Your Observability Bill Is Bigger Than Your Hosting Bill. Here's the Exit.

Your observability bill quietly crossed your hosting bill. You're paying $2,000+/year for a platform whose dashboards three people look at. Meanwhile, your actual incidents — the ones that cost you customers — were a disk that filled up, a cron that died silently, and a DNS record nobody owned.

Here's the uncomfortable math, and the exit path.

The bill you're defending

Typical small-team stack: $99/mo APM + $150/mo log platform + $50/mo uptime checks + a status page at $25/mo. That's $3,900/year to learn things you mostly already knew: the server was fine, the deploy was the problem, and the database needed an index.

Worse: complexity tax. Every platform adds its own agent, its own query language, its own alert-config UI. Your on-call rotation now needs a training doc for the monitoring, not the systems.

What actually catches your real incidents

Audit your last 10 incidents. For most small teams, the causes collapse into five buckets: disk full, dead background job, certificate expiry, dependency outage, bad deploy. Every one of those is detectable with a 10-line script and a cron entry:

# disk
df -H | awk 'NR>1 { gsub("%","",$5); if ($5+0 > 85) print "DISK HIGH: "$6 }'
# dead cron evidence: last run timestamp vs now
find /var/run/jobs -name "*.heartbeat" -mmin +60 | xargs -r echo "STALE JOB:"
# cert expiry
echo | openssl s_client -servername "$DOMAIN" -connect "$DOMAIN:443" 2>/dev/null | openssl x509 -noout -enddate

Wire those into an alert channel and you've covered the majority of real incidents for zero dollars a month. The expensive platform is solving problems you don't have yet.

The honest hybrid

Keep the free tier or a single external check for the things scripts can't do (multi-region latency, SSL from the customer's vantage). Drop the $99/mo dashboard until you have a problem that requires distributed tracing. You'll know when you get there — you'll be hiring someone whose job is dashboards.

---

I run a one-person operation on exactly this philosophy, and I packaged the scripts into two products: the Ops Starter Kit (the monitoring + incident checklists, fill-in-the-blank) and the Automation Starter Pack (the actual cron-ready scripts). One-time price, no subscription, no agent, no seat count.

If your bill is $2,000/year and these replace even half of it, the kit pays for itself in a month. If it doesn't, tell me in the comments and I'll tell you what your stack genuinely needs — for free.