Using Three Mirrors as a Review Template Beats Yelling in Meetings

These past few days, while doing a chip project retrospective, I ran through two approaches: the company's common "Problem-Cause-Measure" table, and the "Three Mirrors" framework from a Huxiu article about Suning.

The Three Mirrors is a retrospective framework derived from the saying: "With bronze as a mirror, one can adjust one's cap and robe; with history as a mirror, one can understand rise and fall; with people as a mirror, one can clarify gains and losses." At first listen, it sounds like chicken soup (fluff), and isn't easy to use in engineering. Still, I built it into three columns: Bronze Mirror for external performance, History Mirror for timelines and cycles, and People Mirror for feedback from others.

Standard tables easily slide into "the market is bad," "the boss decides quickly," or "the team wasn't aligned." In my run-through, at least two lines out of three were adjectives. Power wall, simply put, means that when frequency goes high, power consumption and heat can't be suppressed; this needs to be solved with specific metrics. The Bronze Mirror column forces me to lay out measurable indicators like target power, actual power, area, delivery milestones, and customer feedback.

The History Mirror made me pause. It requires viewing the project within the industry cycle, not just staring at a single node. For example, when this process node first appeared, supply chains and customer expectations changed. Initially, I wrote it as a chronicle, which felt vague. Later, I changed it to "when key variables changed," which became clearer. The Suning example was striking. The article stated that an e-commerce platform carrying the ambition of "Walmart + Amazon" is now being transferred for no less than 1 RMB, with negative net assets in the billions. This example reflects a misjudgment of the cycle. Once scale stopped growing, cash flow and asset quality issues were exposed.

The People Mirror is the softest. Feedback is often unstructured. Customers say "versions aren't stable enough," engineers say "requirements keep changing," testers say "versions are too late." I categorized by role and found that only a few items could actually be implemented. Standard tables treat opinions as to-do lists; the Three Mirrors are more like a glass pane, acknowledging that some problems can't be solved by adding a few measures.

The conclusion is: it depends. If it's just delivery delays or a few defects, standard tables are faster; don't use the Three Mirrors. If it's a retrospective for advanced nodes, multi-team projects, or those with significant external market changes, the Three Mirrors are more useful than yelling. They force stratification, separating facts, time, and feedback, but they rely on the user's self-discipline and can easily devolve into grand narratives.

Actionable advice: Don't use it for annual summaries. First, try it on a small failed module. Put only verifiable metrics in the Bronze Mirror, only external variables in the History Mirror, and only original quotes and evidence in the People Mirror. The column you can't fill is where you haven't truly thought things through.