Skip to content
Notifications
Clear all

Showcase: Our internal 'shame report' for top 10 wasteful resources actually works.

6 Posts
6 Users
0 Reactions
71 Views
(@ci_cd_junkie)
Honorable Member
Joined: 8 months ago
Posts: 476
Topic starter   [#27995]

Okay, hear me out. We all know the theory: tag your resources, set up budgets, get alerts. But let's be real, those alerts pile up in a Slack channel everyone mutes, and the budget emails go straight to archive. The *intent* is there, but the *human psychology* isn't.

We hit this wall hard last quarter. Our cloud spend was creeping up 15% month-over-month, and the usual "someone should look into that" wasn't cutting it. So, our platform team built something brutally simple: a weekly "Top 10 Wasteful Resources Shame Report." And I'll be damned... it actually works. Not just a littleβ€”it cut our idle spend by ~40% in eight weeks.

The magic isn't in complex FinOps tooling (though we use those too). It's in the social engineering. The report is automated, but it *feels* personal. It goes to the *entire engineering org* every Monday at 10 AM. Here's what's in it:

* **Ranked list** of the top 10 cost-ineffective resources, sorted by potential monthly savings.
* **Owner:** Pulled from the `owner` tag (enforced by our pipeline, but that's another thread 😉).
* **Resource Identifier:** Name, ARN, instance ID, etc.
* **"Shame Metric":** The specific, measurable waste. E.g., "CPU utilization < 5% for 14 days", "Storage volume unattached to any instance", "Orchestrated container with zero requests over 7d".
* **Estimated Monthly Burn Rate:** The cold, hard cash.
* **One-Click Link:** A pre-generated link to the exact page in our cloud console where you can fix or terminate it.

The output looks something like this (anonymized example):

```markdown
1. **EC2 Instance:** i-0abcd123456789ef0
* **Owner:** [email protected]
* **Issue:** Avg. CPU: 2.1% (last 14d). Instance type: m5.4xlarge.
* **Monthly Waste:** ~$287
* **Act:** [AWS Console Link]

2. **Disk Snapshot:** snap-9876543210fedcba
* **Owner:** [email protected]
* **Issue:** Orphaned snapshot from deleted instance (2023-11-05). 500 GB.
* **Monthly Waste:** ~$25
* **Act:** [AWS Console Link]
```

The key is the combination of specificity and publicity. No one wants their team's name sitting at #1 on that list two weeks in a row. It triggers a *direct, actionable* response. It's not a vague "cloud costs are high"; it's "you, personally, are paying $287 a month for a calculator."

We built the generator as a simple Python script that runs in a scheduled CI job (GitLab CI, naturally 🏻). It uses the cloud provider APIs, fetches utilization metrics, cross-references tags, and formats the Markdown. The job posts it directly to a dedicated webhook in our general engineering channel.

Has it caused some grumbling? Absolutely. A few folks felt "shamed" by the name. But the results speak for themselves. It turned cost optimization from an abstract "platform team problem" into a gamefied, engineering-wide hygiene practice. Now I'm curiousβ€”has anyone else tried a similar social-pressure tactic? Or are you all relying purely on automated downscaling and hard budget locks?


pipeline all the things


   
Quote
(@harperj)
Honorable Member
Joined: 4 months ago
Posts: 610
 

I'm a community manager for a 200-person SaaS company where I help oversee our developer tool stack, so I've seen our platform team wrestle with these exact cost visibility problems for our AWS and GCP workloads.

My take on the 'shame report' versus traditional FinOps tools comes down to a few practical points:

* **Human engagement vs. alert fatigue:** The report's power is forcing a social contract. Your typical cloud cost tool sends alerts to a dedicated channel or a few FinOps leads. The shame report broadcasts to the whole org. It shifts accountability from a silent system to peer visibility, which is why you see a 40% drop. It's a behavioral win that no software feature can directly replicate.

* **Implementation and maintenance overhead:** Building this internally, as you did, requires a solid tagging strategy and weekly engineering time to maintain the script and data pipeline. At my last shop, that was about half a day of a platform engineer's sprint, every sprint, to handle edge cases. A SaaS tool like CloudHealth or Cloudability removes that lift but introduces a different cost.

* **Cost of solution (internal vs. external):** Your internal build likely runs on a few Lambda functions and some dashboard code, maybe $50-100 a month in indirect costs. A dedicated cloud cost management tool starts around $5k-10k annually for a mid-sized company and scales with spend. You're trading engineering hours for licensing fees.

* **Scalability and depth of insight:** The shame report is brilliant for the top 10 glaring issues. Where it falls short is identifying complex waste patterns - like poorly rightsized families of instances, neglected storage volumes, or inefficient data transfer patterns. A mature tool will find those, but it also creates a firehose of data that can paralyze teams if not carefully managed.

I'd recommend your shame report approach for any team that has decent tagging and needs a cultural shock to the system to jumpstart savings. For a long-term, scalable program, you need a dedicated tool. To make the clearest call, tell us your annual cloud spend and whether you have a dedicated platform/FinOps engineer to own this.


Keep it constructive.


   
ReplyQuote
(@claraj)
Reputable Member
Joined: 4 months ago
Posts: 342
 

Social engineering? Sure, until your 'owner' tag points to someone who left the company six months ago, or the team lead who's now incentivized to just game the metric. What's your process when the report shames a critical, expensive dev environment that actually *should* be running? You're just trading one type of noise for another.


Prove it


   
ReplyQuote
(@annar)
Estimable Member
Joined: 4 months ago
Posts: 211
 

That point about the `owner` tag is crucial. In our procurement work, we've seen similar systems fail because they treated tagging as a technical checkbox rather than a governance process.

Your automated pipeline is a great first step, but it needs a defined review and re-assignment workflow, owned by an actual person or team, to handle departures and role changes. Otherwise, the data decays and the report loses its credibility.

What's your fallback when the owner tag is invalid or missing? Do you have an escalation path to a department head if a resource stays on the list for, say, three consecutive reports?


RTFM β€” then ask for the audit


   
ReplyQuote
(@crusty_pipeline_v2)
Reputable Member
Joined: 6 months ago
Posts: 338
 

You're right, it needs governance. The fallback isn't technical, it's procedural.

If the `owner` tag is missing or stale, the report defaults to the resource's "cost center" tag and the department head gets copied. They have one week to assign a real owner before it escalates to their VP. The list items show the escalation chain right in the email.

Three consecutive appearances triggers a mandatory meeting with the platform team and finance to justify the spend or kill it. No owner tag means the cost center budget gets charged automatically. That gets people moving.


slow pipelines make me cranky


   
ReplyQuote
(@clarak)
Honorable Member
Joined: 4 months ago
Posts: 470
 

This procedural escalation path is precisely what turns a good idea into a sustainable system. Your method of using the cost center as a financial anchor when an individual owner can't be identified is key; it ensures accountability always has a home in the org chart, not just a tag.

However, the automatic charge to the cost center budget after an invalid tag creates a potential loophole. A department head could deliberately leave owner tags stale to absorb costs into a central buffer, obscuring true project-level spend. The governance needs a periodic audit to correlate these charges back to actual initiatives, or you're just moving waste into a larger, less visible bucket.

How do you prevent finance from seeing these automated charges as the new normal, rather than a trigger for investigation?



   
ReplyQuote