How Much Does Turnover Cost?

People are not commodities. That being said, it is absolutely critical to quantify the outcomes of our People programs.

Here’s how I think about it using a Software Engineering role in the US as an example.
• Average cost of hire for the US is ~$4,700.
• Average time to fill in the US is 42 days.
• Recruitment agencies typically charge 25% (15-33%) of offered salary to place folks.
• Median software engineer salary in the US is $120,730 according to the BLS (2021 data...this is obviously dated and on the low side).
• Tech turnover is ~13% on average.

Some takeaways:
• If you're a tech firm in this example, for every 100 folks on your teams, you’re going to lose 13 of them per year, on average. 
• For every person who leaves your team, it’s going to cost you $4,700 to backfill that role (~$35,000+ if through a recruiting agency, using our software engineer salary for example). For 13 roles, that's $61,100 per year in hiring alone, even if you’re not using recruiting agencies.
• You’re going to lose 1.5+ months of development time at 42 days time to fill…that’s ~11.5% of the productivity for that role in a year. 
• For every 8.7 folks your team loses every year, they’re essentially working at -1 headcount on an annual basis.
• Every 5 business days of your time to fill equates to ~1.9% productivity gain or loss for that role.
• That $4,700 cost per hire is ~3.9% of the median salary in the case of a software engineer. Flipping this, every person retained is ~$4,700 - $35,000 cost aversion through hiring alone.

This doesn’t even take into account the benefits and intrinsic value created by these roles nor opportunity costs.
• E.g. SaaS firms likely invest 70-80% of expenditures are in people that are generating significantly more value back to the business.
• E.g. Disengaged teams deliver poor products and services, decreasing the value of the business as a whole, continuously in a vicious cycle.
• E.g. Sales & marketing roles generate revenue and profit for the business at a ratio greater than 1:1 (at least they should).

Note: these are benchmark averages. You can certainly use your own stats to help populate these calculations.