💾 Archived View for dioskouroi.xyz › thread › 29419733 captured on 2021-12-03 at 14:04:38. Gemini links have been rewritten to link to archived content
-=-=-=-=-=-=-
________________________________________________________________________________
It depends a lot of nature of business.
For resource intensive business (e.g. video processing, big data as a service)I would expect 20-40% of revenue, so 2.4 - 4.8 mln/year.
For typical application (e.g. CRM, specialized app, integration) 5-10% of revenue, so 600k-1.2 mln/year.
Mature, stable companies tend to have lower numbers. Hypergrowth startups may have waste in their growth phase.
Happy to help with your AWS optimizations.
I think it depends on your pricing model. If you’re paying 5-10% for a CRM/specialized app/integration you’re likely charging too little. I’ve had bills be around 1-2% and that was willingly knowing I wasn’t optimizing for billing but rather development productivity. I could have cut the costs to under 1% but it wasn’t worth the effort vs. building net new functionality.
About 20,000 a month. We use Amazon as storage and backup only. Everything else is on premise.
What are your on premise infrastructure costs?
$600k/yr but there are two things of note
- we are woefully wasteful with resources. If it was worth the engineering time we could probably cut it down to $120k a year
- it’s not worth the engineering time because the business would rather the devs and infra prioritize feature work over cost cutting every quarter. Their philosophy is that if we can’t afford our bill the company’s got bigger issues.
Isn't that differ nce enough for two engineers? They could certainly get more done in addition to the cost cutting. Seems like an easy case to the biz team.
This assumes that you can hire the two extra engineers. In all likelihood, their business is constrained by the speed that new people can be recruited, vetted, and onboarded. This means that even with infinite budget every engineer is a precious resource and should be allocated to higher-value projects.