A spreadsheet called final_final_v3
The charity's appeal mailing is due to drop on Monday. The data file arrives on Thursday afternoon as a spreadsheet exported from their CRM. Some rows have the title in the first name column. Some addresses have the town in the postcode field. There are records with 'DO NOT MAIL' typed into the surname. A handful of names contain accented characters that come out as question marks in your variable data software.
Your data person spends Friday cleaning it by hand, emails questions the customer does not answer until Monday morning, and the job runs with the press waiting. A week later a supporter rings to complain that they received three letters, one addressed to their late husband.
Why the data is the weak link
Print for personalised mailings is well understood. The data is not, because it comes from the customer's systems, exported by someone who does not know what a mailing house needs, often at the last minute.
- Every customer's export has different columns and formats.
- Checks are done by eye in a spreadsheet, so they depend on who does them and how rushed they are.
- Duplicates are hard to spot when names and addresses are written slightly differently.
- Suppression instructions, such as people who asked not to be contacted, arrive as notes rather than data.
- There is no record of which data version was printed, so complaints are hard to answer.
Data protection sits on top of all of this. What your business must do with personal data is a matter for your own policies and adviser, but a process that handles it the same careful way every time makes those policies much easier to follow.
When the merge goes wrong
A bad merge is a reprint, a postage cost that cannot be recovered and a customer who looks careless to their own supporters or clients. Duplicate letters waste print and postage. Personalisation errors, such as the wrong salutation, cause complaints that land on your customer and then on you. And the Friday scramble to clean data by hand is the kind of pressure where mistakes happen.
A data intake built for mailing jobs
- Customers upload data through a secure page tied to the job, not by email attachment.
- Column mapping is saved per customer, so a regular customer's export is recognised and mapped the same way each time.
- Automated checks run on upload: required fields present, postcode format, obvious field swaps, characters your variable data software cannot print, and likely duplicates using fuzzy matching on name and address.
- Suppression files and instructions are applied as data, with the removed records listed.
- The customer receives a plain report of what was found and chooses how to resolve each issue, with a clean file produced from their decisions.
- Merge proofs are generated from a sample of real records, including the awkward ones such as long names and missing lines, for the customer to approve.
- The printed file is stored with the job, with a record of every change, and deleted on the schedule your retention policy sets.
| Check | Done today by | Done by the intake |
|---|---|---|
| Missing or swapped fields | Eye, in a spreadsheet | Automatic on upload |
| Duplicates | Sorting and scrolling | Fuzzy matching, customer decides |
| Unprintable characters | Found on the proof | Flagged before merge |
| Suppressions | Applied by hand | Applied and listed |
| What was printed | Unclear | Stored with the job |
A calmer Friday
Data problems are raised with the customer as soon as the file arrives, not the day before print. Your data person checks exceptions and decisions instead of scrolling rows. Merge proofs show real awkward records, so the customer approves what will actually print. When a recipient complains, you can show exactly which record produced their letter and where it came from.
If you use a mailing sortation service or postal discount scheme, the clean file is also a better starting point for that step, because it has been checked the same way every time.
Do mailing files cause you stress?
- Data files arrive late and need cleaning by hand.
- You have printed duplicate or misaddressed letters.
- Customers send data by email attachment.
- Merge proofs only show neat sample records.
- You could not quickly show which data version produced a complaint.