The order matters because practice comes before personalization, and diagnosis comes before repair. Steps one through three give you a baseline run and one named fault; step four changes a single variable against that baseline instead of guessing. Cooking for someone else goes last because by then the dish is stable enough that a distracted kitchen will not undo it.
Step 1: Pick a dish you will eat six times
You will cook this dish four to six times before it counts as yours, so the choice at this stage is really a choice about repetition. Pick something you already enjoy eating, then price the ingredient list before you commit: if the main component costs more than a normal weeknight dinner, swap it for a cheaper cut or a simpler version of the same dish. Braised chicken thighs instead of short ribs, for example. A dish you love but can only afford once a year never reaches step four. Where people trip up is choosing something impressive they have never actually eaten, or a showstopper that only gets made for guests. Practicing a dish you do not enjoy eating makes every later repeat feel like homework, and you will quietly stop.
Step 2: Cook it once with zero changes
First runs are for establishing a baseline, not for showing off. Read the recipe through twice before you turn on the heat, set out every ingredient measured and within reach, and follow the times and temperatures as written even when they feel wrong. If it says medium-high for six minutes, set a timer and do not peek-and-stir. The point is to produce a result you can later compare against, so the second attempt in step four has something to be measured from. Seasoning amounts count here too: measure the salt rather than pouring from the box. The usual slip is improvising on the first run — swapping an ingredient, adding garlic early, cranking the heat because nothing seems to be happening. Then if the dish disappoints, you have no idea which change caused it, and step three has nothing solid to record.
Step 3: Name the one step that broke down
Ten minutes after you eat, while the pan is still in the sink, write one line about the step that gave you the most trouble. Fresh memory is specific; a day later you will only remember that it was "a bit stressful." Be concrete enough that a stranger could act on it. "Searing: the pan was not hot enough, the chicken released water and steamed instead of browning, and I turned it too often" is useful. "Timing was off" is not. Note also whether the recipe itself was unclear — vague phrases like "cook until done" or "season to taste" are worth flagging, because those are the phrases you will have to solve yourself in step four and then state plainly when you write your own version at the end. Vague notes feel like progress but give you nothing to change.
Step 4: Change one variable, keep everything else
Here the earlier work pays off: you have a known baseline from the full run and one named fault from your note, which means you can change a single variable and actually learn from the result. Change only that step. If your note said the sear was weak, preheat the pan a full three minutes, pat the meat dry, and leave it alone until it releases on its own — everything else stays identical to attempt one. Write the outcome on the second line of your notes. The tempting shortcut is fixing four things at once, which turns the kitchen into a guessing game: if the dish improves you cannot tell which change did it, and if it gets worse you now have four suspects. One change per cook, one line per result. Two or three rounds usually flattens the weak point completely.
Step 5: Earn the tweak, change one thing
Only now, with a repeatable base, does a personal touch read as intentional rather than as a patch over technique problems. Pick one element and change it per cook: brown the butter before it goes into the sauce, toast and grind the cumin instead of using it pre-ground, finish with lime zest rather than juice, or rest the dish five minutes longer than the recipe says. Test it, then decide whether it stays. The reframe worth holding onto: a signature comes from a small change you can repeat a hundred times, not from a big change you can only pull off once. Adding three flourishes at the same time is how a working recipe turns into something you cannot reproduce — the next attempt tastes different and nobody, including you, knows why. One touch, tested, kept.
Step 6: Cook it for a guest, then write it down
Cooking for a guest is a genuinely different test, because you are talking, pouring drinks, and losing track of the timer — exactly the conditions that expose which parts you actually know versus which parts you have been reading off the page. Pick someone who will tell you the truth rather than someone who will be polite. Serve it, then write the one-page version the next morning: ingredients with real amounts, oven temperature, pan size, and the three or four timing sentences that matter, in your own words. Include the fix from step four so it does not get lost. Do not just bookmark the original recipe. Links rot, and a saved page never contains what you learned. What goes wrong here is treating the guest dinner as the finish line and never writing anything down, which means the dish quietly reverts to a stranger's recipe six months later.
A phone reminder vanishes the moment it is dismissed and leaves no trace of what you did. A paper strip taped inside a cupboard door stays visible, gets shorter each time a finished step is cut off, and is still there the next time the pan comes out. Keep the one-page recipe folded with whatever strips remain.