Designing for Manufacturing Reality — Why Products Fail Between EVT and Production
Most connected products do not fail in the lab.
They fail somewhere between the moment engineering signs off and the moment manufacturing tries to repeat that success hundreds or thousands of times. This gap—often labelled EVT, DVT, or PVT depending on maturity—is where optimism collides with reality.
On the engineering bench, the device works reliably.In small pilot runs, everything appears manageable.On paper, the BOM looks acceptable.
And then manufacturing begins.
Suddenly, yields drop. Programming takes too long. Calibration becomes inconsistent. Units behave differently from one batch to the next. What felt like a finished product reveals itself as a fragile prototype that happened to work a few times.
This is not a failure of manufacturing. It is a failure of design assumptions.
Why Manufacturing Exposes Problems Engineering Never Sees
Engineering environments are controlled by default.
Power supplies are clean.Components are hand-picked or carefully assembled.Debug tools are readily available.Engineers understand the system deeply.
Manufacturing environments are the opposite.
They are optimised for repeatability, speed, and cost. Assemblers follow instructions, not intuition. Test operators rely on fixtures, not oscilloscopes. Variability is unavoidable.
When products move into manufacturing, assumptions that were invisible in development become painfully obvious.
The Hidden Assumption: “We’ll Handle That Later”
Many manufacturing failures trace back to small decisions made early with the phrase “we’ll handle that later”.
Later often means:
After the PCB is frozen
After tooling is ordered
After suppliers are locked
After customers are waiting
At that point, change is no longer a design decision. It is a cost decision.
Component Tolerances: When “Equivalent” Is Not Equivalent
In prototypes, components are often sourced in small quantities, sometimes even from a single batch. In production, sourcing realities change.
Passive components vary more than expected. Small shifts in resistance, capacitance, or ESR can push marginal designs into failure, especially in power circuits, sensor interfaces, or RF paths.
Alternate components introduced due to shortages may meet datasheet specifications but behave differently under real conditions. Timing, noise, and startup characteristics often differ just enough to cause intermittent issues.
Temperature and aging effects, barely noticeable in short tests, become significant over production lifetimes.
When designs operate too close to limits, manufacturing variability becomes a failure amplifier.
Programming and Provisioning: The Bottleneck Nobody Plans For
Programming a device once on a desk is trivial. Programming hundreds or thousands efficiently is a completely different problem.
Manufacturing exposes questions that prototypes never had to answer:
How long does it take to program each unit, including retries and verification?
What happens when programming fails midway?
How are device identities assigned and tracked?
How are keys, certificates, or credentials injected securely?
How do you recover a partially programmed unit?
If these questions are not answered in design, programming stations become bottlenecks, yields drop, and costs rise.
From a business perspective, slow provisioning directly impacts throughput and delivery timelines. From an engineering perspective, it reveals whether systems were designed for repeatability or for convenience.
Test Coverage: What You Don’t Test Will Fail First
Manufacturing does not tolerate ambiguity.
A device must either pass or fail. There is no room for “probably fine”.
When test points are missing, signals are inaccessible, or behaviours are hard to observe, test coverage becomes superficial. Units pass basic checks but fail later in the field, where diagnosis is far more expensive.
Common issues include:
No access to critical internal states, making it impossible to distinguish hardware faults from firmware issues.
Tests that validate functionality but not performance margins, allowing borderline units to ship.
Lack of self-test or built-in diagnostics, forcing reliance on external fixtures that grow complex and fragile.
These problems rarely show up in prototypes because engineers compensate manually. Manufacturing cannot.
Calibration: Where Variability Becomes Visible
Many connected products rely on calibration, whether for sensors, radios, power measurement, or timing.
Calibration that is done manually or informally during development becomes a liability in production.
Manual calibration does not scale and introduces operator-dependent variability.
Poorly defined calibration procedures lead to inconsistent results across shifts, locations, or vendors.
Calibration data that is not traceable to device identity undermines long-term reliability and compliance.
From a business standpoint, inconsistent calibration translates into inconsistent product behaviour, customer complaints, and disputes that are difficult to resolve.
Yield Is a Design Outcome, Not a Manufacturing Problem
Low yield is often blamed on manufacturing quality, but yield is primarily determined by design margins.
Designs that assume ideal conditions produce fragile yields. Designs that account for variability produce stable yields.
When yields are low, the costs compound quickly:
More rework
More scrap
More delays
More tension between teams
At scale, yield directly affects profitability. A design that looks affordable on paper can become unviable once real yield numbers are applied.
Firmware and Manufacturing Are Deeply Linked
Firmware is often treated as independent of manufacturing, but in practice they are tightly coupled.
Firmware that lacks production modes forces manufacturers to use development builds or workarounds.
Debug features that are left enabled unintentionally can affect power consumption, timing, or behaviour.
Inconsistent firmware flashing across batches creates subtle differences that are hard to track later.
Manufacturing-friendly firmware is not bloated or complex. It is explicit, predictable, and designed with clear transitions between production and operational states.
The Cost of Discovering Problems Too Late
The most damaging aspect of manufacturing failures is timing.
When problems are discovered after tooling, supplier contracts, or customer commitments are in place, options narrow quickly. Fixes become expensive, risky, and disruptive.
At that point, teams are forced into compromises:
Shipping with known issues
Accepting lower margins
Delaying launches
Reworking hardware under pressure
None of these outcomes are the result of a single mistake. They are the cumulative effect of designing for prototypes instead of for production.
Designing With Manufacturing in Mind From Day One
Teams that transition smoothly into production share a common mindset.
They design with variability as a given, not an exception.They think in terms of repeatability, not just correctness.They treat manufacturing as a partner, not a downstream problem.
This does not mean slowing innovation. It means aligning engineering effort with real-world constraints.
Manufacturing Reality Is a Business Reality
Between EVT and production, products either mature or fracture.
Those that mature emerge more robust, more predictable, and more scalable. Those that fracture often require redesigns that cost far more than early diligence would have.
From a business perspective, manufacturing readiness determines whether growth is smooth or painful. From an engineering perspective, it reveals whether design decisions were grounded in reality.
At EurthTech, we focus heavily on this transition phase because it is where good ideas become real products—or quietly fail. Designing for manufacturing reality is not about pessimism. It is about respect for the complexity of turning one working device into many reliable ones.
Comments