Two Interactions With Amazon «

Code: , , ,


When I had one day’s notice that Hillary Clinton was going to release her schedule from the time she was first lady, I thought it entirely possible we’d have to scan it, or work with a scanned copy. So I signed up for an Amazon Web Services account in case I needed to rent some computing time via EC2 to process all the images.

There’s one catch, though, new Amazon users were limited to running only 10 (now 20) instances at a time. I sent an email requesting the limit be raised, but there were two snags. First, their billing system didn’t allow them to up the limit until I’d initialized my account by running one (thus using it, even a little). A techie emailed me to tell me this, but every time I tried to run one I got an error message.

The second snag: I’d screwed up the expiration date on my credit card (got a new card in the mail the day before, gave that expiration and hadn’t called to activate it yet). So, at 8:15 PM on a Tuesday night, that techie called me, conferenced in someone from Billing who corrected the error and ran a test transaction to make sure my card would work, then stayed on the line for me to spin up an instance so he could raise my limit.

It was great customer service, especially as I ended up needing to use those instances to process Clinton’s schedule.


A week ago I was writing about refactoring and linked the Amazon pages on my two favorite books. While I was at it, I signed up for their affiliate program and researched how to tag those links. Even after reading through their documentation, I could not figure out how.

They provide a Fancy Link Builder so you can make fancy image links, but I only wanted an unobtrusive plain text link. The Link Builder doesn’t do that: the option for spitting out text links still includes a 1×1 image, and the links include dozens of GET variables — anything but plain. I knew I’d seen short Amazon links on other sites, so I contacted support asking for help fixing my failed attempt.

I got back a form latter informing me what I’d told them, that my attempts didn’t work. It also said I should use the Fancy Link Builder that I’d said didn’t work for my purposes.

I replied and asked how to create a link without GET variables, and I got back a form letter telling me to add a GET variable with my tracking ID.

I replied and asked again and got a form letter with detailed (but out-of-date) instructions on how to use the Fancy Link Builder… to produce a “text-only” link that included an image and GET variables.

I replied to ask again and mention they should update their previous form letter. I got a form letter that insisted none of these were form letters, which I guess means whoever wrote the fourth form letter thought it would be better to try to convince me their customer service personnel typed long, detailed (and once, oudated) messages to my specific question that evinced a lack of comprehension. (Then it suggested I should probably just go post on the affiliate board.)

I did what I should have done a few rounds before and went to find examples of affiliate links. After some tinkering and testing, I figured out how to build the links (add /trackingid-20/ref=nosim/ to the end).


I was a bit surprised by the stark difference between the two customer service interactions, but I realize two different departments at a large company can easily be that different.

And then a difference occurred to me. I went back to look up my emails with the EC2 techie: turns out he wasn’t a techie, his title was Business Development Manager. I guess someone saw the on the end of my email address. There certainly aren’t any case studies on Peter Harkins successfully linking to two refactoring books.

As bad experiences go, the affiliate linking was pretty mild and doesn’t bother me at all. It was a weird request, most of their affiliates are nontechnical and don’t fuss about URL aesthetics, and the image links probably peform better.

The EC2 interaction was great at the time but bothers me now. I hate special clubs and privileges, and that’s pretty clearly what I got. (Though it’s possible some Amazon employee with a title like “Community Outreach Supporter” will notice this post and come explain that every new customer gets after-hours handholding from a manager who conference-calls in a domain expert. Um, sure.) I realize here that I’m complaining about customer service being unfair in my favor, but I can’t help but think it’s wrong.


  1. I did a quick aptitude search ocr and found gocr, ocrad, and tesseract-ocr. I spent an hour or so testing each package with different options on samples from the Clinton Schedule, which I extracted with

    gs -sDEVICE=pnggray -sOutputFile=page.png -r600 -dTextAlphaBits=4 -dGraphicsAlphaBits=4 -dBATCH -dNOPAUSE -dSAFER -dFirstPage=42 -dLastPage=42 -f hrcsked.pdf -c quit && convert page.png page00042.tiff (I also used ImageMagick to crop the image down and got better results because there wasn’t the staticy margins, but I can’t immediately find my notes on that and the IM docs explain how to do it.)

    I found that tesseract worked best, but I strongly suspect that the different ocr programs will rank differently depending on the material.

Leave a Reply

Your email address will not be published.