How Santa Cruz Bicycles Standardized High-Mix Frame Assembly with Connected Torque Data

When Matt Shirley started looking at increasing assembly consistency on his production line, he was not trying to launch a broad manufacturing execution system project. He was trying to solve a narrower problem on the frame assembly line at Santa Cruz Bicycles; the bikes were coming out correctly, but operators were not always building them in the same sequence or with the same level of efficiency.

"They all have the correct torques and all the pieces are put on them in the correct places," said Matt, Production Engineering and Quality Manager at Santa Cruz Bicycles. "But it can be kind of a problem when you're trying to get people to all work with the same amount of efficiency."

Then he named the larger concern: "There's a little bit of quality risk there when not everybody is doing the exact same sequence."

Santa Cruz was standardizing a process that produced built-to-specification bikes but depended too heavily on operator memory and personal sequence.

During the first 15 months with PICO, Santa Cruz had a tool-connected frame assembly station, model-specific digital work instructions, captured torque results, and a few hundred process changes behind it. The immediate value was practical: tighter control over the work, evidence that required torque targets were met, and better data for the next continuous improvement process decision.

Every Bike Down the Line Can Be Different

Most assembly-line intuition assumes batches: run one model, change over, and run the next. Santa Cruz does not work that way.

"We don't build in batches," Matt said. "Each bike during the day could be a new or different bike."

In that high-mix environment, small decisions compound quickly. At the e-bike crank station, Santa Cruz had used seven preset torque wrenches rather than resetting a one-click wrench for each bike.

"We don't want our operators to have to go and twist their click wrench depending on whichever bike is coming through," Matt said.

The operators were good at keeping track of which wrench belongs with which build. But the process still depends on the operator correctly matching the bike to the tool every time.

With Pico’s scan-to-start functionality for automatic process selection, the operators no longer needed to focus on which sequence of steps and torques were correct. Once the build was identified, the system selects the correct torque program for the configuration.

"They don't have to think about it," he said. "I don't worry about it. I know that they're just going to use the correct tool."

This principle is already operating at the connected frame assembly station. The product configuration determines the instruction path, the connected tool returns the fastening result, and the normal workflow does not allow the required torque step to be completed until the result is recorded.

Enforced Sequence Created Feedback Paper Never Did

Matt managed a few hundred process changes during the first 15 months. Much of that activity came from learning how people actually performed the work.

"I can write an SOP that they print out and that they're supposed to follow," he said. "But no one's forcing them to follow that paper SOP. That thing doesn't control their sequencing. So I don't get the feedback from them like I do when we force them to use the tools in a certain sequence."

With paper, an experienced operator can jump ahead, return later, and still finish a correct bike. The process owner may never learn that the written sequence does not match the floor. A controlled digital workflow makes the mismatch visible.

A sequenced process gets argued with, and those arguments can become visible and useful process data.

Matt described one person asking to move a step while five others preferred it where it was. The system made the disagreement visible; the team still had to decide which method should become standard work.

One change was especially concrete. Above a certain torque, the connected tool could react strongly enough that the operator needed both hands to control it. Matt split the operation into two steps: snug and seat the bolt while steadying the product, then release the product and use both hands for the final torque rundown.

That was not a dashboard insight. It was an ergonomics improvement uncovered by closely documenting and controlling the job to see what the operator actually needed.

Put the Decision in the Process, Not at the Bench

Matt's goal was not to compensate for careless operators. It was to remove decisions that the process could directly build in more reliably.

When the product identification determines the torque program, the operator does not have to carry a lookup table in memory. When the tool returns the result directly, documentation becomes part of the work rather than a second scribing task performed afterward. When instructions are tied to the product and its configuration, each station no longer depends on a single generic SOP that covers every variation with ‘if this, do that’ statements.

For a manufacturer trying to do the same thing, the sequence Matt followed is repeatable:

• Start with the joints where a miss is expensive. Critical torques on drive units, cranks, and structural fasteners warrant the use of connected tooling. Low-consequence hardware usually does not.
• Tie the torque program to the product, not the station. Scanning the unit or job lets the configuration select the program for each variation, which removes the lookup step rather than making it easier.
• Make the tool result the condition for completing the step. If the workflow can advance without the result, the record requires documentation. If it cannot, the record becomes the direct control mechanism.
• Keep instructions model-specific where products look alike. A generic SOP is most dangerous when two models are hardest to tell apart at the bench.

This is the same logic behind error-proofing and poka-yoke. Design the process so the wrong outcome is difficult to produce, rather than relying on inspection or recall to catch it later.

That last point mattered at Santa Cruz. A Bronson and a Nomad can look similar at the bench while still requiring different details. The same model can also have configuration differences. Matt gave the example of a Bronson sold with a shock versus one sold without it. Before using product variants, he treated those as separate products but with difficult-to-distinguish differences. With variants, he could maintain a single Bronson product line and change only the steps that differed for automatic selection during assembly.

The process owner still has to define the sequence and maintain the variants. The difference is that the configuration logic resides in the initial setup rather than in operator memory.

That does not make mistakes impossible, and Santa Cruz has not yet measured whether the model-specific instructions reduce onboarding time. It does remove common sources of variation and makes a missed requirement harder to pass forward unnoticed.

Torque Assurance Without Routine Log Review

When asked what justified the system, Matt did not point to a dashboard, a calculated return on investment, or a reduction in warranty claims. He pointed to confidence.

"It's me having the confidence that the torque was achieved properly," he said. "I don't even look back at the logs, because they just can't go on if they don't hit the torque value that we need."

At the connected station, PICO captures the fastening result as part of the build record, without a separate documentation task. No customer issue had required Santa Cruz to review the records, and Matt was not routinely auditing them. The record exists because the work was completed, not because someone remembered to create it afterward or needed to.

When the Whiteboard and the Data Do Not Agree

Santa Cruz tracks production against a daily target. At stations still using click wrenches, operators record the count manually. At the PICO station, the system already knows how many builds were completed and how long the recorded processes took.

That difference became important when lower post-pandemic forecasts led the team to consider combining stations. Doing that well requires a reliable and accurate view of actual cycle time.

"What we're seeing with the data is that we're just not there," Matt said. "When we have them just keep a tally on a whiteboard, they're hitting the numbers that they need to hit. But then when we start to look at the cycle times in our log, they're off, and they're off like a concerning amount."

The data did not prove that the whiteboard counts were false or that anyone was overstating production. The measures came from different stations and operating methods, and Santa Cruz was still working out how to combine multiple processes into a single station-level cycle time.

What PICO exposed was a discrepancy worth investigating. That matters because Santa Cruz found it before redesigning work around a baseline the team assumed it understood and identified true cycle times rather than the scope of work and production requirements naturally fitting into the allotted time.

Before committing to a line change from an unverified baseline, it is worth checking a few things:

• Define what a station-level cycle time includes. If a combined station covers two to four processes depending on the product, that boundary has to be settled before the number means anything, and if any work content overlaps
• Measure the operation without an observer. A stopwatch study captures an idealized rhythm; recorded process time captures the ordinary one. Automated time studies remove that observer effect.
• Investigate the gap before you design around it. A discrepancy found during planning is a question. The same discrepancy found after a layout change is rework.

Operator Adoption Was Uneventful, Not Effortless

Matt described the collective operator reaction as "surprisingly indifferent."

Some operators still preferred click wrenches. Others questioned why an experienced person needed so many instructional slides and why the system required a particular sequence for work they had done for years.

Santa Cruz uses a pro mode that can reduce the instructional content after an operator demonstrates the defined sequence. At the time of the interview, no one had moved into that mode because the team was still seeing out-of-sequence work on slides without the required action.

That is not simply a discipline issue. It may show that the documented sequence needs work, or that the sequence needs to be controlled. Either way, the disagreement is visible.

A First-Time MES User Configured Most of It Himself

The evaluation was intimidating because Matt had never run an MES and was investigating connected torque tools simultaneously.

"It can be really daunting trying to look at a system like Pico when you haven't ever run one before," he said. "Because you know so little, it's really hard to understand what you need. You can't pick between products if you don't know how to use them in the first place."

A tool distribution partner that understood both sides pointed him toward PICO. Matt then configured the station, built the model-specific instructions, set up variants, and managed most changes himself.

"There's no point where I was so stressed out that I was like, I need one of these guys to come here to the factory to set this up for me," he said. "The interface is just so straightforward. I didn't need much hand-holding."

Santa Cruz still used support. Matt said they submitted relatively few help-desk tickets and never needed an on-site implementation team. The deployment was largely self-managed, not support-free.

His advice to another first-time buyer was direct: "Just pick Pico, because the interface is so easy to understand. We've got budget constraints, but really, it's not expensive."

That pattern is not unique to a company of Santa Cruz's size. A 12-person Nebraska engine shop reached a similar conclusion running most of its builds on the free plan.

Where Santa Cruz Plans to Go Next

The case shows what a manufacturer can learn before a large rollout, how the process can be standardized, tool data captured automatically, written sequences improved, and production data trusted for the next decision.