After running our team's CI/CD pipeline on Anomali's "Professional" tier for 18 months, we upgraded to the "Expert" tier six months ago based on promised performance gains and advanced features. The cost jump was significant—2.5x our previous spend. Our goal was to quantify the actual return.
Our primary benchmarks centered on build performance, concurrent job throughput, and cost-per-execution-minute. The results were revealing.
**Key Performance Data (Averaged over 6 months):**
* **Build Time Reduction:** Expert tier VMs showed only a 7-12% improvement in raw execution speed over Professional, not the "up to 40%" suggested for our workload.
* **Concurrent Job Cap:** While we gained +10 concurrent jobs, our actual usage data shows we hit the new limit only 14% of the time. The overhead didn't justify the always-on cost.
* **Cost Efficiency:** We tracked `(total monthly cost) / (total pipeline execution minutes)`. The Expert tier's cost-per-minute was **2.1x higher** than Professional.
The "advanced" features also didn't deliver as expected. The proprietary "Predictive Scaling" often over-provisioned, negating any potential savings. The AI-powered test optimization suggested flaky changes that broke our deterministic builds.
The most tangible benefit was the increased storage for artifacts, but a quick calculation showed using their integrated cloud storage would have been 60% cheaper than upgrading the entire tier.
**Bottom Line:** For teams already on Professional, the leap to Expert is hard to justify based on raw performance and cost metrics. The value proposition simply doesn't hold up under measurement. The resources would be better spent on optimizing pipeline code or provisioning dedicated runners for specific heavy jobs. We are downgrading at the end of this billing cycle.
Numbers don't lie
Interesting that the cost-per-minute metric was worse. We track something similar using Looker dashboards and saw the same pattern when we tested a tier jump with a different vendor last year. The promised concurrency is rarely the bottleneck, it's the actual utilization rate.
Did you try isolating the "Predictive Scaling" feature to see if turning it off changed the cost efficiency at all? Sometimes those extras are baked into the tier cost but you can disable them.
Great point about isolating features. We actually did test the Predictive Scaling toggle for a full sprint. The cost-per-minute did drop a bit, maybe 5%, but it was still way above our old Professional tier baseline.
It feels like they're bundling a bunch of niche "enterprise" features, like enhanced audit logs and custom security roles, that most teams just don't need. You're paying for shelfware. Wish more SaaS tiers were truly modular.
Beta tester at heart
That's a really good way to put it - shelfware. I've seen the same thing with some monitoring tools.
> like enhanced audit logs and custom security roles
This is exactly what pushed us into a higher Docker registry tier we didn't really need. We wanted image retention policies, but had to buy the "compliance pack" with it.
Your test with the Predictive Scaling toggle is interesting though. A 5% drop is something, but I guess it proves the core resources aren't that much better. Thanks for sharing that detail.