csctechtalk

02/12 TechTalk Soundbites
New Frontiers in Big Data
   6 years ago
#csctechtalk01/29 TechTalk SoundbitesThe Evolution of Hanlon OSS for automated server provisioning
   6 years ago
#csctechtalk02/19 TechTalk SoundbitesOpen Source at CSC - Update on Progress and Practice
Jerry Overton
What's most important when creating an enterprise-wide analytics and experimentation community?
Matt
I think its pulling togetehr something tangible today that everyone can contribute to throughout the org
Radhika Subramanian
connect with business users, use-case and how they will use the results in their day-to-day lives
Bob Donnelly
Access, be it from home computers, of where people want to experiment.
John Furrier
I think that sandboxes (cloud) must be available to experiment with different technologies - maker culture meets app dev
Dorothea Gosling
keeping it creative, while focusing on business-relevant outcomes
Jerry Overton
@drbobcsc Absolutely. Speaks to what Radhika was saying about better toosl and access. Totally agree.
Matt
That builds the basic principles for people to grow from, something that impacts our customers end-users
John Furrier
have an horizontally integrated focus - meaning integrated stack development - legacy tech is mostly vertical
Kyle Zellman
Cultural acceptance and data availability
Radhika Subramanian
I agree. Business relevant outcomes.
John Furrier
Developers need a "software factory" environment (aka cloud)
Sorin Costea
@DoroGosling you can't really focus creativity, just reap what seems better usable
Matt
@khzellman Culturally Accepting that 90% of the time I'm abysmal at statistics, the other 20 I'm average
Steve Erhardt
A place to share tools, building blocks, experiences, successes and failures
Dave
@sorincos @DoroGosling Isn't that what makes a great tool? Like the iPhone, it doesn't get in the way of the user.
Steve Erhardt
Also a desire to help each other and learn from each other so that the whole is much more than just a sum of the parts
Sorin Costea
@tohavenostyle to not get in the way is actually lots of background work, careful!
Radhika Subramanian
Need to remember that there needs to be a delivery with ROI..... data professionals sometimes forget that... in the fun experimentation..
Dorothea Gosling
@tohavenostyle agree! that's a great example. And I think that'll be key for big data success. I had my "aha-moment", when I switched from a well known spreadsheet to Tableau.
Adrian Jones
Clear objectives so there is a real purpose
Sankar Vema
we adopted the model of quick viable technology concepts as model for NextGenTech at ATD
Theyaa Matti
A community with people from different disciplines working on the same idea from different point of views. That makes us a great community and everyone learns from everyone
Dorothea Gosling
@sorincos I'd disagree with that. As you brainstorm and cook up new stuff, the bits that are relevant - in a business setting - will bubble to the top. Doesn't mean you have to discard all the other ideas, prioritize ones that have the potential to move
Sorin Costea
@DoroGosling aren't we saying actually the same thing? :-)
Jerry Overton
@DoroGosling Yes, creative and interactive -- I think.
Dave
@sorincos Yes, true in many senses. Getting out of the way is a skill many people should practice and a feature many products lack.
Lisa Braun
We did a report on big data a few years ago, where one of the themes was "let the data lead." Don't go in with pre-set structured queries, but see where the data linkages take you in terms of new insights. Big role for data visualization here. Comments?
Jerry Overton
That scares me a bit, though, because these results have to be presented in context.
Radhika Subramanian
Visualization as a means of discovery is very very difficult. Visualization works well once the results have been computed. Makes it easier to digest
Sorin Costea
risk is to get random ghosts
Theyaa Matti
Data visualization is essential, at the end the data, within big data, has to have a nice representation for everyone to consume. I believe tools like D3 will be major players in the near future.
Radhika Subramanian
I agree with Jerry - blind data discovery is painful.... sounds like fun... but is not!!
Sue Cronizer
One of the biggest problems now is the amount of data which is overwhelming. Visualization gets you part of the way but its difficult dealing with the amount of data we're seeing now
Steven Melanson
Visualization can be one powerful tool when beginning your EDA for any data intensive project. While it should not be the be-all-end-all, it can definitely give insight into potential problems you may run into during training/testing
Lisa Braun
I may not be stating it accurately, but the idea was to discover and explore new linkages...as applied to criminal investigation, for e.g. Don't know exactly who you're looking for or where the data will lead you.
Kyle Zellman
Letting the data lead can be very valuable depending on the problem. If you enter certain projects with preconceived notions of what you're looking for, you may miss the most salient insights in the data
Steven Melanson
@khzellman Not only could you potentially miss insights but you may even try to force the data to agree with your pre-conceived notions which I think is even more dangerous with dealing with prediction
Dave
The big barrier with visualization (really 2D) is dimensionality. Have to move past these visual interpretation to move beyond the couple axis analysis.
Kyle Zellman
@DataScience_SPM absolutely. can lead to false results, which we both know are worse than no results
Sorin Costea
our brain is a great visual pattern analyzer, so that'd be a great way to go. Just the caveat of false matches...
Paul Lavicka
My data is really just 'Little Data'... but need that visualization tool to get the analytics out to my target users.
Jerry Overton
But maybe not, there's more to big data than just sheer volume. I mentioned before that variety and velocity are my personal favs
Jerry Overton
How do you get an organization to become more data driven?
Radhika Subramanian
Deliver success and broadcast the result. This creates the "Jones" effect..
Kyle Zellman
education leading to a cultural shift
Sukanya Kuppuswamy
I think one good way is to present a technology stack that is lean and easy to use
John Furrier
great question - first is to be "data full" meaning start playing with the data to do internal R&D then figure out how to wrangle it and harness it
Matt
at the risk of sounding like a broken record, Actually pick something small, that shows results (usually in $ terms, cost take out etc) then build the trust taht youre not just "mucking about with data"
Jerry Overton
@RadhikaAtEmcien Matches my experience too. People have to see how this stuff can be used. They need examples.
Matt
We also have to be context aware, for many customers uploading business created data (regardless of what) to any analytics/Cloud services is still a non-starter (fear, Security, legality, residency etc)
Carl R
Perhaps the need for and ‘value’ of [quality] data needs to be promoted. Explain why it’s important, the benefits to be had etc. etc.
Radhika Subramanian
@MattMJH Totally agree.... not a broken record... :))
Jerry Overton
@CCR2000CCR Yes, this ties back to an earlier comment around making the success visible.
Theyaa Matti
Having executive support. In large organizations, you need to have support from higher management to get the buy in for big data and analytics