I think the fix here is real but the diagnosis is slightly off. The problem was not the data, it was the sampling rate. Daily numbers on a small site are pure noise and noise is stressful precisely because your brain will find a pattern in it.
I look once a month now, in a spreadsheet, one row per month. Same information, none of the damage, and I still catch the things worth catching. Deleting it entirely also means you cannot tell the difference between "nobody came" and "the site has been down for a week", which does happen.