== Lojas KD - Images Processing ==
Whenever you are scaling a business you need to decide what to automate and
what hold for now and just add one or two more people. The products' images
treatment team in Lojas KD was on the later option for a long time.
Not only they were not being on operations' priority, but as they often used
to suggest improvements and report issues on their processes. Until the day
finally came. A massive backlog of improvements, bugs and wishes was in my
hands.
Summing the estimates (we are talking about of more than 3 years of pure
negligence) I had more than 400 hours to code. Basically a quarter. I
immediately realized that it was not going to happen. Not that way. It was
time to put some expertise to play.
First, I talked to the products team manager and asked permission to hang
around one or two days for a field observation. She agreed, so I get a place
right near the image treatment team leader.
- Can you show me how do you upload an image to a product, after having a
finished jpeg? - I asked.
- Sure thing! - he replied - First, you copy the files to a specific network
folder. Then you access a specific URL in the intranet, so all sizes
variations are generated into another folder. After that you use a form to
upload the files. Finally, you can enter the admin panel and manage the
images...
What a pain! The whole process usually takes around 20 minutes to be done.
It was so wrong and so bad in so many ways... This experience allowed me
some insights. The most important of them was that the backlog was a bunch
of patches over a badly designed process, and implementing everything would
make their lives suck a little less, but won't bring any significant
improvement.
They need (deserve!) a whole different process. One that allows them to
focus on what they excel: edit images. And deal as little as possible with
copying, pasting and uploading pics for half an hour.
After more two days researching and scratching a solution, I came up with a
very different approach. Installed a S3 client on their machines. Created a
bucket on S3 for receiving the finished pictures. And an ec2 machine for
processing batch tasks, such as resize images, write on proper buckets, and
update the database. Their work now would be just upload the finished
picture, and 5 minutes later, check the images order on the admin panel.
So I delivered an unexpectedly better solution in less than 2 weeks, and,
more important, everybody was happier. This story illustrates how being in
touch with the target audience is a tipping point, and how when we take
responsibility for business success things get totally different.