rockyourdata

Please post your comments here

Agile practices are key for the iterative nature of data exploration and model development

Agile is the foundation for #cloudnative and it's not only the only methodology but it drives cultural and behavior change withing orgs and developers in IT

simply put Agile drives more innovation and possibility for higher revenue for customers and cost advantages for higher earnings

@DrZeydy Agreed. Agility is about improving your product iteratively via testing hypotheses. Fail fast.

Agile practices call for building what I'll call the "minimally viable data product" https://en.wikipedia.... . In other words, development of great data products requires embrace of the fail fast, learn&refine fast philosophy.

Agile approach is iterative in nature; it matches the inherent property of growing data and better data interpretation outcomes.

@jameskobielus Never heard of that term but like it.

Time-to-value, baby! Agile is the way to go these days. Waterfall is all but a thing of the past. Agile helps in lots of ways, but only if collaboration is real & effective. The business must work closely with developers in a continuous fashion

Agile depends on requirements and solutions evolving through collaboration between self-organizing, cross-functional teams. That's a great approach for building innovative disruptive products of all sorts, including data products.

Agile and iterative dev is essential - even more so in #DataScience products - hypothesize, test, improve: true for Agile, true for Science.

Agile requires that you don't overbuild your data products. They don't need to do everything, just a few things, but to do them exceptionally well. Simplicity is powerful value, and easier to iterate & refine long term.

@MikeTamir Mmmmmm... science.

Time for Q6, stay tuned. Look at the top of your screen

Agile is aligned to the exploratory investigative nature of the best data science. As you iterate models, bring in fresh sources, and identify heretofore hidden patterns, new ideas spring forth for value your data product can deliver

Please post your comments/replies here

Empathy, you have to know why this data is important and how it will be used #rockyourdata

To build data products, you'll likely need all the skills associated with #datascience: subject matter expertise, computer science (tech/programming) and math/statistics.

@bobehayes Good point Bob!

And a little patience. #BigData

@nanselm2 I agree. Need to build empathy into everything we do. Understand end users' perspective.

you have to master all the cycle from gathering, cleaning, storing and analyze data.

this is Data product dependent: architecture, engendering and dev in addition to Algorithmic understanding all can be vital

You should be next to the Danger Zone, you should have Hacking skills, Math & Stat & Subject Expertise

Takes a variety of skills so the data product will be the result of multiple professionals bringing their complementary skills to bear on a project.

@datascience_tn Now that song is stuck in my head.

You need to have creative imagination to grasp how this, or any product, can address a clear need. You need to identify how data & analytics can add value. And you need to build, test, and deliver it effectively. Product management

Time for Q3

It's important, if you're a data product developer/manager, to start with a business case that identifies how you might monetize it. Check out my blog: http://ibm.co/1NiCvX...

Time for Q3 please

You definitely need a team of multi-disciplinary technical skills, but it all starts with the ability to understand the clients business problem and frame how to go about solving it.

The ability now to integrate structured data and unstructured data, and do it in near real-time so that you can make a decision on something based on early indicators and then merge it and integrate it

if monetization is a criterion of a data product you must have visibility on revenue generation

Time for Q4, take a look at the top of your screen

Well, that's #the $64k question, right? You need someone who understands the data, someone who gets architecture, someone who gets the business side, someone who gets the creative side; and obviously, someone to keep the trains running on time!

Please post your comment/reply here

Waiting for your comments

Please post your comment/reply here

Here's approach I used in developing #datascience skills survey: Generate idea > develop prototype > gather data > test idea (analyze data) > tweak / modify product > validate

Identify the need, frame the business plan, assemble the team and resources, develop and test, deliver, measure, iterate, refine.

It really depends on your choice of methodology Cross Industrial Standard Processes for Data Mining (CRISP-DM) vs Standard Methodology for Analytical Models (SMAM) , but either way it starts with understanding best fit use case.

I basically follow the #scientificmethod.

It's the standard data science workflow embedded within a larger product development/management workflow. Data scientists and subject domain experts are your key developers.

Ingestion, storing, wrangling, making it addressable, and integrating it in the apps

I'm a big fan of the CRISP-DM method.

applications focus on vertical integration and data should be enabled for horizontal integration; these are not mutually exclusive

@bobehayes I'm not familar with CRISP-DM model

For a data product, I like to mix the engineering method (problem-solving) w/scientific method (hypothesis testing). Focus on biz problem first!

John Daus: does that model deal with real time or is it for a stored corpus of data?

@furrier https://en.wikipedia...

A big part of the workflow is the process of identifying whether you have the right sourcing of data to make it a sustainable product. And the right data scientists to iterate/refine the product as an ongoing operation.

Real-world experimentation may be built into the ongoing develop, test, iterate, refine cycle of the data product. In that case, the workflow is one of repeated trials, metric gathering, and feedback loops.

@jameskobielus Agreed on this we also have to keep in mind some Dataproducts are about the piping or more C facing in which case engineering and close cycle time iteration comes to bear

This really depends on the team and product type, but generally speaking, someone starts with an idea, usually triggered by analyzing data within the context of a business process. A prototype is developed, tested, refined then hardened for use