Lean production, at its most fundamental, is about reducing waste through standardization, continuous improvement, and worker empowerment. It was introduced and popularized by Toyota, who used it to rise from an unknown to the largest automaker in the world in less than a decade.
Lean principles have since proliferated far beyond manufacturing and into fields like technology, finance, and even farming. At Pixelz, digital lean production is the ethos and engine of our business.
Every editing step in S.A.W.™ is visible to customers, both during and after production.
I've encountered resistance from retoucher acquaintances of mine when I explain that, at an elevator pitch level, we've built a Photoshop assembly line. There's an immediate assumption that the quality must be lowered, followed by an accusation that the art is being stripped out of retouching.
Why? Are TAG Heuer watches lower quality because they're assembled by many different pairs of hands and robots? Are Disney movies not art because different technicians work on character models, or terrain, motion, color, etc.?
The opposite is obviously true. When Toyota brought lean production principles to auto manufacturing, quality improved to the point cars started lasting hundreds of thousands of miles. When Edwin Catmull was leading Pixar from production of the first ever computer animated feature film all the way to winning 13 Academy Awards, he turned to The Toyota Way for inspiration and instruction on leading creative teams (as seen in his memoir, Creativity, Inc.). I don't know anyone who will argue, with a straight face, that films like Toy Story, WALL-E and Up aren't art.
So let's talk about how we're using digital lean production methods to bring the art of retouching to e-commerce images at scale. It's all about S.A.W.™, Specialist Assisted Workflows.
The simplest way to explain S.A.W.™ is to call it a Photoshop assembly line.
It’s a proprietary production system that analyzes uploaded images, breaks the retouching process down into component steps, and then routes images between specialist human editors and automated processes depending on availability and deadlines.
We’ll illustrate how it works by using the famous 14 principles of The Toyota Way to connect classic lean production principles to modern digital lean production.
In part one, today, we’ll focus on the digital assembly line itself. In part two, we’ll go deeper on the human side and show why empowering workers is fundamental to Pixelz and lean production.
Head of QA Duc Anh and CPO Tejs Rasmussen mapping out post-production processes.
Pixelz Long-Term Philosophy
“Base your management decisions on a long-term philosophy, even at the expense of short-term financial goals.” - The Toyota Way
An early Pixelz production office build-out
Pixelz exists because our founders ran a Danish e-commerce company and got frustrated by repeated bottlenecks in post-production. They wanted to solve it.
“Solve it” didn’t mean build a better version of the same model, i.e. become an outsourcing company with more editors, more skilled editors, or with better communication—although we’ve been all those things through the years while accumulating data. A real solution meant permanently changing the model.
In our case, that goal is to automate product image editing. To develop software that enables anyone, anywhere, to get high quality product photo retouching at scale with the click of a button.
To realize that goal, we’ve structured our organization and made deep long-term investments in research and development with our eyes on the prize.
Pixelz has been working towards S.A.W.™ since our founding in 2010, and we finally began rolling it out in 2016.
Identifying Waste
"Create continuous process flow to bring problems to the surface." - The Toyota Way
Pixelz’ editor software wraps Photoshop with per-image instructions, tracks activity, and routes images between editors at different steps.
We’re a digital lean production company. That means we’re extremely process focused, because the right process will prevent waste and get us the right results.
To get to a point where could identify waste, we needed data. From Pixelz’ earliest days, all Photoshop work has been done in a custom client that tracks every action taken by an editor. Every brush stroke, every clone, each image’s loading time, is recorded. We have detailed data on the retouching of well over 10,000,000 images.
Detailed Photoshop activity tracking for millions of images surfaces waste.
That data allowed us to identify the following three types of waste in post-production:
- Planning waste. Production planning is time consuming. It requires constant attention to ensure images are assigned to the right team and editor, and reassigning when necessary. Sometimes one editor had a huge backlog while another didn’t have any images at all.
- Skill waste. Sometimes highly skilled editors were doing simple tasks like cut-outs and resizing. In a perfect world without waste, editors with more experience work on more challenging tasks.
- QA waste. Rejected images waste editing work itself. We often saw bottlenecks in quality assurance performed by team leaders, and when images were rejected the process of redoing images takes up extra time. Sometimes mistakes had to be corrected by a different editor, again adding extra time to the process and potentially leading to broken deadlines.
So how have we acted to eliminate that waste?
Standardizing and Automating Processes
"Standardized tasks and processes are the foundation for continuous improvement and employee empowerment." - The Toyota Way
Image editors specialize in editing steps and work in a proprietary Photoshop wrapper.
The traditional way to manage retouching of product images is to assign one image, or batch of images, to an editor and say “Go.” That’s the way it’s always been done, and it’s extremely wasteful.
That method of work, where a single person handles every step from start to finish, is cottage industry manufacturing. As in, “I’ll just ride my horse into the village and commission shoes from the cobbler so I have them in four weeks.” That kind of centuries-old cottage industry.
Assembly line manufacturing replaced cottage industry manufacturing for physical products because it was much more consistent and productive. Higher quality, faster, for lower prices.
No matter how complex a retouching process appears at first, it can be broken down into a series of steps. Furthermore, every image is unique — but the steps taken when retouching are not. Whether you’re cutting out a shoe, dress, or necklace, you’re using the pen tool in much the same way.
What if we trained specialists to individual steps and then routed images between specialists? Could we deconstruct image editing to the point that many different editors could work on the same image, assembly line style?
We could, we did, and we found that it comes with a whole plethora of benefits, like:
- Algorithmic traffic control
- Increased automation
- Expedited orders on demand
- Increased consistency
- Surfaced QA problems
- Reduced employee onboarding times
Let’s look at them one by one.
1. Algorithmic Traffic Control
"Level out the workload." - The Toyota Way
Images are routed between editors based on image deadline and editor skills.
There are over 35 different S.A.W.™ steps. Most of our 1000+ editors are trained to about 3-8 skills, and most images go through 4-6 steps, but it’s certainly not unusual for an image to go through 12 or more steps.
S.A.W.™ eliminates planning waste by routing images through steps based on editor availability and image deadline. Each editor’s workstation has only the image they are working on and one other loading in the background; when they finish a step, the system moves the image to another editor (or through an automated process) for its next step.
Leveling out the workload prevents unnecessary bottlenecks and reduces the number of staff actually required to be on the clock.
The beauty of a digital assembly line is that you’re not constrained by geography. With S.A.W.™, different image editors working on the same image can be in different buildings, cities, or even countries. The only thing required to integrate an editor into S.A.W.™ is training as a specialist and the software client.
2. Automation
Some steps are consistent enough that they can be accomplished by automated processes, i.e. Photoshop Actions. As we acquire more data and break image editing into smaller and smaller steps, we are able to train AI to do analyses and increasing amounts of work. Machine learning has a prominent place in all our future plans.
Additionally, we can perform basic sanity checks as images move from step to step. Do image dimensions match those in the specs? No? Flag it for QA.
3. Expedited Orders
A button click returns orders or individual images in 3 hours or less.
One of the great benefits of algorithmic auto-assignment is that it’s extraordinarily flexible and responsive. The best example of this is the case of expedited orders, or “putting a rush on it.”
Traditional rush orders require multiple points of contact: a customer calls their rep, the rep calls an on-site supervisor, the supervisor walks around looking for editors who aren’t working on an urgent project, and then those editors drop everything and work exclusively on the rush images.
S.A.W.™ automates every part of rushing an order and requires no changes whatsoever to the standard process. A client clicks “Expedite” on an order or image from within their account and S.A.W.™ adjusts those images’ deadlines to give them priority. There is no difference to the editor who is working on an expedited image; indeed, they don’t even know. Their focus remains on the single step in front of them.
4. Increasing Consistency Through Specialization
Because steps are small and repeatable, specialists quickly become experts and consistently apply their skill. Consistent individual steps lead to more consistent final outputs.
We have also found that because jobs are less complex, less errors are introduced to begin with. Error, of course, still happens—but S.A.W.™ surfaces problems much faster, allowing for rapid correction.
5. QA Gates Surface Problems
"Build a culture of stopping to fix problems, to get quality right the first time." - The Toyota Way
A technical supervisor web and mobile app enables rapid on-site quality assurance.
After S.A.W.™ was running for a few months, we introduced a QA gate after every step. That means either an automated or human check happens every time an image moves between editors. That may sound wastefully repetitive, but we actually found that it increased efficiency.
Why?
Because breaking image editing down into distinct steps makes it much easier to identify when something goes wrong. That means if a specialist is incorrectly performing a step, we can immediately stop and retrain them. We can also correct that image and only waste the performance of one step, rather than having to redo more steps if QA found a problem after editing was complete.
6. Reduced employee onboarding time
"Standardized tasks and processes are the foundation for continuous improvement and employee empowerment." - The Toyota Way
Pixelz editors and trainers at work.
Isolating skills allows focused training. We have been able to accommodate rapid growth because we can bring specialists up to production level far more rapidly than when they had to master many complex Photoshop skills.
For example, before S.A.W.™ it would take us two months to train staff to the point where they could go live, and up to six months before reaching average productivity. With S.A.W.™, we have cut the go-live time down to three weeks.
This doesn’t just cut down on training cost and resources; it is also motivating for new staff because it allows them to work on real images and be part of the production process much sooner.
Empower People to Make Something Great
"Become a learning organization through relentless reflection and continuous improvement." - The Toyota Way
Bonus quote!
"Making the process better, easier, and cheaper is an important aspiration, something we continually work on - but it is not the goal. Making something great is the goal." - Creativity, Inc. by Edwin Catmull (Founder of Pixar)
Will Pixelz be moving even further into the photo studio?
We’ve come a long way with products like Enterprise, but we’re not done yet. Our core focus, to revolutionize the way visual content is created for e-commerce, remains.
What might the future look like for Pixelz? Is it SaaS? Is it machine learning and fully automated image editing? Is it moving even further into the studio? We have a roadmap, but the truth is that it is constantly changing—by design.
Our goal is to develop a team where everyone, at every level, is thinking lean. Where waste is recognized and eliminated because it improves your day, and improving your day helps the company and our clients. Where each employee has ownership of, and responsibility for, their work.
That’s why I believe that what drives successful digital companies is fundamentally the same as what drove traditional manufacturing.
Because technology might be the vehicle a company rides to success, but people are what drive it. Digital lean production, just like lean manufacturing, is about giving people the tools and desire to make meaningful change.
S.A.W.™ is a Photoshop assembly line, yes. But it’s technology, people, and culture too.
How Digital Lean Production Empowers Workers
Read part two of this blog post, How Digital Lean Production Empowers Workers, to see how empowering workers develops leaders, partners, and the type of deep understanding necessary for digital lean production.