It is the question I get most often, and the mistake hiding behind it is always the same one. People assume improvement scales with volume, that 100 puzzles a day must beat 15. It does not. If anything the person doing 100 is often the one who has not improved since March.
The honest answer
Ten to twenty a day, solved properly, for most players.
Solved properly is doing all the work in that sentence. Fifteen positions you calculate to the end and then review will beat fifty you clicked through on instinct, every single time, and it is not close.
Why more is not better
Past a certain point the extra puzzles stop teaching you anything and start teaching you something worse.
You begin guessing, because volume rewards speed and speed rewards the move that looks right. Do that for six weeks and you have trained a fast guesser rather than a strong calculator. Then you stop reviewing, which is where the actual learning lives, because nobody on this earth reviews 80 puzzles in an evening. And you get tired. Tired calculation is sloppy calculation, and the last forty positions of a marathon session go in one eye and out the other.
I have seen the puzzle counters. Someone with 40,000 solved and a rating that has not moved in two years is not an unusual thing to meet. They did the reps. The reps were the wrong shape.
Make each puzzle count
A puzzle is a rep, and reps only build anything when the form is right.
Calculate to the end before you touch a piece. Not “this looks right”, the whole line, including the move your opponent would actually play. Check the best defence rather than the one that lets your idea work, which is the single most common way a solved puzzle turns out to have been a lucky one.
When you miss, sit with it for thirty seconds and name the reason. Was it the wrong pattern, or the right pattern and lazy calculation? Those are different diseases and they have different cures. Skipping to the next position is the most expensive habit in amateur chess.
And keep the difficulty honest. Around a 70 to 80 percent solve rate is the zone. If you are getting nearly everything, the trainer is flattering you.
Consistency beats marathons
Fifteen a day for a month is worth far more than 450 across one heroic weekend.
Tactics live in long-term memory, and long-term memory is built by spacing. Seeing the same shape on Monday, then again on Thursday, then again a fortnight later is the mechanism. Cramming it once is not a weaker version of that. It is a different thing that mostly does not work.
So the routine is dull, and dull is the point. Fifteen focused positions. A couple of minutes on whatever you missed. One weak theme drilled deliberately two or three times a week, chosen from your worst accuracy score rather than your favourite tactic. That is the whole prescription, and it fits in the time most people spend deciding what to watch.
The dose is only half the question, mind. The other half is which mode you solve in and what the trainer does when you get one wrong, which is what how to use a chess tactics trainer properly is about.