This is not a relationship advice article. I am not qualified to give relationship advice - ask my wife. This is closer to an engineering write-up. I had a problem (I could not detect when something was wrong until it was already too late), so I did what any reasonably stubborn, mildly systems-obsessed person would do. I stopped trying to become a more emotionally intuitive person, and I built a flowchart instead.

Me, several years ago, treating my marriage like a debugging session.

The bug report

My wife is what you’d call an indirect communicator. She doesn’t like to make things awkward. She’s busy. Something bothers her, and instead of saying so, it gets filed away under “not worth bringing up right now.” Which is reasonable, individually, every single time. The problem is that “not worth bringing up” is cumulative. It compounds quietly in the background, like interest on a loan nobody told you that you took out - until one day the balance comes due all at once, and I am standing there genuinely asking what I did wrong, because as far as my own logs are concerned, nothing happened.

You’ve probably heard the old Men Are From Mars, Women Are From Venus framing - men process by gathering information, women process by connecting emotionally. It’s a useful starting point, but it’s also a population-level average being applied to one specific, particular woman, which is a bit like using the average rainfall of Southeast Asia to decide whether to bring an umbrella to one specific wedding in one specific town on one specific day. Some men are naturally good at the feelings part. I am not one of them. So instead of trying to become fluent in a language I don’t have great aptitude for, I decided to build a translator.

The algorithm, zoomed out

Before we go step by step, here’s the whole thing at a glance - the decision tree I actually run in my head now, most of it in about four seconds, standing in the kitchen pretending to check my phone while actually scanning for signal.

There’s also a permanent background process running the whole time called energy matching, which we’ll get to. Let’s go through the steps one at a time.

Step 1: Wait for calm

Never during chaos. Never mid-toddler-meltdown, never five minutes before either of us has to leave the house. I wait until the baby’s asleep and the apartment has gone quiet. I also put my phone away - not just on silent, but out of reach. I used to think I could scroll and still catch what mattered. I was wrong, every time. Deploy only in calm weather, with your full attention - starting the conversation during a storm, or while distracted, just gives the storm more material to work with.

Step 2: The scan

I ask how her day was. If the answer is “fine,” “okay,” or “good” - that’s not data, that’s a decoy. So I don’t stop there. I ask again from a different angle: how was the commute, how did that meeting go, did the thing with her coworker get resolved. Somewhere in there, if something’s actually wrong, it starts to leak out.

Step 3: Flag the repeat

If she brings up the same topic more than once, even in passing, even as a throwaway line - that’s the flagged signal. That’s the thing that actually matters. Everything else in the conversation can be small talk. That one gets a deep dive.

Step 4: The parrot protocol

This one needs a definition, because I didn’t know the term either until I started reading about active listening. Parroting just means reflecting her own words back to her, reworded slightly, as a question - instead of jumping straight to a response or a fix. It sounds embarrassingly simple. It also works embarrassingly well. Here’s roughly what it looks like in practice:

Her: “Work was just... a lot today.”

Him: “Sounds like today was pretty draining?”

Her: “Yeah - exactly. My manager kept adding things last minute, and...”

Notice what didn’t happen there: I didn’t offer a solution, and I didn’t share my own opinion about her manager. I just fed the sentence back to her, slightly reshaped, as a check. That’s the whole move. It signals “I heard you” without signaling “I’m already trying to fix you,” which turns out to be a very different message.

The fork: listen, or solve?

This is the part that took me the longest to learn, mostly because I got it wrong publicly, more than once. Sometimes she wants to be heard. She already knows the answer, she doesn’t need my input, she just needs to say it out loud to someone. Other times, she genuinely wants a solution. Guessing wrong in either direction goes badly - solving when she wanted to vent feels dismissive, and just nodding along when she wanted help feels unhelpful.

And if it is a solutions moment, there’s a second trap waiting right behind the first one: never give a direct order. “You should just do X” lands as a command, not help - even when X is genuinely good advice.

So before I offer anything, I run one more step first: I ask her how she’d handle it. “How would you approach it?” does two things at once - it hands the decision back to her instead of stacking my opinion on top of hers, and more often than I expected, she already had an answer, she just hadn’t said it out loud yet. If she genuinely doesn’t know, that’s my cue to offer something - but even then, the fix is small and it changes everything: frame it as your own approach, with an exit built in.

“Here’s what I would do. But I have my own strengths and weaknesses, and yours are different. The idea behind it stays the same. But the right action for you might look different from mine.”

Same advice. Completely different reception. One sounds like a verdict. The other sounds like an offer she’s free to decline. And asking how she’d do it first usually means I never even need the verdict - she gets there on her own, and my job shrinks down to just nodding along.

Bonus process: energy matching

This one runs constantly, underneath everything else. If she’s upset or excited, I can’t stay perfectly calm - a flat, low-energy response reads as not caring, even when the intent is the opposite. So I match her energy. This does not mean raising my voice, or copying her anger or anxiety - that’s the one exception to this rule. Instead, I use small signals: leaning in, nodding, saying “mm hmm,” matching how fast she talks. If I’m trying to calm things down, I stay just a little below her energy level. Not far below - just a little.

This also applies to complaining, which I learned the hard way. If she’s venting about something, matching energy sometimes means having something of your own to complain about too - even if, on that particular day, your life is genuinely fine. I have been asked, directly, “why aren’t you even complaining?” That is a real question I have been asked in my own kitchen, and I did not have a good answer ready.

Testing in production

None of this worked on the first try. Early versions of the algorithm were closer to interrogation than conversation, which - unsurprisingly - made things worse, not better. I kept iterating. Every wife runs different software, so whatever version of this you build, expect to redraw large chunks of it more than once.

The one exception that overrides everything

Here’s the part I want to be honest about, because a tidy flowchart that always works would be a lie. Some days, none of it works. You time it right, you scan, you parrot, you pick the right fork, you match the energy - and it still goes sideways, for reasons that have nothing to do with your execution. On those days, there’s exactly one rule left in the system, and it overrides all the others.

THE WIFE IS ALWAYS RIGHT

The whole fallback branch is just: import patience. Not because the flowchart failed - some days simply aren’t a logic problem, and no diagram was ever going to fix that.

Built with a whiteboard, some trial and error, and one (1) supportive but occasionally exasperated wife.