Merge fields that do not embarrass you
Lesson 4 of 6 · Mail that gets a phone call
1 · Learn the move · Verified-data override
The mail house prints exactly what the export says, and county rolls are written for the collector, not for a letter. 'KESSEL RUTH A ET VIR'. 'UNKNOWN HEIRS OF'. 'HARROW RIDGE HOLDINGS LLC C/O TAX DEPT'. An acreage of 0.00 on a split parcel. Ask a model to clean the names and it will guess: 'ET VIR' becomes 'and family', a first name appears where there was an initial, 4.87 acres rounds to 5. Every guess is a letter that tells the owner you did not look. The export is the verified data, and it overrides the model's sense of what looks right. So the model transforms fields only by rules you write, and every row no rule covers is flagged and held, never repaired. The flag list is the product. A hundred held rows is a good morning's work; a hundred confident repairs is a hundred wrong letters.
Here are 12 raw owner rows from my export and my merge rules. Apply only my rules. Any row my rules do not cover gets FLAG and a reason, not a fix. Never invent a first name, never round acreage, never drop a C/O line, and never build a greeting for a deceased or unknown owner.
2 · Your turn. You write the prompt
Your export for a 700-piece drop in Sayle County has owner names in roll format, a mix of couples, entities, estates and some rows with no name at all. The last drop went out with 'Dear Et Ux' on nineteen letters. Write the prompt that fixes the fields without inventing any.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.