Last month I audited a client's monitoring bill: $540/month for an APM platform that three engineers had partially configured — and two of them had left the company.
The dashboard was pretty. It was also wrong: it monitored endpoints nobody hit anymore, missed the actual money-path API entirely, and had alert rules that paged the whole team every time a batch job ran.
Here's the uncomfortable math: for most small-to-mid teams, 90% of the value of a $500/month APM subscription comes from five boring checks you can run with free scripts. You keep paying for the other 10% out of habit.
What the $540 actually bought (and what it should have)
The essentials that generate real signal:
- Heartbeat check on the money path. One curl to your critical endpoint every 60s, alert on 2 consecutive failures. Distinguishes "slow" from "dead."
- Disk + memory guard. The #1 silent killer of small deployments. 5 lines of shell, runs in cron.
- Log-noise tripwire. Grep yesterday's logs for ERROR/Traceback counts vs. 7-day baseline. Rising trend = someone should look today.
- SSL + domain expiry. Nothing humbles a team like a site outage because a cert expired. Free to prevent.
- Dead-man's-switch on your own jobs. Your backups and cron jobs should check in when they succeed — and scream when they don't.
That's it. That's the stack. Run it from a $5 VPS.
The part nobody tells you: the checks are easy, the system is hard
Scripts aren't the product. The product is:
- knowing which endpoint is the money path (hint: it's rarely the homepage),
- alert rules that don't cry wolf (2-failure minimum, severity tiers),
- a runbook that says what to do at 2AM, not just "error detected,"
- a weekly pulse review: what alerted, what was noise, what did we fix?
That discipline layer is where most teams fail — and it's exactly why they end up paying $500/month for someone else to have opinions about it.
What I actually recommend
If you have a platform team: fine, keep your APM, but make sure it does the 5 checks above correctly — most don't.
If you're a small team drowning in dashboards: rip it out, run the 5-script stack, and reinvest the $6k/year into the thing that actually earns money.
I packaged my own version of this — the exact scripts, the alert thresholds that survived a year of real incidents, and the runbook template I hand to every client — in the Ops Starter Kit ($14). If you want the deeper automation layer (self-healing loops, scheduled audits, evidence-based reporting), the Automation Starter Pack ($19) adds it. Both are one-time purchases, no subscription, and they fit in a single afternoon to deploy.
(This week both are inside the Ops Mega Bundle at launch pricing if you'd rather grab the full stack at once.)
What's the most useless thing you've paid for in your monitoring stack? I'm collecting horror stories for a follow-up post — the best one gets featured.
Top comments (0)