agiledatascience

[+] Show Hidden Comments

- Agile is a process that was designed to manage rapid, iterative work..

Please reply in the scrolling display directly under the question you're addressing.

- I've seen many successful Data Teams are now moving to Agile and Cross-Functional operation

c. Team-based work, shared responsibilities, uber-collaboration iterating through incremental stages towards ultimate solution #agiledatascience

- The advantage of Agile is rapid iterative development & rapid feedback cycles from customers, very different to older Waterfall style project methodologies

@dez_blanchfield Dez: Is agile usually a "designed" process, or is it usually more of an "emergent" process that naturally coalesces from iterative development ad-hocracy?

- Agile is everything that Waterfall methods like PMBOK & Prince2 are not

Typically prioritizing for the most valuable features or requirements and delivering small packages of those features first #agiledatascience

2 Votes Vote

Steve Ardire

agile enables you to change the kind of analysis you're doing depending on what the data is telling you.

Bob E. Hayes

Agile #datascience is development through collaboration of cross-functional teams.

@dez_blanchfield continuously re-evaluating and re-defining business priorities in tight cycles

@joe_caserta Joe: "Ultimate solution"? Sounds fairly final, cut-and-dried. How many real-world data science intiatives are that closed-ended vs. ongoing/exploratory?

from a software/tech standpoint, #agile is an approach that allow faster development than traditional SDLC processes

- Agile can come naturally to some but it's generally in my experience a designed process, it borrows from the likes of Taiichi Ohno's "Kanban" for example

- Fail & Fail fast requires an agile approach, PMBOK & Prince2 don’t work for Fail & Fail fast

Bob E. Hayes

I like the idea of using data to inform you where your development needs to head. You can leverage behavioral analytics. How do customers use your product/solution? Build what improves usage.

@sardire Steve: If Agile is about changing your analysis methodology mid-project, is there any point in doing upfront planning of the data-science project/workstream in iterative projects?

@joe_caserta - indeed, collaboration is a huge part of it but not exclusive to Agile, short sharp fast "sprints" are the key to Agile ;-)

- Reducing time to value is more easily realised through short sharp agile sprints

Agile is great process for developing new features, but not a good fit for implementing changes on complex systems

#agiledatascience is continuously iterative

@sardire - yes yes yes.. far easier to turn the Titanic around with an Agile methodology than with PMBOK or PRINCE2 ;-)

@8pathsolutions Jennifer: Does "waterfall" (which I'm assuming you're referrig to under "traditional SDLC processes) apply to data science projects? If so, how often? What sorts of projects?

@joe_caserta - totally agree, the epitome of "fail and fail fast" too ;-)

Just about to drop question 2 into the crowdchat stream. Look at the top of your screens for it.

@dez_blanchfield agile formalizes the process to make collaboration most effective.

- I'm sure Jen will agree that the heart and soul of Analytics stems from how Agile allows iteration of Models and Data Sources & input from the business / user / customer..

@dez_blanchfield The pre-chat poll response at the top of this page says no more than half of data science projects may use Agile. Are the rest "waterfall"? If so, is that "waterfall" percentage declining over time?

waterfall can be applied to #datascience projects that need to be integrated into into existing systems and therefore require extensive testing

jameskobielus6

http://www.via-cc.at...

[+] Show Hidden Comments

- There can be such a broad range of answers to that, I'm finding that MVP to the business, is an outcome that meets the initial desired outcome even if it's not entirely correct, or pretty, i.e. "show me the data" ;-)

Let's start on Question Three: Is it meaningful to speak of a data science "minimal viable product" in an Agile context?

- I've seen orgs happy with an MVP being a simple model that's done in Excel to prove the business case that then went on to be built into a tool that was used via webpage from the network

MVP can simply be a predictive model that’s more accurate than random guessing

2 Votes Vote

@dez_blanchfield That's interesting. Like app-dev in general, the desired outcome might be as minimal as "deliver a very finely focused insight derived from the data." Depending on what that insight is, it might be the springboard to more.

- absolutely, the whole premise of Agile in my mind ( and the book's definition ) is about reaching MPV in any way shape and or form by leveraging Agile methodologies and practices ;-)

Bob E. Hayes

If it provides an insight that is useful.

Or a ew predictive model that’s more accurate than the current model/method

Steve Ardire

IMO the MVP had better provide minimum viable insight or else it's useless

@dez_blanchfield Right. The MVP for "data science" might be something you can gain from other analytics tools, perhaps a spreadsheet. Adding ML or predictive analytics (and associated tools/data lakes/libraries) might extend that core insight

@joe_caserta - absolutely, the second you can "show" something that talks to the original question the business was asking, you have an MVP as they will instantly start "using" or "consuming" it ;-)

Moving on to Question Four. Look above.

@sardire - that's awfully Escher of you Steve ;-)

A company that collects proprietary data could build any number of MVPs ranging from new tech to display the data to a new model that predicts outcomes

@8pathsolutions - Jen how long do you think it usually takes to go from Question or Idea to an MVP for most projects, in your experience, days, weeks, months?

- indeed, and the newer tools which come with "out of the box" models and algo's where coding is optional and data ingest is key, you could get to MVP in hours rather than days or weeks or months even !!

@dez_blanchfield Right. Data science is traditionally an ongong exploration workstream without being tied to discrete app-dev workstream. The "product" of data science is insight. The "product" of app-dev may be code wrapped around that insight

@dez_blanchfield the amount of time you need to go from question/idea to MVP depends on whether you have the data, whether you need to preprocess/process the data, whether it requires a model, and how the output needs to displayed - could be days or years

@joe_caserta - ah yes, and accurate is key, I've seen far too many supposed MVP style "I'm done now" outcomes where someone likes what they saw but forgot to check "the math" and it caused real pain ;-(

@bobehayes - or simply speaks to whatever the user or customer or business was seeking in the first place right ;-)

@bobehayes - or simply speaks to whatever the user or customer or business was seeking in the first place right ;-)

@dez_blanchfield I've also seen companies deliver an MVP that is just a shell - essentially no model is built on the back end and the front end just pulls data...