touch this cost
The support tooling ecosystem is built around everything except investigation:
- •Support platforms (Zendesk, Intercom, Pylon) optimize triage and routing - the $1 phase
- •Doc chatbots deflect FAQ-answerable tickets - roughly 20% of B2B volume
- •AI copilots speed up response drafting - the $3 phase
- •Investigation - the $17-40 phase - has no dedicated tooling in most stacks
This is why companies report modest ROI from support AI despite investing heavily in it. They've optimized the $4 in triage + response and left the $30 in investigation untouched.
The annual investigation tax
For a B2B support team, the math scales quickly:
| Ticket volume | Investigation cost/ticket | Monthly cost | Annual cost |
|---|---|---|---|
| 100 tickets/month | $25 avg | $2,500 | $30,000 |
| 200 tickets/month | $25 avg | $5,000 | $60,000 |
| 500 tickets/month | $25 avg | $12,500 | $150,000 |
| 1,000 tickets/month | $25 avg | $25,000 | $300,000 |
These numbers represent the investigation cost alone - not total support cost. And they don't include the second-order costs: missed SLAs, delayed escalation resolution, engineer burnout, and customer churn from slow response times.
What happens when you automate investigation
When investigation is automated - running the same ClickHouse queries, Linear searches, Stripe checks, and GitHub lookups in parallel instead of sequentially - the time drops from 20-45 minutes to under 2 minutes. The cost per ticket drops proportionally.
"Altor diagnosed in 2 minutes what used to take our engineers 45 minutes of copying data between tabs."
Calculate your investigation cost
Take your average fully-loaded engineer cost per hour, multiply by your average investigation time per ticket, multiply by your monthly ticket volume. That's your investigation tax. If 80% of those investigations follow repeatable patterns, that's the amount you can automate away.