The order matters because each step hands the next one its raw material: fixed combos give the tasting card something real to compare, names make those rows worth rating, and only a proven favorite earns a blind taste test.
Step 1: Choose your seven pairings
Everything downstream depends on these seven choices, so settle them on paper before you shop. Aim for three safe pairings you already trust (lemon-ginger, cucumber-mint, lime-mint), two familiar fruits with an unfamiliar herb (strawberry-basil, watermelon-rosemary), and two wildcards you have no confidence in (peach-thyme, blueberry-sage). The spread matters because step 3's tasting card only teaches you anything if the entries taste genuinely unlike each other. Write each combo on a strip of paper and tape it inside the cupboard where the glasses live. Where this falls apart: picking seven variations of citrus water, which collapse into one drink and leave you rating the same flavor seven times.
Step 2: Brew and name each one
Brew the pitcher in the morning, before thirst decides what you drink, and give it a name before you taste it. Muddle first: press the basil or mint against the pitcher wall with the back of a spoon for fifteen seconds until it smells sharp, then add fruit and cold water. Skip that press and you get faint, grassy water that tastes like the fruit was rinsed nearby. Let it sit twenty to thirty minutes in the fridge, then strain, because that window is roughly where flavor peaks. Naming a batch (Basil Bog, Cucumber Static) turns it into something you can rate and defend in step 4 rather than water you happen to be drinking. The mess-up here is pouring and walking away; an unbruised sprig steeps to almost nothing.
Step 3: Rate each invention honestly
By now you have names and several days of pitchers, which is exactly what makes a card worth keeping. Use a lined index card, one row per invention: name, date, rating out of five, one note. Record steep time alongside taste, because herbs turn bitter as they sit, and the same recipe can score a five at thirty minutes and a two at five hours. A rating with no time attached tells you the drink was bad without telling you whether to change the recipe or just drink it sooner. So write "45 min, still bright" or "3 hrs, bitter finish." A frequent slip: logging only numbers, then feeling like the ratings contradict each other when the real variable was how long the pitcher sat.
Step 4: Serve blind, take the reaction
Take the highest-rated row from that card and pour it for someone without explaining anything. A guest's unguarded reaction is the first evidence about your water that did not come from you, and steps 1 through 3 have given you a settled favorite worth testing. Serve it cold, in a clean glass, with no fruit floating on top to give it away. Ask what's in it and write their first two guesses down word for word; if they name basil and strawberry unprompted, you have a recipe worth keeping. The usual way people wreck this: hinting ("there's something green in there") or apologizing in advance ("it's a little weird"), both of which steer the guess and teach you nothing about the drink.
Step 5: Post your top three
Three days of ratings and one blind taste have shown which inventions actually hold up, so promote only those and stop brewing the rest. Write a small menu card with three names, a brew time for each, and a one-line recipe under each name, then tape it at eye level beside the fridge handle. The value is that it removes tomorrow's decision, and that decision point is where plain water or something sweet wins by default. Add a tally mark each time you make one, so the card doubles as a count of what you're actually drinking. Where people go wrong: building the menu too early, or picking by which name sounds best instead of which row scored highest.
A reminder on a phone disappears the instant it is dismissed and leaves no trace of whether you followed through. A paper strip taped where you will see it stays put, visibly shortens as each finished step is cut away, and only comes down when the last one is done.