Cleaning a list before you pay to print it

A list straight out of a practice or field-service system is rarely ready to mail. None of what follows takes long, and at $1.19 a card a list of two thousand with five per cent duplicates is over a hundred dollars of postage into the bin.

1. Collapse the households

   Four people at one address is four cards through one door, and it reads as junk by the second one. Sort by street address and ZIP together, and decide deliberately: one card per household is usually right for an offer aimed at a family, one per person for an appointment or recall notice that is genuinely individual.

2. Deduplicate on something stable

   Names are a bad key — "Rob" and "Robert" are the same person to everyone except a spreadsheet. If your export carries the record id from the source system, dedupe on that. Failing that, address plus last name catches most of it without merging genuinely different people who happen to share a name.

3. Find the rows that cannot be addressed

   Filter for blanks in the street, city, state and ZIP columns. All four have to be present for a card to be addressed at all, so these rows are the ones that will not mail. Some are quick fixes — a ZIP that lost its leading zero when the file opened in a spreadsheet is the classic — and the rest you simply take out.

4. Watch the ZIPs that a spreadsheet has damaged

   Opening a CSV in a spreadsheet will happily turn 02108 into 2108, and it does it silently across a whole column. Check any ZIP that is four digits long. If a lot of them are, re-open the file importing that column as text rather than a number instead of repairing them one at a time.

5. Decide what happens to addresses that fail verification

   Addresses are checked before anything prints, and you choose what happens to the ones that do not come back clean: have them corrected and retried automatically, or skipped. Correcting catches the ordinary cases — a missing suffix, a wrong ZIP for the street — and skipping is the conservative option if you would rather send fewer cards than send cards you are unsure about. Either way you are not paying to print something the post office was never going to deliver.

Remove the people who asked you not to write

Whatever your system flags as no-contact, do-not-mail, moved or deceased, take them out before the upload rather than after. This is the one category where a mistake costs more than a wasted card.

How small a list can usefully be

Smaller than most people assume. There is no minimum to send, and a first drop of two or three hundred of your best-known lapsed customers tells you more than a broad list of several thousand strangers — you learn whether the offer works before you spend at scale, and a tight list is the cheapest way to find that out.