Marrone-Brandon & Analytics

[T master]

One of the big words used after Marrone was hired by both him & the Bills grand pooh bah Russ Brandon was analytics . They talked about using them & have started a new analytics department at one Bills drive .

 

I did graduate from high school so Russ & Marrone are much smarter than i so i had to look up exactly what the word meant which probably won't surprise many of you after reading a few of my posts !

 

The definition is as follows Analytics - Logic - the science of logical analysis .

If they were so big on introducing this to the Bills why wasn't it used in building the current coaching staff ?

 

There was no reason why they couldn't have opened the door for any of the other coaches to stay which could have meant players staying too .

 

While watching the play offs i saw more than one of the Bills old coaches from Gailey's coaching staff prowling the side lines of some of the play off teams such as in San diego - Joe Dealisondris - Carolina - Bruce Dehaven & if any of you can add to this list please do !

 

I know that when a new coach comes in they want there guys with them but last year our O line was as good as it's been in years ! Good enough for CJ to average 6 YPC & allow a very low amount of sacks, & if Bruce Dehaven isn't better than Crossman then i'm going to put in my application for the ST coach .

 

I didn't really care for the coaching choice but we have him now & i think he is a very passionate HC & has the team headed in the right direction . I also hope that he proves me totally wrong in what i thought of him being chosen ! But why didn't they use their brain child of installing this analytics thing while putting together his now staff ?

 

It just makes little to no sense what so ever to be so gung hoe about something then don't use it immediately because he had all the previous years stats & how the Bills did . Not to mention it would have been less change as far as blocking schemes & would have gotten the offense up to speed much quicker !!

 

Well it's all yours Bills fans what do you think ? If this has been discussed sorry but i must have missed it !!

Edited January 13, 2014 by T master

[stuckinny]

it is alot easier math when you start at 0! No where but up

[T master]

it is alot easier math when you start at 0! No where but up

 

Ya but they stayed the same in the win column and at the end of the day thats what counts !

 

Why wouldn't they have used analytics to look at the guys they wanted & the guy that were there I believe Metzelars, Dehaven, Dealisondris would have all stayed if given the option !!

[OCinBuffalo]

As a guy who's been getting paid to do "analytics" for quite some time, I believe I can answer your question.

 

First of all, analytics is not a light switch. You don't just flip it, and suddenly you are doing it.

Second? What the hell do diapers have to do with beer? Quite a lot actually, and we know this because of analytics. I always use this real world outcome of analytics to explain it.

Third: here's your requirement

use analytics to look at the guys they wanted & the guy that were there I believe Metzelars, Dehaven, Dealisondris would have all stayed if given the option !!

Wrong. Sorry, but wrong.

 

You don't approach analytics in terms of looking for the answers to your questions, or what "I believe". Clients constantly struggle with this. Yes, if you go looking for an answer, you'll probably find it: because your methodology was designed to find it. This is doing it wrong.

 

Doing it right = You let questions, not answers, come to you. Questions, such as "Hey! There seems to be a connection between lowering the price of beer, and increased diaper sales, I wonder why?", is where the value of analytics becomes clear.

 

If you started by saying "I think lowering the price of beer will increase beer sales", and go looking into the universe(fancy analytics word) you've created for this purpose, then yeah...somewhere, someplace, you'll find confirmation. But, more often than not, you've inherently biased the entire process towards your predetermined conclusion, and you've also missed a lot(diapers).

 

Instead, the job is to create a universe WITHOUT a specific...something...you are trying to prove. Thus, the weird things, like diaper sales being a function of lowering beer prices, jump out at you.

 

See, if you were just looking to measure beer prices, because you already "know" the "answer"....then you'd never have included diaper sales data in your universe. In this case of NFL coaches, if we just look at football coaches in terms of the play at their players positions, and build a universe on that?

 

I'm sure you can prove what you are looking for too, but it's doubtful it's right...or at the very least, it's doubtful that we have the complete picture.

 

Thus, building a proper universe should take time. You don't want everything you have, or maybe you do. Figuring our what is relevant, on a macro scale...is why you hire guys like me, who've done this before. We're objective. We aren't looking to prove anything other than: we know how to do it right.

 

So finally, what the hell does beer have to do with diapers?

 

Sale of diapers, when purchased as part of a small order(5 or less items sold) that also includes beer, increased when beer prices were lowered. The reason this was found? The sale of ALL items was looked at relative to beer prices(or prices of ALL items), for small orders, medium orders and big orders, and diapers jumped to the top of that list in terms of significant increase. In order to find out why, the customer's credit card info was added to cube(another fancy analytics word). Then, it was determined that these small orders were bought by: men.

 

Who runs out to the store for diapers? Who is most likely to buy beer? Who makes buying decisions based on beer prices? Who doesn't know/care about diapers in general, never mind their prices? See? More questions, not answers. But, once you answer the questions with the right data?

 

Boom. You want to sell more high margin diapers? Cut your beer prices, and you will, because men don't care about diapers, or their price, but they do care about beer, and its price.

 

Get it?

 

If we didn't build the data warehouse properly, to allow for these kind of cross-product analyses, or the ability to add dimensions, like the credit card data, if we didn't allow for what we call "drill across": this diapers/beer thing never happens.

 

That's why analytics takes time, and this is why it's not a light switch.

 

IF you want to use analytics to determine which coach to keep, and which to fire?

 

You have a hell of a lot of work to do.

Edited January 14, 2014 by OCinBuffalo

[Dibs]

.....

 

IF you want to use analytics to determine which coach to keep, and which to fire?

 

You have a hell of a lot of work to do.

 

Great post!

 

IMO due to the limited data pool, the immense number of variables, and the unpredictable nature of the individual(players, coaches, team unity etc).......trying to use analytics to effectively determine coach appointments or retention is nigh-impossible(Spoon!).

Edited January 14, 2014 by Dibs

[OCinBuffalo]

Great post!

 

IMO due to the limited data pool, the immense number of variables, and the unpredictable nature of the individual(players, coaches, team unity etc).......trying to use analytics to effectively determine coach appointments or retention is nigh-impossible(Spoon!).

I could do it.

 

The question is: who is paying me, and can they afford our rates?

[3rdand12]

As a guy who's been getting paid to do "analytics" for quite some time, I believe I can answer your question.

 

First of all, analytics is not a light switch. You don't just flip it, and suddenly you are doing it.

Second? What the hell do diapers have to do with beer? Quite a lot actually, and we know this because of analytics. I always use this real world outcome of analytics to explain it.

Third: here's your requirement

 

Wrong. Sorry, but wrong.

 

You don't approach analytics in terms of looking for the answers to your questions, or what "I believe". Clients constantly struggle with this. Yes, if you go looking for an answer, you'll probably find it: because your methodology was designed to find it. This is doing it wrong.

 

Doing it right = You let questions, not answers, come to you. Questions, such as "Hey! There seems to be a connection between lowering the price of beer, and increased diaper sales, I wonder why?", is where the value of analytics becomes clear.

 

If you started by saying "I think lowering the price of beer will increase beer sales", and go looking into the universe(fancy analytics word) you've created for this purpose, then yeah...somewhere, someplace, you'll find confirmation. But, more often than not, you've inherently biased the entire process towards your predetermined conclusion, and you've also missed a lot(diapers).

 

Instead, the job is to create a universe WITHOUT a specific...something...you are trying to prove. Thus, the weird things, like diaper sales being a function of lowering beer prices, jump out at you.

 

See, if you were just looking to measure beer prices, because you already "know" the "answer"....then you'd never have included diaper sales data in your universe. In this case of NFL coaches, if we just look at football coaches in terms of the play at their players positions, and build a universe on that?

 

I'm sure you can prove what you are looking for too, but it's doubtful it's right...or at the very least, it's doubtful that we have the complete picture.

 

Thus, building a proper universe should take time. You don't want everything you have, or maybe you do. Figuring our what is relevant, on a macro scale...is why you hire guys like me, who've done this before. We're objective. We aren't looking to prove anything other than: we know how to do it right.

 

So finally, what the hell does beer have to do with diapers?

 

Sale of diapers, when purchased as part of a small order(5 or less items sold) that also includes beer, increased when beer prices were lowered. The reason this was found? The sale of ALL items was looked at relative to beer prices(or prices of ALL items), for small orders, medium orders and big orders, and diapers jumped to the top of that list in terms of significant increase. In order to find out why, the customer's credit card info was added to cube(another fancy analytics word). Then, it was determined that these small orders were bought by: men.

 

Who runs out to the store for diapers? Who is most likely to buy beer? Who makes buying decisions based on beer prices? Who doesn't know/care about diapers in general, never mind their prices? See? More questions, not answers. But, once you answer the questions with the right data?

 

Boom. You want to sell more high margin diapers? Cut your beer prices, and you will, because men don't care about diapers, or their price, but they do care about beer, and its price.

 

Get it?

 

If we didn't build the data warehouse properly, to allow for these kind of cross-product analyses, or the ability to add dimensions, like the credit card data, if we didn't allow for what we call "drill across": this diapers/beer thing never happens.

 

That's why analytics takes time, and this is why it's not a light switch.

 

IF you want to use analytics to determine which coach to keep, and which to fire?

 

You have a hell of a lot of work to do.

I read the whole thing.

Wheres my certificate ?

: )

if you take out some of the big words and long sentences i understood what you are impressing upon us.

Well done to.

 

 

 

Ya but they stayed the same in the win column and at the end of the day thats what counts !

 

Why wouldn't they have used analytics to look at the guys they wanted & the guy that were there I believe Metzelars, Dehaven, Dealisondris would have all stayed if given the option !!

I think its one of those things thats going to take a while to get dialed in.

I think its the future current future though.

Just something the team needs to get skilled at , and then make sound decisions. There is the rub .

I miss Joe D

Edited January 14, 2014 by 3rdand12

[DC Grid]

Analytics in the NFL is the new hotness I get that. Putting aside questions about whether or not advanced analytics would really make a difference for the Bills, the Jags provide proof of how it can fail, especially when implementing it fast from scratch. According to their awesome analytics department that they cobbled together then immediately relied upon last offseason, Gabbert was the 12th best QB in the NFL.

[section122]

I'm not going to quote the entire post but ocinbuffalo that is one of the best posts I have ever read on this site!

[OCinBuffalo]

I read the whole thing.

Wheres my certificate ?

: )

if you take out some of the big words and long sentences i understood what you are impressing upon us.

Well done to.

Fresh out of certificates.

 

But, if you want to buy one...

 

And look at it this way: as long as I'm not oppressing you...

I'm not going to quote the entire post but ocinbuffalo that is one of the best posts I have ever read on this site!

Thanks. Not bad for 3 drinks into it. However, after reading it again now? I forgot to explain some stuff....

 

but that would have made it looooooonger.

Analytics in the NFL is the new hotness I get that. Putting aside questions about whether or not advanced analytics would really make a difference for the Bills, the Jags provide proof of how it can fail, especially when implementing it fast from scratch. According to their awesome analytics department that they cobbled together then immediately relied upon last offseason, Gabbert was the 12th best QB in the NFL.

I sincerely doubt most teams have devoted the resources required to do this properly...yet. This isn't necessarily a bad thing. And, yeah: attempting to Big Bang analytics into being is usually an exercise in misery. You need to do it in stages, which is why the Bills not having a fully functioning analytics department overnight is: good.

 

In contrast: It appears the Jags did exactly what I am talking about: "let's prove that Gabbert isn't that bad", and, threw together a quick universe that never had a chance of being right.

 

Right way:

Warehouse any and all data that has even the slightest chance of describing or influencing QB play. Then, start playing around with that data, and see what emerges regarding Gabbert. That should take about 18 months to do right, with a small team.

 

Also: I don't see any reason why the league doesn't create some common Big Data for teams to use. Same reasoning for this as there is for the Combine. There's little point in each team creating their own, because for some of it(I'd guess 40%), ALL teams will need the same exact data.

Edited January 14, 2014 by OCinBuffalo

[T master]

As a guy who's been getting paid to do "analytics" for quite some time, I believe I can answer your question.

 

First of all, analytics is not a light switch. You don't just flip it, and suddenly you are doing it.

Second? What the hell do diapers have to do with beer? Quite a lot actually, and we know this because of analytics. I always use this real world outcome of analytics to explain it.

Third: here's your requirement

 

Wrong. Sorry, but wrong.

 

You don't approach analytics in terms of looking for the answers to your questions, or what "I believe". Clients constantly struggle with this. Yes, if you go looking for an answer, you'll probably find it: because your methodology was designed to find it. This is doing it wrong.

 

Doing it right = You let questions, not answers, come to you. Questions, such as "Hey! There seems to be a connection between lowering the price of beer, and increased diaper sales, I wonder why?", is where the value of analytics becomes clear.

 

If you started by saying "I think lowering the price of beer will increase beer sales", and go looking into the universe(fancy analytics word) you've created for this purpose, then yeah...somewhere, someplace, you'll find confirmation. But, more often than not, you've inherently biased the entire process towards your predetermined conclusion, and you've also missed a lot(diapers).

 

Instead, the job is to create a universe WITHOUT a specific...something...you are trying to prove. Thus, the weird things, like diaper sales being a function of lowering beer prices, jump out at you.

 

See, if you were just looking to measure beer prices, because you already "know" the "answer"....then you'd never have included diaper sales data in your universe. In this case of NFL coaches, if we just look at football coaches in terms of the play at their players positions, and build a universe on that?

 

I'm sure you can prove what you are looking for too, but it's doubtful it's right...or at the very least, it's doubtful that we have the complete picture.

 

Thus, building a proper universe should take time. You don't want everything you have, or maybe you do. Figuring our what is relevant, on a macro scale...is why you hire guys like me, who've done this before. We're objective. We aren't looking to prove anything other than: we know how to do it right.

 

So finally, what the hell does beer have to do with diapers?

 

Sale of diapers, when purchased as part of a small order(5 or less items sold) that also includes beer, increased when beer prices were lowered. The reason this was found? The sale of ALL items was looked at relative to beer prices(or prices of ALL items), for small orders, medium orders and big orders, and diapers jumped to the top of that list in terms of significant increase. In order to find out why, the customer's credit card info was added to cube(another fancy analytics word). Then, it was determined that these small orders were bought by: men.

 

Who runs out to the store for diapers? Who is most likely to buy beer? Who makes buying decisions based on beer prices? Who doesn't know/care about diapers in general, never mind their prices? See? More questions, not answers. But, once you answer the questions with the right data?

 

Boom. You want to sell more high margin diapers? Cut your beer prices, and you will, because men don't care about diapers, or their price, but they do care about beer, and its price.

 

Get it?

 

If we didn't build the data warehouse properly, to allow for these kind of cross-product analyses, or the ability to add dimensions, like the credit card data, if we didn't allow for what we call "drill across": this diapers/beer thing never happens.

 

That's why analytics takes time, and this is why it's not a light switch.

 

IF you want to use analytics to determine which coach to keep, and which to fire?

 

You have a hell of a lot of work to do.

 

WOW in my little pea brain this analytics thing is made way to illogical i guess for the lack of a better term with the universe thing !

 

It sounds as though there is a lack of common sense in the thing or as it is apparent i'm missing something but thanks for the explanation

I think ??

[BuffaloBill]

 

 

Great post!

 

IMO due to the limited data pool, the immense number of variables, and the unpredictable nature of the individual(players, coaches, team unity etc).......trying to use analytics to effectively determine coach appointments or retention is nigh-impossible(Spoon!).

 

Determining which variables matter is what analytics is all about. The science generally shows that popular belief or "gut feel" is often incomplete if not outright inaccurate. That is the beauty of applying the science.

[Over 29 years of fanhood]

Analytics is a new word for an old practice of database analysis. Difference is now with computers' processing power the types analyses and the amount of data you can crunch to find patterns is vast.

 

I am not convinced the Bills have effectively built robust Analytics as Russ had prioritized.

 

Marrone like to use the word occasionally then talks about isolated stats like third down conversion rates, so I'm not sure they aren't misusing the term a bit.

 

I guess any application of math to football might seem groundbreaking....

 

 

[zonabb]

OCinBuffalo good post. I took my fair share of stats in my to get my phd and I use them sparingly... mainly regression and spatial, so I have a grasp of statistical modeling and prediction. Anyway I've been so annoyed with the general discourse of analytics because frankly, no one really knows what it is, can't explain it, couldn't apply it, etc. Good laymen's outline.

 

It pains me to hear WGR55 people screaming for always going for it on 4 and 1 because "analytics" say so. No, probability says so... Unless you can explain how down and distance or location on the field change or some other variable impacts the probability and therefore the decision. But a blanklet "always for go for it on 4 and 1" that hold all else equal is not analytics. Quoting someone else's analytics makes you seem smart I guess but I'd be impressed if someone in the local media actual could DO IT rather than regurgitate it.

 

Everyone here would be wise, as a follow up to the "beer and diapers" example, read Steven Leavitt's "Freakonomics." Although no football examples (there is a sumo one), it provides some good real world example of the power of statistics and using big data and algorithms.

 

I certainly wasn't going to tackle writing the response you did, although 3 drinks probably helped! Oh, and I've always wanted to develop a model on whether or not QB draft position and college attributes have any bearing at all NFL success...assuming of course everyone QB played on the same NFL team under the same offence with the same coached and same teammates : ) Then of course one adding in team and coach specific variables (whatever those are!).

 

Anyway, that's enough stats consideration for me for the day.

[Over 29 years of fanhood]

 

Everyone here would be wise, as a follow up to the "beer and diapers" example, read Steven Leavitt's "Freakonomics."

 

Fantastic book and easy read, but it was Dubner too(don't neglect the coauthor man!)

 

Did anyone read the follow up book?

 

I always assume when Russ was talking Analytics he envisioned an application akin to the movie moneyball

[Dibs]

Determining which variables matter is what analytics is all about. The science generally shows that popular belief or "gut feel" is often incomplete if not outright inaccurate. That is the beauty of applying the science.

 

I totally agree.......and as it happens I spend a fair bit of time providing simplistic data analysis on this site to counteract the oft posted "popular beliefs" or "gut feel" opinions that many others espouse.

 

My point was that there are many areas relating to football(or more accurately humans) that no amount of analytics can determine. Some may disagree with this.....I just hope that the Bills analytics department aren't amongst those.

 

Don't get me wrong though......I think analytic questions should be asked in all areas & data subsequently collected. I just don't think it can be useful in all of the ways that some people seem to think it can.

[Over 29 years of fanhood]

 

 

Great post!

 

IMO due to the limited data pool, the immense number of variables, and the unpredictable nature of the individual(players, coaches, team unity etc).......trying to use analytics to effectively determine coach appointments or retention is nigh-impossible(Spoon!).

 

Appointments perhaps not, but certainly there is an application in assessing performance and improvement.

 

As a six sigma black belt I believe you can Quantify the universe. If one thinks an outcome is immeasurable or without correlating metrics, I think one has yet to measure the right things, hasn't tested variable combinations correctly or is missing an interaction. It's usually one or more.

 

 

I do wonder how/if bills coaches today use data to make decisions, assign depth charts, call plays and formations situationally, etc.

 

When the gm talks about it I assume he's determining how to assess which players and positions to throw money at and which are more commodity skill set not worth paying premiums for.

 

 

[CodeMonkey]

I do wonder how/if bills coaches today use data to make decisions, assign depth charts, call plays and formations situationally, etc.

 

assign depth charts - Maybe

call plays and formations situationally - Probably not

 

When the gm talks about it I assume he's determining how to assess which players and positions to throw money at and which are more commodity skill set not worth paying premiums for.

This is what I believe he is talking about as well. And include who to draft (determine BPA for example) as well.

[OCinBuffalo]

Analytics is a new word for an old practice of database analysis.

Boy you can say that again: In college? Operations Research/Management(2 classes). In my rookie Big 6 days? Data Warehousing. In my Valley time? Business Intelligence. For the last few years: Analytics.

 

We'll see if Big Data supplants Analytics...but yeah, it's all just marketing terms, with the only real change being: how many processors you can chain together.

 

The other thing that doesn't change? Look at the end of this post.

OCinBuffalo good post. I took my fair share of stats in my to get my phd and I use them sparingly... mainly regression and spatial, so I have a grasp of statistical modeling and prediction. Anyway I've been so annoyed with the general discourse of analytics because frankly, no one really knows what it is, can't explain it, couldn't apply it, etc. Good laymen's outline.

 

It pains me to hear WGR55 people screaming for always going for it on 4 and 1 because "analytics" say so. No, probability says so... Unless you can explain how down and distance or location on the field change or some other variable impacts the probability and therefore the decision. But a blanklet "always for go for it on 4 and 1" that hold all else equal is not analytics. Quoting someone else's analytics makes you seem smart I guess but I'd be impressed if someone in the local media actual could DO IT rather than regurgitate it.

 

Everyone here would be wise, as a follow up to the "beer and diapers" example, read Steven Leavitt's "Freakonomics." Although no football examples (there is a sumo one), it provides some good real world example of the power of statistics and using big data and algorithms.

 

I certainly wasn't going to tackle writing the response you did, although 3 drinks probably helped! Oh, and I've always wanted to develop a model on whether or not QB draft position and college attributes have any bearing at all NFL success...assuming of course everyone QB played on the same NFL team under the same offence with the same coached and same teammates : ) Then of course one adding in team and coach specific variables (whatever those are!).

 

Anyway, that's enough stats consideration for me for the day.

Everything can be quantified. There's an illusion in this world that "my job is different, becasue it all depends on my decisions/knowledge, and I'm so good/what I do is so different, that"....horsecrap.

Fantastic book and easy read, but it was Dubner too(don't neglect the coauthor man!)

 

Did anyone read the follow up book?

 

I always assume when Russ was talking Analytics he envisioned an application akin to the movie moneyball

Yep, the diapers/beer thing is what I've chosen to use...because it has a sort of humanizing effect, and people can relate to it.

 

Yes, Moneyball is interesting, but, there's been a huge debate going on for years about the specific methods there, and whether those methods are riddled with confidence bias. I've devoted very little time to it, but it seems there are good arguments on both sides.

I totally agree.......and as it happens I spend a fair bit of time providing simplistic data analysis on this site to counteract the oft posted "popular beliefs" or "gut feel" opinions that many others espouse.

 

My point was that there are many areas relating to football(or more accurately humans) that no amount of analytics can determine. Some may disagree with this.....I just hope that the Bills analytics department aren't amongst those.

 

Don't get me wrong though......I think analytic questions should be asked in all areas & data subsequently collected. I just don't think it can be useful in all of the ways that some people seem to think it can.

It depends, as it always does: on not only your defnition of useful, but the client's. In this case, what if I can in fact quantify something, but few others can understand/relate to it? Even though my work is excellent: it doesn't matter.

Appointments perhaps not, but certainly there is an application in assessing performance and improvement.

 

As a six sigma black belt I believe you can Quantify the universe. If one thinks an outcome is immeasurable or without correlating metrics, I think one has yet to measure the right things, hasn't tested variable combinations correctly or is missing an interaction. It's usually one or more.

 

I do wonder how/if bills coaches today use data to make decisions, assign depth charts, call plays and formations situationally, etc.

 

When the gm talks about it I assume he's determining how to assess which players and positions to throw money at and which are more commodity skill set not worth paying premiums for.

Ah, a kindred spirit. As I said above, most people think their jobs are special. Answer: nope. Most people therefore think what they do is specialized and custom to them, yet, if you ask them what makes it good, they immediately talk about "industry standards", and how good they are at meeting them It's cognitive dissonance: something can't be specialized/customized, and standard, at the same time. (Well, if you understood what we did at my firm, you'd understand how we preserve this...illusion? delusion? but whatever)

 

Why should we expect football to be any different, or the delusions to be anything but worse? When a coach talks about "their way"...but then says the, apparently magic, words "get off the ball!"...we have our conundrum. How can we all talk about "getting off the ball", yet at the same time make any claim to specialized coaching instruction? Evaluation? If we all agree what "getting off the ball" looks like, then how can we claim to evaluate it...differently? Thus, this becomes a question of competence: you either coach/evaulate properly, each time you do it, or you don't. True/False is very easy to quantify.

 

No. The fact is, as PFF has shown, all it takes is: proper, consistent measurement techniques. Say what you want about "not knowing the play", that argument fails when PFF evaluates every player, on every play, using an objective and standard evaluation instrument. The large # of plays ensures that not knowing the play: barely matters. And, even if it does? The standardization means: we always fail in the same way. Thus, when comparing results one player to another? Any failure is consistent, and therefore: irrelevant.

 

Next: be careful about being truthful abot your life experiences. Apparently there are insecure people patrolling the board, and they can't handle it. Six Sigma is interesting. Of all of the methodologies in management consulting, this one has the best chance to actually stand the test of time. We'll see. I start fires for cookouts with most of the other business books I read...never run out of kindling.

assign depth charts - Maybe

call plays and formations situationally - Probably not

This is what I believe he is talking about as well. And include who to draft (determine BPA for example) as well.

We'll see...

 

The biggest problem I see? The same problem I've seen everywhere else = you can deliver wonderful intelligence, but, if the client can't understand it/make use of/is convinced by it in the proper time frame?

 

It doesn't matter.

 

This is not to call other peole stupid. (Well, it is to call some people... ). The problems we've found in this field are pretty consistent: only the top 20% of client staff has the grey matter to actually use these systems effectively. This is especially true when we add the time dimension: as in, here's the answer, now, how long until s/he realizes it, and can make a decision based on it? If that window of time passes, you've accomplished nothing. Again, it may just be a matter of how you've defined your stuff. It may be as simple as how you've laid out the User Interface. It can be a lot of things. EDIT: Let me be clear, IQ(or EQ, or 7 levels of intelligence...something...rears its head here) cuts both ways: if the analytics consultant isn't able to listent to client, and see HOW to lay it out, you're just as screwed.

 

This isn't an experience thing either. This is purely about IQ, and to a lesser degree, confidence of the manager in his/her abilities, in question. Age has nothing to do with it. Neither does title. For every young ass-kicker I've seen use analytics well, I've seen a young dolt not get past the concept of "over time, not point in time". Same thing with CEOs.

 

I don't see this part changing at all in football.

 

So, analytics will only be useful to some. The real question then, is also always the same: can the organization put aside it's inherent power structure temporarily, such that when those who see the intelligence staring back at them, they are able to act on it? Or, will those with inferior ability, but, a title that says they can: block whatever might be done, because "they just don't see it"?

Edited January 17, 2014 by OCinBuffalo

[shibuya]

As a guy who's been getting paid to do "analytics" for quite some time, I believe I can answer your question.

 

First of all, analytics is not a light switch. You don't just flip it, and suddenly you are doing it.

Second? What the hell do diapers have to do with beer? Quite a lot actually, and we know this because of analytics. I always use this real world outcome of analytics to explain it.

Third: here's your requirement

 

Wrong. Sorry, but wrong.

 

You don't approach analytics in terms of looking for the answers to your questions, or what "I believe". Clients constantly struggle with this. Yes, if you go looking for an answer, you'll probably find it: because your methodology was designed to find it. This is doing it wrong.

 

Doing it right = You let questions, not answers, come to you. Questions, such as "Hey! There seems to be a connection between lowering the price of beer, and increased diaper sales, I wonder why?", is where the value of analytics becomes clear.

 

If you started by saying "I think lowering the price of beer will increase beer sales", and go looking into the universe(fancy analytics word) you've created for this purpose, then yeah...somewhere, someplace, you'll find confirmation. But, more often than not, you've inherently biased the entire process towards your predetermined conclusion, and you've also missed a lot(diapers).

 

Instead, the job is to create a universe WITHOUT a specific...something...you are trying to prove. Thus, the weird things, like diaper sales being a function of lowering beer prices, jump out at you.

 

See, if you were just looking to measure beer prices, because you already "know" the "answer"....then you'd never have included diaper sales data in your universe. In this case of NFL coaches, if we just look at football coaches in terms of the play at their players positions, and build a universe on that?

 

I'm sure you can prove what you are looking for too, but it's doubtful it's right...or at the very least, it's doubtful that we have the complete picture.

 

Thus, building a proper universe should take time. You don't want everything you have, or maybe you do. Figuring our what is relevant, on a macro scale...is why you hire guys like me, who've done this before. We're objective. We aren't looking to prove anything other than: we know how to do it right.

 

So finally, what the hell does beer have to do with diapers?

 

Sale of diapers, when purchased as part of a small order(5 or less items sold) that also includes beer, increased when beer prices were lowered. The reason this was found? The sale of ALL items was looked at relative to beer prices(or prices of ALL items), for small orders, medium orders and big orders, and diapers jumped to the top of that list in terms of significant increase. In order to find out why, the customer's credit card info was added to cube(another fancy analytics word). Then, it was determined that these small orders were bought by: men.

 

Who runs out to the store for diapers? Who is most likely to buy beer? Who makes buying decisions based on beer prices? Who doesn't know/care about diapers in general, never mind their prices? See? More questions, not answers. But, once you answer the questions with the right data?

 

Boom. You want to sell more high margin diapers? Cut your beer prices, and you will, because men don't care about diapers, or their price, but they do care about beer, and its price.

 

Get it?

 

If we didn't build the data warehouse properly, to allow for these kind of cross-product analyses, or the ability to add dimensions, like the credit card data, if we didn't allow for what we call "drill across": this diapers/beer thing never happens.

 

That's why analytics takes time, and this is why it's not a light switch.

 

IF you want to use analytics to determine which coach to keep, and which to fire?

 

You have a hell of a lot of work to do.

 

 

Awesome read