Tech IndustryFeb 4, 2019
MicrosoftSvcuuhds

Do Data Analyst jobs have future

Throughout my 10+ years career I’ve been at different jobs - SDE, PM, Product Owner. For the last 3 years I started to look more deep into Data Analyst/Data Science area - watching online courses, reading books, doing some experiments after work, and I enjoyed it a lot. I’ve completed a couple of certification programs, and each time I got to work with data at my current job I felt happy, passionate and caring. I think I really want to turn my hobby into the job and start looking for positions. However I do not have Data Analyst experience, it is all just bits which I did at my previous jobs where I took any chance to work with data, analyze it, visualize etc. I am familiar with Data tools in Excel; PowerBI; Azure ML; Kusto; have some pet projects in R and plan to learn Python. Most significantly I have patience and very detail oriented when it comes to analysis. I can spend hours without noticing if I need to get answers or visualize trends or find correlation etc. So basically I have 2 questions: - is it real to find Data Analyst jobs given my job history; - do you think Data Analyst jobs have future? I don’t want to switch if it goes nowhere...

Add a comment
Nvidia $$$$:) Feb 4, 2019

If you are happy with 100-120k salary then data analyst ..want to make more move to ML

Facebook alt_qq Feb 4, 2019

Nah data analytics roles in fang pay more than that At MS it falls on the eng track so the base pay keeps pace with SWE/PM/DS if you're in the right team/org

Nvidia $$$$:) Feb 5, 2019

Do you know the Salary range for data analyst ?

Facebook alt_qq Feb 4, 2019

If you're at MS just find an internal data analyst position to apply to, you can spin your story and skills around to sell yourself I think data analyst position at MS technically fall on eng track so your base pay at least keeps pace with SWE (though equity falls way behind)

New
zorkan Feb 4, 2019

Yeah, but I think it depends on what you do as an analyst. The problem with DS/Analyst title is that the variance between teams and companies is high. For modeling work, I think the AutoML stuff could lower the demand over the next few years. AB testing, feature engineering, and coding up + training (classification/regression) models is not hard. Everything you would do for a kaggle competition will probably be automated away in less than 5 years time. The advent of deep learning and cloud compute is shifting the value away from domain expertise and towards data acquisition, cleaning, and visualization. Eg; compared to hiring ML researchers to invent/code up better models and algorithms, you will see much better ROI as a business if you invest in getting tons of quality data, then train simple but larger models (more cycles) in the cloud. Cloud compute providers (A, G, M) are working very hard to lower the value of modeling skills, because domain expertise in this area is the complement of the cloud market (reducing cost of one increases demand for the other). But if your analyst role is only doing data cleaning/visualization/BI stuff, then you are probably fine. I think if you find something in this field that makes you passionate and you see the value in it, then you should go for it either way. https://www.forbes.com/sites/forbestechcouncil/2019/02/04/why-there-will-be-no-data-science-job-titles-by-2029/#2eaff9433a8f

Apple Pneumonia. Feb 10, 2019

AutoML is vaporware, don’t believe that marketing garbage. Data modeling and analytics are very complex how are humans going to automate that before autonomous vehicles? I think you have it all wrong.

New
zorkan Feb 11, 2019

I agree with you on many levels. Especially for the more complex problems (eg; problems requiring guided user manipulation of entities, reasoning and understanding, multi-step planning, causation, ect..). I think to automate this stuff at scale you basically have solve something equivalent to automatic program generation. The main critique of AutoML is that it does not capture the way machine learning is changing (streaming data in motion, online learning systems, federated learning, RL, ect..). But currently, there is a massive ammount of low hanging fruit that can be boiled down to simple & static classification or regression, and domain experience is becoming less valuable than data collection for this class of problems.