Comments
Big Data & Your Customer Smarts
You must login to participate in this chat. Please login.

Anybody out there?

Prospector

It has been a while since the last post.

Prospector

This one was an eye-openner for me.

Prospector

Sentiment analysis was a new subject for me. From the examples provided, it seems to be an important element of big data analysis and it would be interesting to know companies that are offering such analysis.

Prospector

3. I didn't quite understand what you mean by "uncovering the long tail of model features" but I will look into the subject of large combinations, interactions of additional variables.

Prospector

2. In terms of buying behavior, it was really interesting to read that as few as 5% of relevant variables yield

95% of the sales lift. However, this important variables may not be included in the analysis if there are issues with validity.

Prospector

A couple of points from my side:

1. From professional experience, the target of validity is one of the hardest ones to achieve when it comes to big data analysis.

Prospector

Thank you Daniel for a very informative & interesting seminar.

Prospector

thank you for a good info 

Prospector

Thank you Daniel.

Prospector

Daniel,


Will you be presenting again in the future? Thanks.

Prospector

Good presentation. Thank you!

Prospector

Good info, thanks!

Prospector

Many thanks for the presentation and follow-up chat!

Prospector

Well, maybe just tired enough to keep the shift key down too long, Beth.

Blogger

@BEth what do you mean? We're never exhausted! 

Blogger

Thanks for the great lecture! Signing off now...

Blogger

Computational Marketing it is!

Blogger

Think Computational Marketing - it is the future!

Daniel -- interesting. Thanks. And it looks like we've exhausted our audience. Any closing thoughts before you sign off?

Blogger

Yes, those are more marketing terms I hear a lot from the domain experts I speak with in my Computational Marketing work, but I can translate it all down to machine learning results and techniques.

Hey everybody, we're closing in our hour mark and will be signing off momentarily. If anybody has a question for Daniel, now's the time to ask! 

Blogger

@Daniel, I've seen the terms uplift modeling or net-lift modeling? That's all about influencing customer behavior, right?

Blogger

sounds good..to encourage people to buy cofee for others in case of a future treat. That too is influencing customer behavior.

Blogger

@Beth, I think influencing positive thinking will lead to a number of benefits, including specific ones like you mentioned. If you can identify your high value customers, then special promotions for a TV would make a lot of sense.

@Beth yes, worth plugging!

Blogger

@kicheko hey,maybe they'll send you free coffee. That could make some people quite happy. Oh, I heard today that Starbucks was planning to give free coffees to people who buy for others.http://www.nbcnews.com/business/starbucks-free-coffee-promo-prods-lawmakers-come-together-8C11363331 They don't really expect it to work on a political level, but it will promote goodwill for Stabuck, regardless.

Blogger

Daniel -- is the goal just to influence positive thinking or is it to take positive action (like buying that $1500 TV or whatever)?

Blogger

I'll check that radio show out for sure!

@bethschultz-They shifted focus to product mix and optimization at store level. I mentioned this before.

Prospector

Ariella/Daniel -- I'll have to shamelessly plug the radio show we did just last week with Ben Hamner, director of engineering at Kaggle. He shared great tips on how to participate in and win a data science competition.  Tips for Participating in a Data Science Competition

 

Blogger

David,


Do you have any white papers or case studies associated with big data and customer loyalty programs?

Thanks.

Prospector

@bethschultz-Yes.

 

Prospector

Category 4 segment - sure it might generate a negative consequence, BUT you have to stay in contact with your customers. The difference is to use technology to "influence" them into thinking good of your product.

@Ariella, - Before i know it, coffee houses be will calling me to consult for coffee festivals and i could choose to play along :)

Blogger

@Daniel yes, kaggle is endlessly fascinating. New discoveries on that platform pop up all the time.

Blogger

@pkilaruaod -- I thought that might be the case, as we've seen some instances of that sort of decision-making as retailers look to deliver value pricing messages. Is it a grocery store?

Blogger

@Daniel, - I believe that even klout has enough data to produce a more accurate version of itself in the future. I'm willing to trust it will be more accurate with time and so will be any dependent algorithms

Blogger

Daniel, regarding customer segmentation and your category 4 (Reminding their subscription is up may trigger a re-evaluation of cost or quality) -- as much as that might trigger positive result it might also lead to negative consequence, no? 

 

Blogger

@bethschultz-Iforgot to mention they scrapped all loyalty cards.

Prospector

To get a sense for what types of problems are being solved with machine learning, please visit: www.kaggle.com which sponsors all kinds of machine learning contest, many for prize money. It is a fascinating look into the field of data science.

@kicheko LOL, a number of ppl post things about coffee regularly. Some even put it in their profiles.

Blogger

OK, thanks for explaining that. I wanted to be sure I had the term correct!

Blogger

Yes, "feature engineering" is the process of figuring out what data points are going to be predictive for the problem being solved using machine learning. It is a very important part of the process.

@kicheko I agree, facebook did well implementing that, but is it usual to use hashtags on facebook posts? Certainly will help us to analyze facebook data, or at least reduce the error/residual data that we normally don't use for the analysis.

Prospector

@ariella: Klout used to call me an expert in coffee.:) i have no idea where that came from...

Blogger

Daniel, during the lecture I think you used a term called "feature engineering." Did I hear that correctly?

Blogger

Yes, impromptu sentiment analysis should lead to a cummulative view of the customer over time.

I use Klout, but only inconjunction with other tools. Klout scores might be one feature variable for my algorithm.

@kicheko can't stand the whole idea behind klout and the way people would jump through hoops just to raise their scores. Lately, I haven't seen much about it, so I wonder if it's faded into oblivion.

Blogger

Daniel - so, just to be clear, the purpose of  impromptu sentiment analysis is to gather the data and feed it into the data warehouse for more advanced analysis at some point?

Blogger

@kicheko yes, it was to help gain that kind of search, but if people do keep their posts limited to friends, I believe that their posts will not come up to those who don't have permission to see them. I'm sure there is a way to get beyond that, but I believe that's the official word on their nod to privacy settings.

Blogger

@Daniel: there's a time the  tool called Klout became popular sort of measuring people's influence...maybe not the same as credibility but similar. Problem with Klout may have been premature implementation because it had lots of innacuracy.

Blogger

Oh yea, hastag analysis is definitely part of the sentiment equation!

Sentiment analysis should be historical in nature so as to avoid rash decision making. So you'd collect current data, but add it to the data warehouse for predictive analysis later on.

@ivofilipe: Thanks....the fact that you can tweet any random person on twitter is a powerful feature Twitter has over facebook. Though FB did itself well to introduce hashtags.

Blogger

They are focusing on product mix and optimization at store level. I have not seen their results yet. This retail chain is one of the early pioneers of loyalty cards.

Prospector

Yes, credibility analysis has not yet made it to the mainstream. It is a heavy area of research. That's why I like it so much, it is a differentiator.

Daniel, the idea of impromptu sentiment analysis you mentioned is interesting -- but I would fear in some cases it might lead to some reactive, rash, decision making. Is that a fair caution?

Blogger

Daniel, with only knowing what you've told us so far about credibility analysis, I would imagine it could really shake up social marketing measurements!

Blogger

I think that a locality-based focus can work in conjunction with a customer focus, since it is important consider local preferences, culture, etc.

@kicheko that's a good point! We use twitter to share what I like to call announcement data. Data we share to inform people, to be helpful to someone. There's a scientific paper called "Sadilek-Kautz-Silenzio_Modeling-Spread-of-Disease-from-Social-Interactions" where the data used was from twitter posts.

 

Prospector

@pkilaruaod -- interesting question. I would see that as a move from good to bad (at least from the customer perspective). Does it believe a focus on sales will ultimately lead to improved customer experience? 

Blogger

Capturing loyalty card engagment via POS is critical in helping to define the makeup of the loyalty customer. Items in basket, basket size, frequency of shopping, channel of shopping (in store vs online), etc. Thanks. Good ideas discussed about focusing on your 'best' customers and driving marketing strategy towards those folks!

Prospector

Credibility analysis is the darling of machine learning research. Google this term to find the latest research papers on the subject. It is fascinating!!

 

@Ariella:  I appreciate your explanation

Prospector

While the trend has been on customer focus, recently one of the major retailer is shifted focus from customers. Instead they are focusing more on individual store sales. What are your thoughts on their strategy?

Prospector

Daniel, you mentioned "credibility analytics" in the lecture. Can you tell us just a bit more about that? Are you seeing credibility metrics worked into CLV modeling at this point -- or are we way too early for that?

Blogger

For a fee-based service, try www.datasift.com

 

@Daniel thanks for the link!

Blogger

Hi everybody. If you have a question for Daniel -- even if you've placed in on the board previously -- raise it here again! We'll try to get to as many as we can in our remaining time.

Blogger

Hello all, the link for the free sentiment analysis tool is: http://www.sentiment140.com/

Awesome presentation!  Great discussion on the best socila media sites also

Prospector

thanks for the session, very insightful!

Prospector

Thank you for the presentation

 

Prospector

Nicely done. Thank you Dan.

Prospector

@Beth @Daniel Thanks!

Blogger

Thanks good stuff!

Prospector

@MJ yes, there are those endorsements that seem to come out of nowhere. LI does push people to do that, and I suppose they expect you'd reciprocate. But I don't pay much attention to 2nd and 3rd level contacts. It's really only an issue if you want to establish a connection; harder to do with someone ony on a 3rd level connection.

Blogger

Good presentation

Prospector

Many thanks for the presentaion and chat!

Prospector

Thank you for great presentation.  Love the focus on the customer!

Statistician

Does anyone know the links to the free twitter feeds for sentiment analysis?

Prospector

@mikerman yes, I use it LI more for professional contacts.

Blogger

@bboshn1764:  Yep, I think so; and Linkedin more "work" oriented.

Prospector

@Ariella:  Its just that there is so much going on with the website.  Trying to figure out who is in your 1st circle, 2nd circle; peoplpe endorsing you for skills that you may or may not have,...just seems much ado about nothing.

Prospector

@ivofilipe: and if its not personal, then selling it to sentiment analyzers will not land you in court.

Blogger

@mikeman- good point

Prospector

ksuhr32502 I agree that small business is at a disadvantage.

Statistician

@Mikeman- Ariela may be right. More sentiment on FB and TW

Prospector

@bboshn1764:  Absolutely, a crossover between Facebook and Linkedin--but I think it's all about who and what you want to target, and the message that you want to send.  To my mind, Facebook for "general" contacts, and Linkedin for more professional.

Prospector

@kicheko yes, exactly!

Blogger

@Kicheko I agree with you, I think it depends of in which country you are. We Portugueses doesn't use twitter has a more personal social network, comparing with facebook of course.

Prospector

@Ariella: Have seen...have to say Omg..now everything makes sense! of course selling data is lucrative and this is what investors see.

Blogger

@BethSchultz, unfortunately, it's still underdeveloiped in terms of using it to make decisons, but that is starting to change

Prospector

Good proactive move by PayPal!  Wish more companies were like that.  It is getting better though.

Statistician

@michaeljackson confusing in what way? I don't think people post so much about their feelings on products and services there as they do on something like Twitter and FB.

Blogger

@Mikeman - LKin does get confussing at times

Prospector

@mj: i would say the reverse of what you have said...i find facebook to be personal and twitter more impersonal.

Blogger

@kicheko you're most welcome. It starts off by saying  that Twitter earned $47.5 million from selling off its data to a fast-growing group of companies that analyze the data for insights into news events and trends.

Blogger

@bboshn1764:  LinkedIn is a bit consfusing to me.

Prospector

Social networking makes it hard for the small business because they now have 4 different avenues that must be checked for orders and trouble shooting than just email or a simple phone call. Customers want an immediate response but if you do not have the people monitoring all this it will not happen for the customer and then the customer gets angry.

Prospector

@Beth: although customer sentiement issues and viral blow-ups may not be a large impact to the bottom line, it could definitely reduce credibility, and in turn, customer loyalty over time.

Prospector

@Ariella: Thanks for the link.

Blogger

Credibility analysis -- I like that idea. I think it could really change how we view social commentary!

Blogger

@Mikerman - appears the same may be on both

Prospector

@Ariella: Even investors seems to see this power. they don't seem to be asking too many business model related questions ahead of the IPO...seems good.

Blogger

Sentiment analysis is especially critical for new product launches.  It often allows early corrective action before too much damage can be done.

Statistician

Here's the link http://online.wsj.com/news/articles/SB10001424052702304441404579118531954483974 It h as this great quote: "Businesses have spent years understanding the past," says Mr. Spivack. "Twitter has found a way to measure the present."

Blogger

I'm curious to know how the Internet of Things will extend the power to learn the customer. Will be tremendous that..

Blogger

@kicheko:  Twitter seems a bit more personal....Facebook is a tad impersonal (in my opinion)

Prospector

United/guitar -- the classic example!

Blogger

I rarely look at facebook for feedback and even less with linked in.

Prospector

@bboshn1764:  But very different audiences between Linkedin and Facebook ...

Prospector

Does sentiment matter? Thoughts?

Blogger

Twitter maybe because it's quickly accessable

Prospector

@kicheko there just was a WSJ article about data mining in Twitter.

Blogger

I feel twitter is stronger since you can send without having to login anything. It is quick as well.

Prospector

@bbo @ mj: I think twitter is stronger...a lot of that because its brief and also because its mostly open

Blogger

Does Linkedin get more results than facebook?

Prospector

@fcotigno -- so is the business making decisions based on what you learn via the CLV with social metrics, or testing it out yet?

Blogger

Facebook or Twitter??? Which is best?

Prospector

Does one media appear to have more impact than another

Prospector

@jdavis87 yes, it's not just a loss now but down the road.

Blogger

Which seems to be the best

 

Prospector

(Ah, chat is back, now!   :)   )

Prospector

@BethSchultz, Yes, but difficult to line up to sales

Prospector

Very impactful point, 1 poor CX experience equated to 1M potential lost sales opportunities, even assuming only 1% of those followers were going to purchase new product in next 3-5 years that's 10k+ lost opportunities.

Prospector

@bboshn1764 you mean because the insight is more reliable on it for big data?

Blogger

Blogging is a great way to attract customers and potential stakeholders

Prospector

@bboshn1764 -- I would think the answer depends on where your customers are!

Blogger

@bboshn1764: I think so.

Prospector

Of those of you who do CLV measures, are you including social metrics at this point? Why or why not?

Blogger

Is one social media better than another

Prospector

(Comments now seem paused for me--maybe this will kick them to start up again ...)

Prospector

Nothing as bad as mishandling a customer with a problem in the days of twitter

Blogger

Love the story about social media influence...so true.

Statistician

@kicheko yes, though some bonuses only are given out in January 

Blogger

(Comments now seem paused for me--arggh.  Maybe this will kick them to start up again ...)

Prospector

@Beth: self-fulfilling prophecy- Nice one. Got it!

Prospector

@aru_sas -- the archive is available immediately after this program ends. 

Blogger

@Ariella, - End of year especially sounds like a good time to sell cars..holidays and all

Blogger

@fcotigno -- thanks. ;-)

 

Blogger

The live streaming is being blocked by my company's firewall. When will the archive be available

Prospector

Hospice, I think there's more to learn via big-data beyond what business strategy dictates -- otherwise, you could create a self-fulfilling prophecy

Blogger

@kicheko true, and supposedly luxury cars sell a lot more at bonus time

Blogger

Feature engineering?

 

Prospector

@BethSchultz http://www.marketingscience.info/ 

Prospector

This segmentation is harder to apply for brand new technologies I think.  Many times people will say they aren't interested when they really just don't understand the benefits yet.  Marketer has to do their job to educate the customer to move them out of segment one to segment 2 or 3.

Statistician

I think Daniel said, "feature engineering?" Is that  term familiar to those of you doing machine learning?

Blogger

@jmulders940 We live in OLAP world so we're kind of behind the curve from that perspective, but its good to take the time to pre-build the tables that will be utilizied most often for ad-hoc requests and C level needs. Data pulls are simple refreshes for 75% of my world's workload

Prospector

Determining your valued customers has more to do with business strategy that may not require big data to be achieved, right?

Prospector

I think i have a good example of how timing is a key item on how the science works(slide 10). You sell more on payday:)

Blogger

usually people make a point of skipping 13 rather than 12 

Blogger

@fcotigno -- I'll have to look up Ehrenberg-Bass. Thanks!

Blogger

@Beth It's an example of "I didn't really say everything I said." 

Prospector

@jdavis87, I'd say 50% is a conservative estimate

Prospector

See Ehrenberg-Bass about customer retention--customer churn is to be expected

 

Prospector

@jdavis87 -- sounds like your company is a step ahead! Kudos for attention to quality data to start.

Blogger

@jmulder - Indeed. Important to focus on the "Smart" in place of the "Big"

Blogger

@mjulder940 of course, that's true.

Blogger

@jmulders940 -- absolutely!

Blogger

Although, to Daniel's point, focusing on 4 could have a negative impact if you're not careful!

Blogger

"use by" is completely dependent on the type/goal of the "current" analysis

Prospector

Everyone remember GIGO - garbage in, garbage out?  Seems like that holds true for data too

 

Prospector

I'd agree influencing behavior is a game-changer. Would you guys?

Blogger

@Beth: I'd say more 50-60% in my world, we try to save 1/2 workload time for QA and the analytics piece

Prospector

we may need a "use by" date stamp like on groceries. 

Blogger

@Hospice, please hit the left arrow --- there is audio.

Blogger

@Beth: I never thought of it so but yeah it is. Its in the way you market that you influence customer behavior and then measure the response...respond and return

Blogger

80 percent of the time spent cleaning the data and getting it ready for the analytics... does that jive with what you folks experience?

Blogger

@HH it's on, try a refresh

Blogger

@Beth yes, some of us have more low-brow frames of reference :D

Blogger

Veracity = Does it help solve for X? :)

Prospector

you can never have too many Vs

Blogger

(System working just fine now--issues gettig it running today ....)

Prospector

@kwsmithr -- classic Bohr is not so widely known as classic Yogi!

Blogger

Thanks for clarifyng "BI".  Everyone seems to have a different view of what it really encompasses.

Statistician

@kicheko -- isn't that the real goal? Not just understanding what your customers want but telling them what they want?

Blogger

I like that, "the big in big data."

Blogger

Do you think you can progress from BI to big-data analytics? Or there evolutionary steps along the way?

Blogger

Sounds good. I'm surprised that influencing customer behavior is also part of big-data

Blogger

"Computational marketing" -- does that resonate with you?

Blogger

Thanks kicheko -- I think Daniel is your guy.

Blogger

Slide Deck is located next to Mr. Gutierrez' picture

Prospector

@Beth: I am looking forward to learning how big data is becoming smart data...optimization of methods/cutting the unnecessary, saving costs etc

Blogger

That's the right attitude Michael!

Blogger

almost time to learn

Prospector

While we're waiting for the show to start, anybody care to share what you're most interesting in learning about big-data and your customer smarts?

Blogger

And hello to all from Chicago! If you're a first-time participant, please note that the audio player will appear at start time in the window above. Click on the left arrow to start (sometimes you have to click more than once).

 

Blogger

Hello from the DC Metro Area!

Prospector

Hi from Sunny Ohio!

Prospector

cant wait to hear what they have to say about Big Data!

Prospector

I just downloaded the slide deck, looks good.

 

Prospector

Hi all. 20 minutes to start time! If you haven't downloaded today's slide deck, please do so. You'll find it right above this window, next to Daniel's picture.

Blogger

Hello, I just log in, waiting for the broadcast to start.

 

Prospector

hello, I see there are earlier birds here. I'll break for lunch and then return in time for class.

Blogger

Hello from GA.

Wow! Today we have slide!

Very nice slide. Including lots of interesting cases.

I am really enjoying the free best practices taining on Data Governance

Prospector

Lookig forward to understanding the latest advancement with Big Data.

Prospector

I'll be there!

Prospector

Greeting in advance to all the attendees for my lecture "Big Data & Your Customer Smarts." I am looking forward to it.

hello to all

Prospector

should be interesting hope to attend

Prospector

Looking forward.

Prospector

If you aspire to become a Data Scientist, a graduate school program can boost your opportunities. Marquette University's PhD program in Computational Sciences is co-sponsoring an Open House scheduled on November 2, 2013 (see http://www.marquette.edu/chem/documents/Fallopenhouseflyerfinal.pdf for information). There are also MS degrees in Bioinformatics and Computing that feature courses on analytical methods. The Computing program includes an online option. Marquette offers classes to both degree-seeking and non-degree-seeking students. The programs are in the Mathematics, Statistics, and Computer Science Department so they offer the cross-discipline courses required for success.

Prospector


INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +