The pressure to add a battery subsystem to the Direct Solar DC Grinder came from multiple directions. Stakeholders wanted operation during cloud cover. Users asked for it. Competing products had it. From a requirements standpoint, the case looked strong.
The decision to not add batteries was one of the more consequential architecture decisions in that project, and it was not obvious until we mapped the full consequence tree across all four layers of the system.
The Consequence Tree
A direction change in a physical product does not affect just the layer where the decision lives. It propagates. The battery decision touched mechanical geometry, electronics, firmware, and the supply chain and service network. When you look at any one of these in isolation, the cost looks manageable. When you look at all of them together, you see what the decision actually costs.
For the solar grinder operating in remote agricultural settings, the service network consideration was decisive. Lithium cells require specific handling for replacement and disposal. In the operating environments we were targeting, that infrastructure did not exist. The battery would eventually fail. The service event would be a problem with no good solution in the field.
Solving It in the Control Layer
The problem the battery was supposed to solve was inconsistent output during variable solar irradiance. The battery was the obvious solution. It was not the only solution.
The alternative was to solve the irradiance variability problem in the control layer. Variable input, constant output through adaptive motor control. The mechanical and electrical architecture stayed unchanged. The firmware got more sophisticated. The service network problem disappeared entirely because there was nothing to replace.
When to Pivot and When to Hold
The framework I use starts with the consequence map before any direction decision is made. Map the change across every layer. Estimate the cost in each. Identify which consequences are recoverable and which are not. The recoverable ones are risks. The unrecoverable ones are constraints.
What does this change affect in every layer of the system? What does it cost in each? Are there alternatives that solve the same problem with fewer cross-layer consequences? Map this before the decision is made.
The solar grinder shipped without batteries, achieved 90% efficiency, and eliminated 15% of production cost compared to the battery-equipped design it replaced. The consequence map made that decision possible.