When you get measured for a pair of jeans at ApparelWerks, you’ll get custom measurements: seat, waist, front rise, back rise, inseam and outseam. If you get a number wrong, even by half an inch, your jeans won’t fit properly. ApparelWerks delivers garments that delight customers with their fit, which makes accuracy the product itself.
The catch: those measurements are scribbled on a paper job sheet traveling with the garment through its lifecycle. When the company started, the only solution to get numbers into ApparelWerks’ system was to type them in manually. They closed that gap with Datalab by turning minutes of error-prone keyboarding into a verifying glance.
About ApparelWerks
ApparelWerks is a made-to-measure and made-to-order apparel manufacturer born from the collaboration between Rick Levine, an engineer from a family of artists, and Steven Heard, who has decades of experience making clothing (including at the Levi Strauss & Co Valencia Street factory in San Francisco). Together, deep hands-on manufacturing expertise meets an innovative approach to how clothing gets made.
What makes ApparelWerks unique is its philosophy towards clothing production. Their goal is to solve a familiar frustration many have experienced shopping for mass-produced clothing. Buying jeans that fit is hard! Worst case, you can try two pairs of jeans in the same size, yet they fit differently. ApparelWerks addresses this problem with one-piece flow, a manufacturing method inspired by the Toyota production system. Rather than cutting and sewing in bulk batches, a team makes one garment, follows it through the process, and then makes the next. A problem in a pattern is corrected as soon as it’s found, rather than after a stack of a hundred incorrectly cut pieces is already sewn.
ApparelWerks serves two types of customers, and each informs the other. First, they are a pattern development house for brands that create clothing. They support pattern creation and design for manufacturing. Recently, they also began experimenting with making custom women’s jeans for retail customers, as well.
A pair of custom jeans coming together on the ApparelWerks shop floor
How ApparelWerks produces clothing
A custom jeans order for an ApparelWerks customer starts with a handwritten job sheet, with measurements taken from an individual. These measurements are unique to each person; every measurement counts.
The customized nature of their manufacturing means people are at the forefront of their production system. Their manufacturing team consists of skilled sewers and patternmakers. Their judgement and skillset crystallize in a physical job sheet that has to move with the garment through production.
Every garment going through their manufacturing process has a unique identifier, a five-character code encoded on a woven QR-code label, sewn into the finished piece. It’s assigned at intake and travels with the garment throughout the entire manufacturing process, providing an easily trackable key for production. After the customer gets their jeans, they can use the code to order another garment in their exact size. Instead of an anonymous, mass-produced piece, every garment is identifiable as it goes through its lifecycle.
The job sheet is a working document. Someone reads the measurements and creates a pattern, producing a cutting file. Using that file, someone pulls the fabric and cuts the pieces of cloth; they go into a tray with the traveler, the garment is sewn and a team member performs QC. A garment accumulates roughly 15–20 measurements across two passes — the customer’s order measurements and the QC measurements of the finished piece. Both sets have to make it into the system, accurately, every time.
The bottleneck: excellent at sewing, not at data entry
When Rick and Steven first started their production, handwritten job sheets soon became an analog bottleneck. They required a painstaking amount of manual effort; when a garment was finished, someone would mark the sheet complete, then type in 15-20 measurements by hand.
“We were great at sewing and making great-fitting jeans, not so good at data entry.” — Rick Levine, Co-founder, ApparelWerks
Do this for every job sheet and the minutes add up to hours. Rick’s mission was to automate the manual process of getting measurements into the system — without compromising quality, and without adding more tasks and training for pattern makers and sewers.
How Datalab solved this
Rick started with the earlier Google product, but found it required a lot of additional massaging to make the quality work. It didn’t decrease the total amount of manual effort needed to get the data into the system.
Datalab came up in a web search, where Rick saw a press release about Datalab’s performance on handwriting that claimed it beat the industry norm. He tested his hardest job sheets on the hosted API to see what would stick, and it held up.
“You’ve cracked a problem nobody else has solved. What attracted me is your claim to be much better than industry norms for cloud-based handwriting recognizers. And you proved it.” — Rick Levine, Co-founder, ApparelWerks
The new pipeline, transformed by Datalab
Rick had seen costs balloon in the AI and services space, and wanted a pipeline where every API call was deliberate. It’d be fast and simple for the sewer, but tightly qualified so the shop wasn’t paying to process documents that didn’t need processing. What he built stitches Datalab into a workflow that bridges analog and digital while keeping human judgment in the loop:
- Scribble and snap a photo. The sewer finishes QC, takes a photo of the job sheet with their phone and emails it to a dedicated inbox, one only accepting images from the ApparelWerks team.
- A cost gate. The system scans the image at low resolution for visual landmarks, such as a grid in the upper-right corner, the garment QR code, and other identifiable markers. This first gate is a check that decides whether the document is worth running through recognition at all. (The same email pathway is used to associate pictures of finished garments with their database records, no OCR scanning needed.)
- Call Datalab in accurate mode. If the image clears the gate, the system makes an API call to Datalab in high accuracy mode (
processing_mode=accurate) and gets a clean markdown file back. - Parse and convert. ApparelWerks built a parser to read the columns, gridded measurement data out of the markdown, then apply business rules and conversions (for example, turning a height that may be recorded in feet and inches into plain inches for the system).
- Human review. The extracted measurements surface in the admin interface as tentative values, waiting for approval, shown side by side with the original job sheet photo. A reviewer can quickly compare the two and confirm the measurements are correct.
Human oversight very much remains a critical part of their workflow. Two sets of eyes still matter, one at QC and one at approval. What’s changed, though, is that the reviewer’s job is confirmation, not manual transcription. This frees them up to focus on the sewing and pattern-marking itself.
A handwritten ApparelWerks job sheet parsed into structured measurements
Impact
What used to take about five minutes of hunting through fields and typing now takes roughly thirty seconds of visual confirmation. But for Rick, speed is only one part of the benefit.
“It’s as much minimizing the aggravation and perceived friction as it is the time saving and the accuracy. We’re trying to make it as pleasant as possible for the people using it.” — Rick Levine, Co-founder, ApparelWerks
Their team, either measuring customers or working on the sewing floor, will never stop writing numbers and jotting on job sheets; it’s part and parcel of their workflow. Instead, ApparelWerks built a process around them that makes their lives easier. They take a picture, send a quick email, and the data-entry problem disappears. It’s a time-saver and gives back headspace to focus entirely on making clothing, without a tedious, error-prone chore that breaks their concentration.
The economics fit a small shop, too. At a few thousand garments a year and roughly a cent per call, Rick’s usage sits comfortably within Datalab’s free tier.
What’s next
Job sheets were the first problem to solve. ApparelWerks is now exploring Datalab for extracting text and data from patterns and technical drawings, the analog artifacts carrying the shop’s hard-won knowledge. As Rick puts it, it’s one tool in the toolbox, but an easy one to reach for: once he understood Datalab’s API and wrote the calling harness, it became simple to try it on new problems.
For a shop where every inch matters, the goal is to remove the tedium and distractions so the team can spend its attention where it belongs: making great clothing.
You can learn more about ApparelWerks at apparelwerks.com. Explore their custom jeans options at dillonmontara.com.