Three messy sheets, cleaned.
Same engine that runs in the workbench. Names and numbers are anonymised; the shapes of the mess are exactly what we get every day.
1 · A contact list copied out of three different phones
2,140 rows in
1,986 clean rows
154 duplicates merged
37 rows flagged for a human
Before
name,手机号,城市 张伟,+8613800001111,河南 濮阳 张伟 ,138-0000-1111 , 濮阳市 李娜,1380000111,郑州 王芳,１３８００００２２２２,郑州 赵磊,12345,安阳
After
name,phone,city,flag 张伟,13800001111,濮阳, 李娜,,郑州,phone has 10 digits 王芳,13800002222,郑州, 赵磊,12345,安阳,phone has 5 digits
Two “张伟” became one row instead of two calls to the same person. The two broken numbers were kept and labelled, not silently dropped.
2 · Order data with money written five ways
860 rows in
860 rows out
amount → real numbers
212 issues listed
Before
order,date,amount A-01,2024/3/7,¥1,280 A-02,2024年3月8日,1280元 A-03,2024.13.45,abc A-04,45000,￥98.5 A-05,,1 200
After
order,date,amount A-01,2024-03-07,1280 A-02,2024-03-08,1280 A-03,2024.13.45,abc A-04,2023-03-15,98.5 A-05,,1200
Columns became sortable and summable in Excel — the whole point. The impossible date and the “abc” amount are flagged by row number so someone can call the customer.
3 · A 2,000-row survey with invisible junk
2,004 rows in
1,958 rows out
46 empty rows removed
0 manual fixes for format
Before
id,score,email 1,８,user1＠mail.com 2,7,user2@mail.com 3,,user3@mail.com 4,9,USER4@MAIL.COM
After
id,score,email 1,8,user1@mail.com 2,7,user2@mail.com 3,,user3@mail.com 4,9,user4@mail.com
Full-width score digits and the invisible character after the second email are the classic reasons a “clean” file still fails to import. One pass fixed both.