Be Social

+

Complex variable project – Part 2

July 6, 2012


So the project is finally done and I can talk about it a little bit.

When I started out I thought this project would be the second most complex job I ever worked on.

Data
A list of about 5000 targets was used. Specialized list – Physicians and psychologists.
The list was postal sorted, split into English and French because of different return envelopes.

There was a high number of variables for everything from language to practice profile, geographic representative territory and location of the physician that came into play. Since the program runs for nearly a year with at least 5 iterations, the analysis program was completely automated and tested, the “CRM” part of the process now takes around 20 minutes. Initial programming took two days. This came in handy when Transcontinental returned the files incorrectly not once, but twice. Being able to verify the data and quickly say “It’s good” or not was priceless.

Creative
The ingenious creative called for local pictures to appear. So that if you lived in Montreal, a background picture of Downtown Montreal would appear, and if you lived in a farming community, it would be a less urban photo. A storyline would walk the physicians through the benefits of the product, and allow them to request a sample.

Programming
In all a matrix of over 300 background images split among 5 waves was used. Some pictures were used more than once, others were used only once. These 5 waves each discussed different topics, but the background images and protagonists would change based on local elements, the rural or urban nature of the profile, and the socio-demographic makeup of the physician’s practice. A predominantly blue collar area would receive a storyline of a line foreman, construction worker, or other blue collar profession, while a predominantly white collar practice would receive a storyline of an office worker, travel agent, etc. In all, ten separate stories were written, each discussing the benefits of the product from a different angle and featuring a different protagonist while keeping the communication appropriate and relevant to 1. Location, 2. Practice socio-demographic profile, 3. Physician, 4. Representative, 5. Provincial reimbursement percentages and, of course, language preferrence.

Because each layout had some common elements we were able to reuse the basic template. The text, however, changed completely from one mailing to the next, so the layout program had to be rewritten and re-validated with every mailing. This took about 4 hours per campaign.

Quality Assurance
By far the biggest challenge in this project was quality assurance. With the number of variables, the number of variable elements, multiple layers with visibility, transparency, etc. it was imperative that the layout validation be very thorough. Fortunately, we have a multidisciplinary team. I’m very good with data, Rose is very good with text, and Willy is very good with creative verification. However, the high number of permutations still meant we had to look at over a hundred pieces to ensure the layout program was performing correctly. This took probably 80% of the time we spent on this project. 4 hours to program, and 2 days to verify. In reality it was a bit more than that because we had 3 people verifying the various permutations, so it’s more like 6 man-days.

In the end, though, the program was validated and it was time for production of the print-ready PDF files.

Technology Issues
That’s when we hit a little snag. Production time was incredibly long. 100 records took about an hour to process. The combination of 3 high resolution variable images full-sheet variable images on top of which there were other variable images with transparency for the representatives meant that XMPie could not reuse graphic elements as it usually does. By changing the design ever so slightly we managed to reduce that somewhat by placing the transparent images on a solid colour background instead of a variable image. This reduced the production speed to about 35 minutes per 100 records. It was still awful. Because the desktop version of XMPie only allows one production file at a time, and it can’t automatically break up a list of 5000 into 50 chunks of 100, I was stuck launching a new production file manually every 35-40 minutes for 2 days straight, pretty much. The job had to be cut into 100 record files because the postscript file, which is the intermediary step for XMPie’s technology weighed in between 4GB and 8GB per 100 record chunk. That’s about as big as we can make it and the file will still work at the printer’s RIP, albeit slowly.

For subsequent jobs, we used the server in France; something we said we’d never do because it takes so long to transfer files from France, but the PDFs are created 4 at a time on the server, and it’s able to automatically break up the file into manageable chunks. That means you just set it and forget it… Well, theoretically, anyway. Unfortunately the server locked up twice so far, and we had to restart at least three times because the job is so demanding that the server would lose track of something and just stop substituting images into the documents.

Future Directions
So these were some of the technological challenges we faced on this job. I recently ran into some of the folks at GMC Technology while I was speaking at Interquest in Toronto, (see post here) and I discussed this project. They suggested they should be able to significantly improve production speed because they use a proprietary file production method that doesn’t use Adobe’s slow Distiller process. I’m eager to test it with them to see what it can do. I’ve used GMC’s software before, when I was at Transcontinental. It really is top notch software, though it’s very expensive, and I’m not sure that we have the volumes to justify the expense, but that’s a decision for our president. My job is to validate that the technology provides superior results. Then we can run the numbers and see if it’s worth upgrading to it from XMPie. The fact is that XMPie’s technology has given us so much grief and headache that some of that expense may be justified by lost time on our part.

Anyway, the technology choice is to be continued, I’ll keep you up to date as it happens..

As usual, don’t hesitate to contact me if you have questions about marketing, or the technological processes that are necessary to make relational marketing work. I’d be glad to help.

Tagged with:

Leave a Comment

Please keep in mind that comments are moderated and rel="nofollow" is in use. So, please do not use a spammy keyword or a domain as your name, or it will be deleted. Let us have a personal and meaningful conversation instead.