This article is about quality control for data driven programs produced on a web press with in-line finishing and in-line inkjet variable imaging. This great video shows what one of these press configurations would look like. 

Introduction:  addressed direct mail has opportunities for both personalization and customization.

Personalization is created from the individual's data profile - the address block - first name "Dear Maureen" - your relationship, eg: "a member since 1991", your point level, your geographic location, the departments you transact in, etc.  Customization is created from product information, the sales channel, promotional offers and creative. Here is an example showing both.

Maureen buys designer clothes and travels past her local store to go to her favourite regional mall store (as her local store is only a "c" store without the designer department) to redeem her preferred offer of "no sales tax offers" on purchases over $150.  Imagine 2.5 million personalized and customized pieces with 257 national stores and 10 variable offers for each store type; "A", "B", "C". How do you manage the controls for this type of program with so many variables?

The control process involves a series of QC planning steps.

1.  A detailed production grid - which is the matrix with all versions, segments, cells, language and counts and creative or offset versions.

2.  A 'torture' test file that creates a record for each of the variables on the grid. It must include the longest and shortest value in each field and maximum fields of line addresses so you can check the position of the address block.  This is a cross section of fictional people that have been programmed to represent every segment and cell in the complete run.  One  might be Mary Smith, travels over 10 miles to buy housewares - at Store 25, receives offer 57.  Dela Jazman lives within a 2 mile radius of Store 31 and shops for children's clothing and men's fashion, receives offer 22.

3.  Offset and imaging form setups show all layout for all option.

4.  Imaging 'lives', where the printer uses your test file to demonstrate their form setup versions are matching the data file that you approve before production begins.  (This doesn't attempt to go into the data processing prep for hygiene or logic to flag records - that is another topic).

5.  Confirmation of data counts by versions and live records from processed printers data to go back to your programer to check the original client data for their QC check.  Seed names added to the versions you want to get back to you in the mail. The printers data must include a sequential number for every record that is imaged on every piece.

Then you are ready for the production run approval for colour and imaging final ok.  At the beginning of the run (the run starts once you have approved colour) add the 'torture test' data set before the live data again at the beginning of the run.  QC that again against your approved proofs.  Add a pull every nth record where programming has added pull records based on the database that are both pulled during the press run. Don't build in too many pulls for examples 1 sample every 2,000 pieces will generate 12,500 pulls to QC samples.  Pick a number that is manageable based on your run length - here you might pull 1 every 50,000 = 500 pulls.

Here is an example of why you would include this final QC step at the press approval.  This is an issue that really happened. The personalization on press required four inkjet heads. Unfortunately we didn't realize they were out of sync until three months after the selfmailers were mailed at the time of analytics which couldn't be done because the data didn't make any sense.  What happened was, three of the heads personalized the letter version "a" the other head personalized letter version "b".  The program was based on a "gift with purchase" and there was a matrix of five possibilities so everyone got the offer of the person (record) before them.  

At least the program was for women only, so some guy didn't get a free lipstick with his perfume offer. The program went out in the mail.  Imagine the response analysis and not knowing why an individual responded.  It was nothing that six people in data entry couldn't fix over three weeks of visually looking at every single record and manually correcting the data in the files so analytics could report something that made sense.

Before you begin the quality control process ask the supplier for their written process and compare to your project requirements and your agency process and identify the gaps,  Believe me, the supplier will not have a customized process based on the specific requirements of your job as each job has unique challenges.  And there is always time to do it right the second time.