Analyzing Amazon: Was Talent the Tipping Point for HQ2?

The ability to find the workers needed ended up being a crucial element in Amazon’s decision to split its second headquarters between New York and northern Virginia. But what is Amazon looking for in its workers? Is it so hard to find?

We’ve had a lot of interest in an analysis we did for the Wall Street Journal about how Amazon and Google compare in terms of what they are looking for in workers. Since analyzing hiring is a key tactic in workforce development, strategic human resource planning, and corporate competitive intelligence, we thought we’d go into more detail about Amazon’s hiring here.

Amazon Software Engineers: Programming with a Side Order of Data Science

Amazon’s problem may not just be finding enough people for the job titles they need but finding enough people with the particular skills they require for those jobs. Amazon’s conception of certain jobs can be significantly different from that of other employers.

The largest numbers of job postings by Amazon in the last 12 months are for Software Development Engineers (4,897), Software Development Managers (1,269) and Solutions Architects (1,135) out of 42,192 posted jobs overall.  But software developers at Amazon look very different from developers overall.

For example, Amazon job postings are five times more likely to ask for C++ (62% vs 13% national), and three times more likely to ask for Object Oriented Analysis and Design (43% vs 13%).  The demand for OOAD, a specific approach to developing software, is an important sign of how Amazon operates internally.

But a bigger point is that Amazon’s software jobs are also heavily influenced by the need to work with big data. Some of the prominent skills for developers at Amazon include Python (35.8% of Amazon jobs vs. 19% of all software engineer jobs nationwide), data structures (35.3% vs. 4.5%), machine learning (21.4% vs. 2.9%), and big data (11.3% vs. 4.9%) —all of which are big data skills. (Python, while it is used in many applications, is a favorite in big data work because it is relatively easy to program). That shows that Amazon software jobs are as much, or more, a part of the data economy as the software economy.

Why New York and D.C.? Comparing Labor Markets

Based on the skills Amazon needs, what’s the advantage of locating in Queens and Crystal City, Va., (just outside Washington)? There are two advantages, based on our data: concentration and cost.

Using our job posting database, we looked at the demand for the skills Amazon wants in several metro areas. In each we identified the “location quotient,” or whether the demand for a specific skill is higher or lower than the national average. Based on the table below, you can see demand in the New York and Washington metro areas is high—but certainly not as high as in the ferociously competitive tech regions of San Jose and Seattle.

Concentration of Software Developer/Engineer Postings with Specific Skills, by Region

 New York MSA concentrationNY MSA job postingsWashington DC MSAWashington DC MSA postingsSeattle MSA concentrationSeattle MSA postingsSan Jose MSA concentrationSan Jose MSA postingsNationwide postings
Machine Learning1.52,0552.51,1437.11,97718.62,90421,068
Big Data1.83,9773.42,5593.51,59110.92,80034,705
Natural Language Processing1.947732634.121716.85014,039
All Software Developers1.467,3782.845,0092.625,8687.240,261753,884

Significantly, it also appears to take less time to find software developers in New York and Washington. An important metric is posting duration: how long a job posting stays open gives an indication of how hard it is to fill jobs. In San Francisco and San Jose, posting duration for software developers is 47 and 46 days, respectively, compared to the national average of 39 days. The average duration in New York matches the nation at 39 days, and Washington is slightly higher at 43 days.

New York is the “law of numbers” choice. Even though it doesn’t have the same dramatic concentrations of big data software demand as in Silicon Valley, Seattle, or Washington D.C., the sheer size of the city and its talent pool makes it a player. In raw numbers of software jobs posted, New York ranked first or second among the cities under consideration.

By contrast, the D.C. metro area (including Crystal City), is a great pick for concentration of software developer jobs, even if the absolute numbers are smaller. Excluding the West Coast tech hubs, the D.C. region punches well above its weight for relative concentration of demand, generally coming in at two or three times the national average. Yet the shorter posting duration suggests employers have less trouble finding talent.

The Washington region has another advantage: cost. The average advertised salary for a software engineer with machine learning skills is $116,000 per year in the Washington metro area, compared to $144,000 in San Jose, $135,000 in New York, and a national average of $122,000.

Real-time labor data analysis can give a significant advantage to analysts on many fronts. As a source of corporate intelligence, companies can use it to site facilities near the best talent pools or scope out what their competitors are doing. And for workforce and economic development officials, this analysis can provide key insight into what businesses need, and what a region has to offer them in terms of talent.

You can conduct this kind of analysis using our Labor Insight™ analytics dashboard. Find out more about Labor Insight, as well as our other products for competitive intelligence and workforce development.