Every layer of management compresses information. By the time a problem on the front line reaches a founder or executive, it has been through one or two rounds of "let me see how to frame this." That is not malice; it is the natural compression of bad news travelling upward. The result is that leaders consistently overestimate how well things are going.
The compression is mechanical, so no amount of "my door is open" fixes it. Each messenger softens by maybe 20%, out of self-preservation, loyalty to their team, or simple summary. Two layers of 20% softening turns "this is broken and customers are noticing" into "some friction we are actively managing." Every leader believes their culture is the exception; the failure mode is universal precisely because everyone believes that.
The cheapest fix
Talk to people two levels below you, regularly, in unscripted settings. Not skip-level surveys, not engagement scores, but actual conversations with engineers, support reps, account managers, anyone close to the work. You are not bypassing your managers; you are calibrating your model of reality.
The mechanics matter less than the consistency: thirty minutes, one on one, no agenda document, no manager present, and a cadence you actually keep (a couple per week reaches an entire mid-sized org each quarter). Tell your directs plainly what you are doing and why ("calibrating, not auditing you"), because a skip-level programme discovered by accident reads as surveillance. And protect the sources absolutely: what you hear travels upward into patterns, never sideways into "someone on your team said."
What to actually ask
Forget "how is it going." Ask: what would you change if you had a magic wand for one day? What is the most frustrating part of your week? What is the thing nobody is talking about that everyone knows is broken? Then listen, do not defend.
Two more that reliably open doors: "what are we pretending works?" and "what did you stop raising because nothing happened?" The second is the most important question in the set, because its answers are a map of where your escalation paths already failed. Expect the first conversation with anyone to be guarded; the honest material arrives in the second and third, after they have watched what you did with the first. The listening discipline is absolute: the moment you defend a decision or explain the context, the channel you are trying to open closes, politely, forever.
What to do with what you hear
Almost nothing immediately. The point is calibration, not action. Patterns across many conversations are the signal; one-off frustrations are noise. Over months, the picture you build is more accurate than any dashboard your leadership team produces.
The "almost" carries real weight though. When a pattern does emerge across five conversations, act on the pattern at the system level (fix the process, fund the tooling, adjust the priority) without attributing it to any conversation, and let the fix be visible. People talk; within a quarter the whole company knows that things said in skip-levels quietly get fixed, and the quality of what you hear compounds. One caution for founders who love the practice: undermining your managers by second-guessing their calls from skip-level anecdotes destroys more than the calibration gains. The information flows up to you; decisions still flow through them.