OTHER PLACES OF INTEREST
Danny Flamberg's Blog
Danny has been marketing for a while, and his articles and work reflect great understanding of data driven marketing.
Eric Peterson the Demystifier
Eric gets metrics, analytics, interactive, and the real world. His advice is worth taking...
Geeking with Greg
Greg Linden created Amazon's recommendation system, so imagine what can write about...
Ned Batchelder's Blog
Ned just finds and writes interesting things. I don't know how he does it.
R at LoyaltyMatrix
Jim Porzak tells of his real-life use of R for marketing analysis.
HOW DID YOU GET HERE?
In fact, even if you aren’t using SPSS, it’s worth looking at what they are doing. They talk about Python, R, Visualization, all the cool stuff. Highly recommended.
* * *
The cool folks at Dataspora helped run a meetup of the Bay Area R users, and got some great collected info about R. If you want to see how R is being used in real world ways, this is a must read. The whole blog is pretty fun (for us analytic types), so worth adding to your RSS reader and keeping up.
Besides the cool stuff from Google and Facebook, my old pal Jim Porzak talks about how he did database and marketing analytics with R.
BTW, I consider these guys at Dataspora to be like Juice Analytics : great examples of folks who are using open source and modern “mash-up” approaches to both analyzing and communicating their findings. There are lots of analysts out there who delight in throwing around eigenvalues and goodness-of-fit tests… but to have both the ability to detect the pattern and visually explain why it’s important… well, that’s a rare thing.
Keep your eyes on these 2 groups for more cool stuff soon…
* * *
I’d write a whole thing here, but http://www.sics.se/~joe/bluetail/vol1/v1_oo.html says it so nicely that I’ll just refer you to that page.
Hacker News has a whole list of rebuttals and comments, but they tend to be the usual “oh, it works for this one point case, so it must be great for all things” responses. See http://news.ycombinator.com/item?id=474919 to join in the fun.
* * *
Everyone thinks analytics is all about the data. I hear amazing stories of petabyte warehouses, state vector learning algorithms, and every other aspect of the data. But it really comes down to the story: What is the data telling us about the world? This is the context, the relevance, the “impact”, all those pieces.
This comes, in part, from shared meaning. Once we all understand what a metrics is all about, then we can communicate with it. But if we make up something new just to be different, then we lose some of that shared meaning, communication disappears, and the story fades away. That story is the center of it all: if you can’t tell the story, then all you have is meaningless data.
And when it’s missing, its just painful to watch. Here are some examples:
RYG It’s very common to have Red/Yellow/Green status reports on projects. But did you know that people have different interpretations of RYG? My most recent team has this standard:
This seems strange to me. I think in terms of Star Trek: Green means everything is good, but when Yellow alert is called, well, get ready to bring the noise. And Red alert means that people wearing red shirts should just start writing their wills.
But in this version, we have a page full of “Yellows” which means that all is going fine… which, in Star Trek terms, should be Green. And given that I think this way, so do lots of others. So, when we review our status with others, we have to spend time explaining that our metric, though named the same as theirs, has a different meaning… and that they shoudn’t freak out just because we have a page full of Yellows; that just means all is fine. And, not surprisingly, we wind up spending time just converting Greens to Yellows when presenting our status internally, but keeping them Green so that we don’t get confused among ourselves.
Lack of shared context just creates dysfunction, but people in power like things the way they like them. The story suffers when the writer can’t communicate with a shared language.
2009 Super Bowl Ads Watch them anywhere; I saw them at Yahoo! Big Games Ads. For $3million per 30 second spot in some cases, one would expect the ads to really attempt to resonate. Instead, many threw shared meaning into the air with no real story, or a story relying on assumed shared context that just wasn’t there.
Who were Disasters:
Call them assumptions. Call them the “duh” factors. Call it what you will, but everything, every interaction, should be a story based on shared understandings. Try to use your audiences’ language; reference their POV in the story. Put a beginning, build to an end, wrap it up. Stories are how we communicate throughout history, so make the story understandable with shared terms (or at least define your needlessly made up ones).
These examples all show flops because of lack of story or lack of shared context, or both. All your visualization, all your analyses, they all need to flow into a story based on shared context.
Take a few minutes, watch some of these ads, think about how you communicate your data: Are you creating a story like Bridgestone or Monster, or are you throwing junk up and hoping your user will see something in it like Sobe?
* * *
A friend asked why we can’t wear sports shoes to work. I mentioned that he already was…
Here are some brands of “standard” affordable business shoes, and their parent companies:
Johnston and Murphy: Owned by Genesco, makers of Lids, HatZone, and other sports stuff
Cole Haan: Nike
Bostonian: part of Clarks, the comfy shoes guys
Some are “sporty work shoes”, and are part of a holding company:
Some of these shoe makers are still independent from holding companies and trying to make it on their own.
Allen Edmonds: Still independent
Florsheim: Back from the dead, owned by the founder’s family and the Weyco group (BTW, we all love marketspeak: from their annual report… “Weyco Group designs and markets moderately priced and better-grade men’s branded footwear for casual, fashion and dress lifestyles. The principal brands of shoes sold by the Company are Florsheim, Stacy Adams and Nunn Bush.” Moderately priced and better grade?)
Apropos of nothing, Zappos has a “Made in the USA” option on it’s site (http://www.zappos.com/n/es/d/722000390/page/1.html). However, you can’t search for “made in Italy” or other well known shoe countries.
* * *
Or so it seemed. I was installing Jedit to see if they had fixed any of its annoying gnats. (Side note: its on version 4.3pre16. If you have to release something 16 times before it’s considered done, face it, you need to adjust your release system).
The installer I’d almost always used was the Java one; it was pretty low impact. I chose my folder (c:\\program files\\jedit) and away it went. The java installer requires you to make your own start menu entry, so I went in explorer to find my folder… and it wasn’t there.
Now, I have “show hidden and system files” set, so I didn’t think that was it. And when I ran an Administrator Command Prompt and did a dir in my Program Files directory, nope, not there.
So, I reran the installer to make sure that indeed, I was putting the write name. This time, I did the browse to force it to pick the right directory… and my jaw dropped. The Java installer was showing the jedit directory already created! I alt-tabbed to my explorer, refreshed, no jedit; reran the dir, no jedit.
I installed this time in JeditAgain. Same thing. No presence in a dir or File Explorer window.
This was strange. Time to search.
Someone at StackOverflow, the Yahoo! Answers for programmers, had nailed it.
Java File.canWrite() on Vista and super hidden files is where a guy showed how Java can write files that he can’t see, but Java can… And the answer is:
UAC again. Yes, fans, when User Access Control is on, the java app writes are redirected to C:\\Users\\
And when I went there… yep, all my versions of Jedit were there, and a bunch of other stuff besides. Whenever a program was writing to it’s own directory (uncompressing files, whatever), it was putting them in this VirtualStore area.
Yet another reason to really hate Vista. I know, it thinks its doing it for my own good… but jeez.
BTW, if you want to run a java jar as an admin, you need to run it from the administrator command prompt.
* * *
Great read, and fantastic validation for an amazing amount of effort.
If you want to read more about R, feel free to peruse my R Statistical System section collecting my articles and tutorials.
* * *
I’ve blogged about Piwik myself, but some of these others were new to me. Worth taking a read. (And not just for the plug for Yahoo Web Analytics…)
* * *
Over the years, lots of different types of bidding systems have tried to unseat eBay. Some have bid extension (last bid extends expiration by 15 minutes, so no more sniping), some have various step requirements, some only handle B2B and bulk items (UBid is the most famous of these), but none have really lasted in the US. (Note that eBay is a 2nd stringer in most other countries: good, but usually not the leader. Who knew?)
Now, from Baton Rouge, La., comes a strange new play: UTurnBids.com. It’s kind of hard to explain, but basically, its like a reverse auction… only you have to be low and unique. So, let’s say 201 people are bidding. The first guy bids 1 cent for the item, thinking “I’m the lowest!”. Next guy bids 1 cent also, with the same logic. Now, they’ve canceled each other out. On down the line, as each person picks a low bid, if its duplicated, then it cannot win. This implies that in the search for uniqueness, you would try to go to 2 digits, and potentially 3, for your bid. At least, that’s the hope. They also hope you would bid multiple times to find the non-unique one, and they can make money that way.
http://www.uturnbids.com/ContentImage/How-It-Works.aspx has a picture.
Clever idea. Of course, without lots of bidders, it seems that lots of very low bids (all odd numbers, of course) are winning: the Recent Winners shows some good grabs.
Anyway, thought it was interesting. I haven’t signed up to see if it tells you that your bid was duplicated or not. Also, their business model is the “bid credits” system, meaning that you pay to keep bidding (there are some free ones for small items limited to 5 bids a day).
I wish them luck, as I do for any company trying to make it in tech based in La. It’s a nice twist, and it’s fun to learn about how much effort goes into bidding systems. I’ve had the pleasure of chatting with some Yahoo! research scientists about pricing and bid theory, and there are more twists there than you ever expected. See the Auction Theory category in Wikipedia for a collection of topics.
* * *
Kevin Hillstrom has a great letter about how Web Analysts need to go beyond the limits of their tools to actually answer business questions.
While I agree with everything that he says, I also don’t want to let the web analytics tool suppliers off the hook so easily. I expand beyond his points in my (long but worth it) post, What Web Analytics is Missing.
I delve into similar issues as to what Kevin mentions… but I also point out that we are hindering the analysts by not providing the right tools. Both sides, the analysts and the tool vendors, need to work together to up the game.
When folks ask me “what’s new in the web analytics world?”, I am sad to say that the answer has been “not much”. This is building pressure: startups will find ways to solve the problems, analysts will get sick of paying fees to companies who can’t integrate enough or the right types of data, or provide analysis of business metrics instead of web behaviors.
Read Kevin’s note (and the comments) and read my post What Web Analytics is Missing. If you are an analyst, here are some great ways to raise your game. And if you are a tool vendor, well, here is your roadmap of product features, nicely laid out for you.
* * *