I hate analytics gurus



  • We just had a meeting with a data analytics specialist (you know, the "R" programming language type). In this case, he was hired to figure out all the reports for an app we are making.

    The guru showed up in a stained shirt. He didn't bring his laptop, so we had to load his powerpoint with report specs from an email.

    He had initially marked each slide with a numeric designation, but as he was revising the presentation, the numbers got all jumbled up. He never bothered to fix this before submitting his work.

    We went through his powerpoint report by report. In the end, we ended up crossing off about half of them as either useless or too vague. A few times, we have also merged several different slides into one report, with an added filter. It's like the guru was trying to pad his numbers.

    Each report had some glaring issue he failed to address.

    Some representative exchanges:

    Me: Wait, if you have a timeline graph comparing average revenue per customer with total revenue, wouldn't that other line hug the axis as the number of customers surpasses like 10?
    Guru: Umm... I guess...
    Me: That would make the graph kind of useless, right?
    Guru: You can remove it if you don't need it, I guess.
    (wait, wasn't it your job to figure this shit out?)

    Or

    Me: Wait, you only added one analytics report. But on google analytics, we now have like 20.??
    Guru: That's just an example, you can add more when you get the data.
    (wait, wasn't it your job to figure this shit out?)

    Or

    Me: This report tracks the number of created, paused and completed orders per day. But wouldn't one order appear multiple times in each column? What if someone pauses and resumes order multiple times per day? Do we group everything by the order creation time or the status change time?
    Guru: Umm.... maybe by status time... or creation time... hm...
    (wait, wasn't it your job to figure this shit out?)

    Or

    Me: Wait, isn't this slide just a duplicate of the slide 1-3-5?
    Guru: Oh right! You can then remove the other one.
    (wait, wasn't it your job to figure this shit out?)

    Or

    Me: You only have charts in this presentation. Shouldn't we have tabular displays for some of these reports.
    Guru: Hmm, I guess that would be useful.
    (wait, wasn't it your job to figure this shit out?)

    Or

    Me: I thought you would include some sales funnel tracking. That's the lifeblood of the system, after all. But I see nothing here....?
    Guru: Good idea! Drop me an email and I'll see to add some of that.
    (wait, wasn't it your job to figure this shit out?)

    WTF do these data analytics people do all day?

    He didn't imagine any report that an average backend dev couldn't produce on their own. Boss clearly expected the guy to produce secret math juice. Instead we got half-complete pile of crap we can pick for ideas as we design our own reports.

    First UX and design, now data analytics... I have yet to meet a specialist that doesn't disappoint.


  • sockdevs

    @cartman82 said in I hate analytics gurus:

    WTF do these data analytics people do all day?

    Make shiny things that dazzle management types



  • @cartman82 "gurus"


  • Discourse touched me in a no-no place

    @cartman82 said in I hate analytics gurus:

    WTF do these data analytics people do all day?

    They write stuff on FB and Twitter about how data analytics is so great.



  • Sounds like your problem is crappy consultants, not their specialty. I've never directly worked with data analytics people but one company I chatted with at a conference seemed reasonably competent based on the types of questions they were asking me about what our company was doing at the time.


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    WTF do these data analytics people do all day?

    The guru showed up in a stained shirt.

    They eat leftovers in the common room.


    P.s. mobile compose sucks! How can i quote a line then another when compose would not just hover?



  • "Knowing R" and "being a good data analyst" are not the same skillset. I'd wager the vast majority of quality data analysts are not using R, because R sucks ass compared to, say, Tableau. Or even SAS.


  • Discourse touched me in a no-no place

    @blakeyrat 10 years ago, I'd have agreed with you. These days, R seems to be where it is at; it's what the training courses all seem to be on, and it's what my scientific colleagues all wanted to learn. They didn't want SAS even though we had all the licenses and everything…



  • @dkf said in I hate analytics gurus:

    my scientific colleagues all wanted to learn.

    Science people are miles away from web analytics or business analytics people.

    You're talking about a community that actually likes Latex for laying out documents. They're miles away from rationality when it comes to selecting software.

    That said, SAS has been going downhill for like a decade now, which is why Tableau is currently eating their lunch. It'll be around for decades yet because it's so firmly installed in so many places, though.


  • Discourse touched me in a no-no place

    @blakeyrat said in I hate analytics gurus:

    You're talking about a community that actually likes Latex for laying out documents.

    That varies very much between communities. The maths people adore it, but the biologists won't touch it, and use Word. The biologists love R. And Excel.


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    Me: Wait, if you have a timeline graph comparing average revenue per customer with total revenue, wouldn't that other line hug the axis as the number of customers surpasses like 10?
    Guru: Umm... I guess...
    Me: That would make the graph kind of useless, right?
    Guru: You can remove it if you don't need it, I guess.
    (wait, wasn't it your job to figure this shit out?)

    Which is why you use multiple Y-axes. The bad part is that the guru was so completely stumped by your question that he decided to toss it rather than explain how it should work.

    Sounds like he just copypastaed some stuff he saw elsewhere that looked pretty, and has no idea how any of it works.


  • :belt_onion:

    @blakeyrat said in I hate analytics gurus:

    because R sucks ass compared to, say, Tableau.

    mainly because they're... two different things?
    There's an R server plugin that let you feed data from Tableau to R to do deeper analytics and then feed back to tableau. Tableau is great at making things pretty and doing a few rudimentary metrics, but it is nowhere near as powerful as R.


  • sockdevs

    @dkf said in I hate analytics gurus:

    These days, R seems to be where it is at; it's what the training courses all seem to be on, and it's what my scientific colleagues all wanted to learn.

    It may just be me, but I'm getting a whiff of 'hipster hype' about that statement, as it doesn't sound all that different from the NoSQL 'revolution' and the soaring popularity of NodeJS.



  • @darkmatter said in I hate analytics gurus:

    Which is why you use multiple Y-axes. The bad part is that the guru was so completely stumped by your question that he decided to toss it rather than explain how it should work.
    Sounds like he just copypastaed some stuff he saw elsewhere that looked pretty, and has no idea how any of it works.

    Looking at his graph, he does seem to use multiple Y axes (left and right).

    But he failed to explain how overlaying these two graphs (avg revenue per customer vs total revenue) would be useful to anyone. And in which proportion should the axes be.



  • @darkmatter said in I hate analytics gurus:

    There's an R server plugin that let you feed data from Tableau to R to do deeper analytics and then feed back to tableau. Tableau is great at making things pretty and doing a few rudimentary metrics, but it is nowhere near as powerful as R.

    First time I heard of Tableau.

    Looks fancy on their video.



  • @mikehurley said in I hate analytics gurus:

    Sounds like your problem is crappy consultants, not their specialty. I've never directly worked with data analytics people but one company I chatted with at a conference seemed reasonably competent based on the types of questions they were asking me about what our company was doing at the time.

    Oh I am sure there are competent data analysts and many areas where their shine. And maybe even this guy is competent in other areas or situations.

    I am just amazed that I've never had an experience where some supposed domain expert comes in and completely blows me away with the depth of their knowledge.

    Since I do all sorts of different things half-competently, I expect the person who focuses on one thing to be really REALLY good at it. But so far, no such luck.


  • Discourse touched me in a no-no place

    @cartman82 said in I hate analytics gurus:

    Since I do all sorts of different things half-competently, I expect the person who focuses on one thing to be really REALLY good at it. But so far, no such luck.

    Real experts tend to be very good. They also tend to be expensive.


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    But he failed to explain how overlaying these two graphs (avg revenue per customer vs total revenue) would be useful to anyone. And in which proportion should the axes be.

    It shows whether more customers causes diminishing returns per customer, which could apply to certain businesses. Would likely need labor or other cost metrics to measure whether the diminished returns per customer causes operating margins to drop into the red.



  • @darkmatter said in I hate analytics gurus:

    It shows whether more customers causes diminishing returns per customer, which could apply to certain businesses. Would likely need labor or other cost metrics to measure whether the diminished returns per customer causes operating margins to drop into the red.

    Wait, wouldn't more customers increase return per customer, as your fixed costs become smaller and smaller piece of the pie? That's why all these startups try to amass as many customers as they can before they start monetizing. Right?

    Either way, I don't see how this graph would show you anything without some kind of meaningful relation between graphs.

    Eg. you have 10,000 customers. Your total revenue is 17 M dollars. Average revenue per customer is 1,700 $.
    So on the left axis you have like 10 - 20 M dollars, and on the right, you have 1,000 - 2000 dollars. But nothing's stopping you to stretch the right side axis to $0 - $10,000 or $1,500 - $1,800. And this will create a completely different picture.

    Don't you need some kind of fixed relation between scales for this to make any kind of sense?


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    as your fixed costs become smaller and smaller piece of the pie?

    That's why i said some businesses. Uber's fixed costs expand at the same rate as their ability to get more customers. A restaurant's fixed costs stay the same and profits go up.
    Or say a certain product might fetch $1mil from your top 10 customers, but as you expand you can't find anyone else willing to pay that much , so you start taking less and less... then the original customers don't want to pay $1mil for a $100k product and you end up making $10mil for 10 customers or $10mil for 100 customers but having to have 10x as much support staff.


  • :belt_onion:

    not that i do this for a living or anything :trolleybus: if i was a contractor i'd inquire as to your pay rate since you appear to have an analytics guru spot you need filled.

    oh wait, i just answered your idiot guru's questions for him so i guess you don't need either of us anymore :sadface:


  • :belt_onion:

    @darkmatter said in I hate analytics gurus:

    A restaurant's fixed costs stay the same and profits go up.

    Which isn't exactly true if you're talking a restaurant company with more than one store. Maybe you get 1,000 customers/day if you have 1 store in Footown and each customer is $10 income and that store costs $5,000 to run: $10,000 - $5,000 = $5,000 profit.

    Now consider they open a 2nd store on the other side of town... but the town only has about 1,500 people that would ever want to eat at your chain. So now you max out at 1,500 people for $15,000, but your two stores cost $5,000 each to run... so you make... $5,000 still, but have twice as many places for things to break unexpectedly and cost more money.

    So now you have more customers, but the profit stayed flat, so you should have kept it at 1 store (or opened the other store in Bartown instead of Footown).


  • BINNED

    @darkmatter said in I hate analytics gurus:

    opened the other store in Bartown instead of Footown

    Those Footown barbarians just don't understand quality food. You know what they call a Quarter Pounder with Cheese in Footown?



  • @darkmatter said in I hate analytics gurus:

    oh wait, i just answered your idiot guru's questions for him so i guess you don't need either of us anymore

    Ha, it has worked!



  • @darkmatter said in I hate analytics gurus:

    Now consider they open a 2nd store on the other side of town... but the town only has about 1,500 people that would ever want to eat at your chain. So now you max out at 1,500 people for $15,000, but your two stores cost $5,000 each to run... so you make... $5,000 still, but have twice as many places for things to break unexpectedly and cost more money.
    So now you have more customers, but the profit stayed flat, so you should have kept it at 1 store (or opened the other store in Bartown instead of Footown).

    Ok, but the proposed graph still makes little sense to showcase that. You'd want to see costs as well, as you mentioned above.

    Also that still doesn't tell me how do I pick scales for two sets of lines with loosely related metrics.


  • BINNED

    @cartman82 said in I hate analytics gurus:

    Also that still doesn't tell me how do I pick scales for two sets of lines with loosely related metrics.

    Depends whether you want to make the data look good or bad


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    Also that still doesn't tell me how do I pick scales for two sets of lines with loosely related metrics.

    the scales are independent, as long as you don't make one exponential and the other logarithmic you'll be fine. i'd go with linear auto-fit for each independently in this case.

    @cartman82 said in I hate analytics gurus:

    Ok, but the proposed graph still makes little sense to showcase that. You'd want to see costs as well, as you mentioned above.

    well, replacing the 2 charts with 1 chart of profit shows you the end result, which is that profit didn't go up from having more customers. But having those 2 revenue charts helps you figure out why it didn't go up.


  • :belt_onion:

    @darkmatter said in I hate analytics gurus:

    But having those 2 revenue charts helps show you why it didn't go up.

    from my restaurant scenario, the revenues per customer don't change even though profits don't go up with more customers. So the 2 charts help show you that it is the increase in costs not the decrease in revenue per customer that is your problem.

    The one double-chart isn't the end-all be-all answer, it is one of many to help pinpoint an answer.



  • @darkmatter said in I hate analytics gurus:

    the scales are independent, as long as you don't make one exponential and the other logarithmic you'll be fine. i'd go with linear auto-fit for each independently in this case.

    Ok, that sounds reasonable.


  • Winner of the 2016 Presidential Election

    @blakeyrat said in I hate analytics gurus:

    They're miles away from rationality when it comes to selecting software.

    There's at least some rationality to selecting R over any closed-source options: depending on what you're doing, being able to be sure that the program is doing what it claims to be can be very important.

    Of course, that runs into the same problem OpenSSL did: just because it's possible for people to do that doesn't mean they will, in practice. Or even have the right skill set to properly evaluate it if they do.



  • @darkmatter said in I hate analytics gurus:

    Which is why you use multiple Y-axes. The bad part is that the guru was so completely stumped by your question that he decided to toss it rather than explain how it should work.

    That misses the point that "average revenue per customer vs total revenue" is a fundamentally useless report. The average revenue would just be total revenue divided by number of customers, so all it's showing you is the ratio (1/customers).

    @darkmatter said in I hate analytics gurus:

    from my restaurant scenario, the revenues per customer don't change even though profits don't go up with more customers.

    A "profit vs average customer revenue" graph might be interesting, but it wasn't what the guru proposed. Read again what you initially quoted: "timeline graph comparing average revenue per customer with total revenue"


  • :belt_onion:

    @Kian as I stated, the restaurant scenario specifically wouldn't show any average revenue per customer vs total revenue changes.... but that is not true in every industry. And the lack of something interesting is interesting sometimes.

    I don't know offhand what cartman's industry is or I could guess if it was going to be an uninteresting graph (though you might still make the graph at least once to at least show that it is not needed in the future). Yes, if you knew the industry specifics like this supposed guru would given that they're paying him to know things, then you could make more pertinent graphs. But the proposed graph isn't as outrageously dumb as cartman's initial post implied without more information about the industry/company. A lot of the rest of the guru's presentation was awful, so i'd find it hard to give the guru any benefit of doubt though.


  • Dupa

    @cartman82 said in I hate analytics gurus:

    First UX and design, then programmers, now data analytics... I have yet to meet a specialist that doesn't disappoint.

    FTFTCM (for testing candidates' mettle).



  • @cartman82 was this guru brought to you by the same gut feeling that brought your other "interesting" coworkers? Because then you should just replace all the "I hate ..." threads with "I hate my boss's guts"



  • @homoBalkanus said in I hate analytics gurus:

    @cartman82 was this guru brought to you by the same gut feeling that brought your other "interesting" coworkers? Because then you should just replace all the "I hate ..." threads with "I hate my boss's guts"

    No idea where he unearthed the guy.

    The Boss was crestfallen when the guy didn't produce any math whatsoever, so I hope that will be enough not to hire him again.



  • For some reason, we are still working with this guy.

    We have now hired him to work out something like google analytics reporting. Why we can't just go to google analytics and copy their ideas? Because CEO and her first lieutenant butted their nose and want this guy's pretty reports pronto up in their ivory tower.

    So now we have a big conference call with my Boss, CEO and 1st Lieutenant, where the Guru is to present his big report. He skypes in. Music starts blaring through headphones. Guru called in from a cafe. We can hardly hear him.

    I mean, he seems to have some novel idea, but jeesh. Can you possibly be more unprofessional?


  • Discourse touched me in a no-no place

    @cartman82 said in I hate analytics gurus:

    Can you possibly be more unprofessional?

    Don't ask that question! You may find out that yes, he can. :scream_cat:



  • @cartman82 Guru sounds like a wonderful job. Any bullshit you say is awesome and you can do anything and still get gigs



  • @cartman82 said in I hate analytics gurus:

    Can you possibly be more unprofessional?

    Four Words:

    Banana.

    Hammock.

    Work.

    Attire.


  • Discourse touched me in a no-no place

    @perverted_vixen Depends on how old your profession is, surely?



  • @dkf said in I hate analytics gurus:

    @perverted_vixen Depends on how old your profession is, surely?

    The oldest naturally, and even we reject banana hammocks as professional attire.

    in fact we reject pretty much all clothing as being unprofessional attire. The best professional attire is the one that you were provided with naturally.


  • sockdevs

    @perverted_vixen said in I hate analytics gurus:

    The best professional attire is the one that you were provided with naturally.

    Hey! That's cyborgist! :P


  • Impossible Mission - B

    @perverted_vixen said in I hate analytics gurus:

    @dkf said in I hate analytics gurus:

    @perverted_vixen Depends on how old your profession is, surely?

    The oldest naturally

    You're a politician?!?


  • sockdevs

    @masonwheeler said in I hate analytics gurus:

    @perverted_vixen said in I hate analytics gurus:

    @dkf said in I hate analytics gurus:

    @perverted_vixen Depends on how old your profession is, surely?

    The oldest naturally

    You're a politician?!?

    No, that's the profession with the oldest people, not the oldest profession.



  • @masonwheeler said in I hate analytics gurus:

    @perverted_vixen said in I hate analytics gurus:

    @dkf said in I hate analytics gurus:

    @perverted_vixen Depends on how old your profession is, surely?

    The oldest naturally

    You're a politician?!?

    i said PROFESSION, not disease.

    I keep myself clean and healthy thank you very much, and i have regular checkups too! because that's just the right thing to do to the jons and janes you service in my profession.


  • :belt_onion:

    @cartman82 said in I hate analytics gurus:

    I mean, he seems to have some novel idea, but jeesh. Can you possibly be more unprofessional?

    Lets try

    He skypes in. Music starts blaring through headphones. Guru called in from a cafe. We can hardly hear him.

    Are you sure it was a cafe?


  • Winner of the 2016 Presidential Election

    @cartman82 said in I hate analytics gurus:

    butted their nose

    Is that a euphemism for stuck their heads up their asses?


Log in to reply
 

Looks like your connection to What the Daily WTF? was lost, please wait while we try to reconnect.