- by SethBreedlove, Data Doctor
- 10/10/2014 2:42:20 PM
@ Matt, having worked in healthcare, I feel your struggle. When it came to analytics, what I remember most was suggesting something, hearing we can't do that, then six months later hearing, okay, yeah we need to do this.
But I do believe the industry is changing from a reactive to a more proactive model. Much is due to Medicare telling hospitals to cut out the crap and pay for their own mistakes. Excuse my English. There can be no incentive to change when the old way earns more income for a hospital, so that is where the change happens first.
- 9/17/2014 2:57:36 PM
@Cordell, that's an interesting question. Will data scientists become the coders of the next revolution? I like to think data scientists have more cache than that. It seems far less straightforward, given all the variety of data they'll need to be able to work with, than a coding job.
- 9/17/2014 2:47:14 PM
@MattH, given the healthcare crisis in this country, I would think the sorts of projects you've described would represent great opportunities for data scientists -- such big challenges to solve with such widespread impact. But perhaps I'm too idealistic (since it's not my salary at issue).
I'm curious, too, whether your HR folks understand analytics/data science enough to even know where to begin looking for appropriate potential job candidates. Just earlier today I talked to a marketing analytics director I know in another industry who said he finally had to take the hit and use an external recruiting firm to find on-target candidates for his purposes because internal HR had no clue how to do it, and as a result kept sending him people that had no experience for what he was looking for.
- by MattH17, Prospector
- 9/17/2014 1:48:46 PM
I'm in the healthcare industry. I work for a health system that owns an HMO- I'm assigned to the HMO.
The types of work we should be doing are things like predictive analytics re: readmissions, chronic conditions, LTVM, leaver/stayer analyses - the list goes on and on. The potential financial impacts of these initiatives is very large. Others in our industry see ROIs on data science in the >500% range.
I liken the struggle of trying to explain to executives what data science can do for them to what it would be like to explain an airplane to someone who had never seen one. They simply cannot wrap their heads around what they are missing out on.
- by Cordell, Blogger
- 9/17/2014 12:21:42 AM
First about @jamescon's comment on computers being fallible. We should remember that analytics only improves our odds of being right. Even machines with great algorithms don't nail it 100% of the time. Secondly the movement towards self service analytics is a bigger threat than wage inflation. Probably because of the higher cost of data scientist companies taking a longer view will hire scientist build (or at least design) systems that make them invisible to users. The question is will DS become the coders of the next revolution? And get paid like them as a result.
- 9/16/2014 5:03:22 PM
Jim, you ask, "Will data scientists be equally selfish if they resist having machines do what they do today?" I don't think so -- data scientists thrive on their ability to use machine-learning and AI, after all. Likewise, they'd take it as a challenge to come up with a way to prove computers wrong should those computers one day say, hey, we don't need you. (Somebody needs to provide a power supply, after all.)
- 9/16/2014 4:56:14 PM
@MattH -- thanks for sharing. I'm curious what industry you're in, and what type of challenges/problems you're data scientists would be addressing. Does the potential impact compensate at all, if if just a little, for the lower salaries, or are the salaries so below desired targets that the type of project and implications just don't matter?
- by CandidoNick, Data Doctor
- by MattH17, Prospector
- 9/16/2014 11:21:02 AM
The portion of your post relating to companies unwillingness- or inability- to adequately compensate data science professionals is what struck me the most.
As somebody who is currently trying to build and recruit a data science team at a company far outside the Fortune 500, I can absolutely attest to the nightmare this presents. Corporate HR has not caught up to the point where they are capable of digesting the true ROI of data science. "We already have analysts, we pay them $XX,XXX. Why would we hire analysts at $XXX,XXX?"
The talent is not willing to work for the wages offered (in my opinion, justifiably so. Have you seen tuition prices lately?) and corporations are not willing to adjust the legacy payscale systems. It's a vicious cycle that I am having an incredibly hard time finding a solution to.
Outside of your Googles, Amazons, etc., how can the rest of us ever hope to make headway? The solution we have adopted is a system of hiring bright, inexperience people and training them up. Perhaps bringing in contracted assets to assist and allow job shadowing to occur. It's a long term strategy that may not ever pay off, but I see no alternative.
- by Jamescon, Editor
- 9/16/2014 9:50:17 AM
If senior managers won't approve analytics projects, is it because that could lead to the managers being made obsolete? If yes, is it because they are selfish, putting their own careers ahead of the good of the corporation?
Will data scientists be equally selfish if they resist having machines do what they do today?
OK, I took that to what may be a level of absurdity. In reality the human/computer relationship isn't an either/or thing. Computers help people make decisions, they can't make the decisions for them. At minimum, a human looks at what the computer produced and says, "Yes, do it." or "Wait, something isn't right." In the latter case, more investigation is needed. It could be about hunches (that dreaded word in data science) or about experience on the part of the human.
What we tend to lose sight of in these discussions is that computers are fallible. We've all seen it in things like a simple calculation that sums wrong or a web search that "misses" key results.
We blame the computer for those goofs. Actually, the problem probably results from operator error or, in many cases, a human error far back on the road to results. We forget that fallible humans made those computers. Are you ready to have computers make final rulings with no human oversight on risk-the-business and risk-the-employee decisions? As a data scientist, what would you do if a computer said, "We don't need data scientists"?