Summary: After salvaging a crash-and-burn 80-hour production job, I spent six months reinventing a complex digital pipeline that spanned multiple creative and business units. The outcome using new processes and automation was a scalable system of producing assets at warp speed. Time savings was 95% (2.5 FTE’s), and human error was reduced to near-zero.
Last March I found myself halfway across the country buffing my brain at the Adobe Summit in Las Vegas. The conference was pretty much over for the day, and just as I was getting into happy hour… a panicked call came from the home office in New York.
My team had just received the Mother of all Job Requests — at the end of a grueling week producing digital marketing and web pages for a new iPad on att.com. Then, WHAM! Out of the blue, another device manufacturer decided to make a major product launch of its own — and my team was called back into action.
The upshot was that we had six working days to produce about 2,750 product image assets for our e-commerce catalog. One of our in-house client/partners had placed the order, with a hard deadline. They would not give us more time to do the work, even though I begged.
I estimated the job would take about 80 hours. But how would we do it? My team was burned out and maxed out. And our laborious process was never built to produce at this scale or speed — even with overtime.
Back in New York I jumped into the trenches — spending the next week working shoulder to shoulder with my team to crank out the work. Let me tell ya, it was one helluva crash course. Some interesting patterns began to emerge. Basically, the whole process was a disaster.
The problem itself was almost textbook perfect — a crossover challenge between people, process, and technology.
Change was desperately needed in all three areas.
1. Untangle the Problem
Catalog work is recurring business for us. The digital design team creates product images for the e-commerce portions of the company website. It’s not the highest-profile work, nor the most glamorous. But it’s incredibly important to the business. It helps drive sales.
The work is driven by an in-house “business owner” whenever a new product comes out. For example, a manufacturer might introduce two new phones, each in three colors, accompanied with some 200 accessories. (And yes, that can happen — especially with a major product launch by Apple or Samsung.) Design will crop and resize each source image to a variety of sizes and proportions for up to 15 merchandising placements across our websites.
Catalog work was one of the early Big Messes I identified after starting my job at AT&T, and I did plan to fix it someday… on my own time. This crisis forced me to deal with it sooner — but I did not rush. Once we got out of the blue hole, I chipped away at the problem for the next six months.
A sane and sustainable solution would be:
- Easy to learn
- Scalable
- Elegant, with minimal steps
I knocked on every door in the supply chain to take stock of where requests and source images came from, how we processed and repurposed them, and where they went after they left my shop.
The new process slowly lurched into place. I would build a prototype automation that was 4x better, then throw it away for one that was 40x better. I needed a comprehensive Photoshop template, so I enlisted my geeky designer colleague Adriel Caldelas who fleshed out the artboard concepts to perfection. I created some prototype actions, then tested and rewrote them until they were bullet-proof. I discovered allies within AT&T with expertise on the nuances of AppleScript and Automator.
By the end, I had standardized image specs, simplified client requests, revamped my team’s production methods, built custom droplet-style apps, and created documentation for everyone.
How amazing was the solution?
Not to brag, but it reduced the level of effort by more than 90% and slashed the defect rate to nearly zero. Most important, I’d zapped the worst grunt work for my team… and replaced it with more creative and meaningful projects. (OK, that is bragging. But it’s pretty great, right?)
In the process, I had changed the people, our process, and the technology.
We reduced the level of effort by more than 90%, and slashed the defect rate to nearly zero.
I totally agree with my friend Clair Carter-Ginn who recently said in an interview, “Do not rely on a tool to fix your process!” She and I have learned, in our own ways, that you’ve got to actually understand and solve the root problems first. We love tools! All hope for improvement depends on them! But you can’t drop a tool on a situation and expect magic to happen. This is much deeper work.
2. Clean Up the Supply Chain
Problem – Conflicting Priorities: When our client had a new batch of products or accessories to add to the catalog, they’d set up a master ticket… then chop the big job into three smaller jobs before droping them into my team’s queue. Why? Well, the final images were sent to three different implementation teams. Three silos. To our client it made perfect sense. To me, it was ridiculous. Triple the work. Each batch was downloaded three separate times, by three separate designers, and then opened, checked for accuracy, silhouetted, and resized. And in the end, this yielded only one-third of the total job. There was no coordination between designers doing nearly identical work. How could there be? The system was stacked against us. The problem was upstream.
What Changed: We glued it back together at the front end. One batch, one ticket. My designers can roll out all the assets in a single beautiful flow. Then they organize it at the back end and give it to our client the way they like it — three tidy folders. (This also eliminates white noise because we now get the same work done with only one-third the number of tickets.)
Problem — Poor Image Quality: In the system I inherited, as much as 25 percent of every batch of source images might be low-rez. The first thing a designer had to do, before even starting work, was to inspect for sub-par images and send a report to the client. The client would pass our feedback to their client… who then reached out to the OEM for better images. And yes, better images almost always showed up. Why was this accepted as normal and never addressed in any sort of institutional way?
What Changed: Back into the supply chain I crawled — two degrees out — with a spec sheet for our client’s supplier. I did a couple of training sessions to make sure everyone was cool. Then I flipped the switch. Voila! Our client now owned QA. Garbage in, garbage out, my friends. My goal on the design side was to motor through the work without stopping. Deliver the assets and close those tickets. My designers didn’t take it well. They’re a pixel-perfect bunch. “Oh, but what if we see a jaggy image?” one of them tried arguing. “Are you really telling us to just do it anyway and send back crappy work?” Me: “Yessirree! That’s the only way those guys are going to learn.” I am teaching my team to prioritize. Do responsible work, yes, but let’s knock out this production stuff so you can all get back to work on higher-value projects.
Agreed? Agreed!
3. Reinvent the Work
Problem – Inconsistent process: To make all the different sizes (with extremely nit-picky naming conventions), designers had come up a few tricks to speed up their work. Somebody had written a couple of Photoshop actions. But a small job of 20 items — where each one had to be saved out in 15 different sizes/proportions, one at a time, to make 300 final assets — could literally take an entire day, sometimes longer. File naming was entirely manual and where the most errors happened. Everyone did the work differently, and we couldn’t fairly estimate how many hours any particular job might take.
What Changed: I front-loaded the new process with planning and prep work — the part that requires thinking, and visualization skills. Everything else, we gave to the robots. Templates and automation now do the heavy lifting of producing all the correctly named sizes and formats. I have a soft spot for storyboards — they really helped me visualize things as I worked out the kinks.
- De-fragmenation: Client supplies source images all in one clean batch (no extras, please)
- File Names: Client must pre-name all source files (GIGO)
- Image Check: Designer inspects the source images in one sweep, and cleans up any images that need it
- Image Prep: Using a Photoshop action, designer sets all source images to a standard size
- Templates: In a master Photoshop template, artboards are set up for each of the needed sizes
- Smart Objects: Each artboard contains an instance of a smart object; the designer swaps in a prepped source image one time, and BOOM! all the artboards update by magic
- Automation for naming: Via AppleScript, artboard layers are renamed based on multiple variables including the source file name
- One Click: Adobe Photoshop’s “generator” function creates all 15 individual assets, named correctly, in about 10 seconds
- Organize: Use a rule-based Automator droplet app to merge final files for the project into three consolidated folders, for delivery
Problem – Extra loops: The existing process included a full client review of every single asset we produced, prior to delivery. It was essential. Image mixups and file naming mistakes were common. And no wonder — our entire process was manual. This review step added an extra day to every job.
What Changed: Once production got automated, our defect rate plummeted. Mistakes became an endangered species, and the review loop was no longer needed. So I killed it.
Problem – Cluttered design: The design team had been adding fake reflections and shadow effects to every item. I didn’t see the value — it was time consuming work, and ugly. So I went on a hunt to find out why we were doing this. The best answer I got was, “It’s what we’ve always done.” So I asked: “Does anybody in this company think it will hurt sales if these effects go away?” Crickets.
What Changed: Those shadows and reflections… I killed them. Today every item — across the board — sits on a plain white background. The site looks cleaner and more consistent, and it destroyed our worst productivity sinkhole. Designers have not complained that the work got easier.
Problem – Too Much Management: The old production process involved an Art Director to ensure that the reflection and shadow effects were consistent and believable. But this was a fail because the work still looked fake and cluttered anyway.
What Changed: With the effects gone… did we need an Art Director on these jobs anymore? No, we did not. So… I killed this, too. Designers would be OK without a nanny this time. And Art Directors were thrilled to focus on higher profile projects.
4. Lock-In the Wins
OK! Now it’s time to swap in the new process — and make it stick.
What Changed: I created a one-sheet to show how the two primary teams (Design, and our client) will work together. It establishes a common language, outlines the process step by step, and gives our client boilerplate language to speak to us — replacing their rambling job requests of yore.
- Communicating with all stakeholders
- Training sessions for everyone in the pipeline
- Documenting the new workflow
- Standardizing format of requests from our client.
- Standardizing the production process on my team
- Micromanaging for 90 days by watching every job request to ensure end-to-end compliance
- Shouting this achievement to my manager
Measures of Success:
- Can generate up to 600 assets in less than an hour
- Reduced 10 hours of work down to 20 minutes
- Slashed defect rate to virtually zero
- Built to scale
- Freed up 2.25 FTE’s for more high-value work
- Eliminated the biggest pain point for my team
Bottom Line:
It took a crisis for me to fix a longstanding problem — and overhaul our production pipeline to deliver digital assets faster, at high volume, and with greater accuracy. Here’s what I learned: Some obstacles will take all the innovation, endurance, and persuasion you can muster. But if you succeed, it might be spectacular!