Data science job ads that do not attract candidates, versus those that do

There's no shortage of great talent for companies such as Apple, Google, Intel, Facebook, Wikipedia and some exciting startups. But what if you are not one of these?

I received the following job ad in my mailbox (see below in italics), from a third-party recruiter, and it's probably for a data science position at Nike near Portland, Oregon (my guess). Basically, it's a 6-month gig to build an A/B platform.

I discuss here a few aspects that make this job ad unlikely to attract talent, as well as remedies.

It was a mass-mailing, not customized to the recipient. The recipient will think that you did not check if he/she is a good fit, and could have the impression that he/she is competing with hundreds of other candidates - essentially everyone in your database.
For a location such as Portland, Oregon, you might want to offer telecommute or at least be OK with a candidate in Seattle. Write it in the job ad!
You are asking for 3 clusters of non-compatible skill sets: (1) communication, (2a) developer - web architect (big data), (2b) developer - dashboard, BI, and (3) statistics / machine learning guru.

This skills mix is not found in typical employees. Granted, this is a consulting job, but you should then advertise it as a consulting opportunity. You would have to offer $300/hr to motivate the few guys that have this rare mix of talent and knowledge. There are consultants or small consulting companies (with this mix) willing to do the job for $300/hour (motivation is even bigger if you reimburse their weekly hotel/travel expenses, and consulting firms are better equipped than individual consultants as they can use a mix of cheap, junior with expensive, senior consultants to lower the cost). Or do you really need to build such a platform in the first place? Why not get one from a vendor instead?

You mention that machine learning / statistics is a plus (not a requirement), yet the core of the job is developing an A/B testing platform. Something where deep statistical knowledge is most critical.

For a regular employee, you would want an engineer with some statistical knowledge, train him so that he becomes an expert in experimental design (aka A/B testing) - at least in digital experimental design. The guy needs to acquire a deep knowledge of Internet/server architecture, traffic flows and web metrics, to understand source of biases in this type of A/B testing. Yet he needs to be very knowledgeable about statistical theory and non-parametric, robust confidence intervals in the context of big data and most importantly - in the context of doing TONS of A/B tests, whether the data is big or small.

In short, you need someone with some very specialized, narrow (not comprehensive) statistical knowledge (not an expert in logistic regression) and very specialized (but not comprehensive) knowledge about Internet architecture. These people exist, they are domain experts. Would a guy like me, who developed his own home-made Map-Reduce / Hadoop environment, qualify? No, I would surely be turned down for not having Hive/Pig etc. And that's a mistake.

In addition, you want the person to also be a developer. Here of course, I would not qualify: I'm a web developer, I developed this very website for instance as well as analytic API's, I managed developers (consultants), and haved used scripting languages for a long time, but I'm not the developer with the kind of team experience that you really need; companies made the mistake in the past to believe that I am a developer -- don't assume that you can turn a creative, independent guy like me into a useful team developer, it just does not happen, what I offer to a team is different. And while I'm great at designing the architecture of simple and useful dashboards (and I love it), I do not "code" them - yet I'd still be a good asset for the design part of the dashboard. Maybe dashboard creation could be done (at a low cost) using an automated tool (vendor) or templates requiring little to no coding, rather than having a consultant do it. Then you can remove "dashboard creation" from the list of tasks, broadening the pool of applicants.

My recommendation: hire two consultants - a domain expert like me for a limited number of hours, maybe 200 - (or better, one that has worked with true Hadoop if you can find one). And a developer who will closely work with the domain expert - at least in the initial step when the architecture / design of the platform is discussed, as well as during final steps when you test it on real data. Maybe hire a business analyst for the dashboard creation, or use one from the employer, or purchase/customize a product.

Here's the email with the job ad:

Greetings Vincent,

I came across your profile on our database. We may have interacted with you in the past on earlier job opportunities. I wanted to touch base with you to find out your current availability to consider new contract opportunities with our clients Nationwide. Please let me know as soon as possible about your availability and interest so that I can initiate further discussions and provide details of a few new roles we are trying to service currently.

Details of current opening which might be of interest to you:

Title: Backend Web Developer - big data preferred

Location: Hillsboro, OR

Contract Duration: 6 months

Responsibilities:

Regular communication with your teammates about your project(s)
Communication with key project stakeholders to ensure that features are being implemented according to plan.
Build tools to provide actionable insight/develop analytics dashboards
Creatively figure out how to use web data to inform decision-making at our client's open source developer group
Develop and evolve the backend data processing for a suite of web-based analytic tools, eventually including an A/B testing platform used across the group to inform all product decisions
Work with web developers to continually acquire new data
Regular communication with your teammates about your project(s)
Communication with key project stakeholders to ensure that features are being implemented according to plan.

Qualifications:

Exceptional communication and organizational skills
Active listener
An advanced degree in Computer Science
A combined 4+ years of experience in software engineering - big data preferred.
Experience in cloud technologies such as MapReduce, Hadoop, Hive, and / or Pig
Strong database/SQL skills
Experience in one or more scripting languages
Application, website, and/or software development
Analytics, statistics, machine learning, and A/B or Multivariate testing experience is a plus
Work effectively and efficiently
Have great attention to detail