Gain Confidence in Your Matchback Solving the Attribution Puzzle

By Jim Coogan

Catalogers rely on matchbacks to compare catalogs mailed against the response rates from their various housefile and prospecting list segments. However, matchback attribution rates, which are the percentage of orders that can be allocated back to a catalog mailing, are decaying. It’s important that the attribution of orders can be as high a percentage possible. When attribution rates are low, it’s difficult or impossible to know the true demand that’s coming from your catalog.

How can catalogers determine the reasons causing low attribution rates in matchbacks? Consider implementing the following tactics:

Run a matchback requiring an address match rather than a name or household match. Some mailers use a multipass allocation, first matching against individual or household, then matching against address.
Before beginning any of the legwork, make sure all files being used for analysis are DPV/ZIP4 verified.
Compare the match keys between the merge and mail files, and the match keys in the matchbacks. Differences in the match keys and tight vs. loose match key logic may account for orders that don’t match.
Pull a sample of orders that aren’t attributed in the matchback and compare them to the mail file. If you pull a geographic area, you can compare the unallocated orders against the mail file and see which orders should have matched the mail file.
Take the unallocated orders from a matchback and put them in a merge purge as the lowest priority.
Run a matchback internally or using a different service bureau or co-op database to see how another matchback vendor’s data looks and its attribution percentage.
Compare orders against the entire merge and see which ones may be coming from unmailed segments. If some of the unallocated orders fall into the unmailed segments, check if those segments should be mailed because you have a lot of response coming from unmailed segments.

What are the reasons that emerge when you look at why orders can’t be allocated back to the mail file?

Some address correction may be on the mail file but not on the matchback file, so look at the ACS on the sets of data being compared. Just over 1 percent of consumers move each month. This can be up to 15 percent of the unallocated orders from the past year alone because some housefiles aren’t updated with NCOA changes. If you’re not updating your housefile with NCOA changes you may find a surprising number of unmatched orders.
If it’s a matchback based on name and address match, the name may not match exactly or a married couple may have two last names. Run a matchback requiring an address match rather than a name or household match. There are a lot of variations in names, but much less variation in addresses between the mail file and the order file.
Addresses may not be standardized or the fields for addresses may not align between the mail files and the order files.
Address hygiene issues, including missing or incomplete ZIP codes.
Duplicates where a customer has ordered twice. Is the second order from that customer being allocated?
Retail orders are notoriously suspect because associates often don’t collect all the right address information. Look at all of your unallocated orders and check to see how many are close to your retail store locations. Consider segmenting reporting within your retail store market areas and outside your retail store market areas.
Orders outside the date range of the matchback. The rule is for the matchback to look backward 120 days to previous mail files if an order doesn’t match the most recent mail file.
Bill-to vs. ship-to addresses where the mailing is to the bill-to address and the order came from the ship-to address, or vice versa.
Actual new-to-file customers that came from channels other than a recent catalog mailing. Compare your unallocated orders to new-to-file customers to see how many new customers are flowing from noncatalog sources. The flood of affiliate orders is a common reason for transactions that can’t be allocated to a mail file.
If you have orders that are generated from sources other than catalog prospecting, include those orders in your analysis. These are orders that should not be attributed, and this can obviously reduce the number of your unattributed orders.

There are always going to be orders that can’t be attributed in a matchback. The challenge is to minimize those orders so you can trust your matchback reporting and use it to build financial forecasts for all future campaigns. If you spend enough time digging into unallocated orders, you’ll improve your catalog reporting.

Jim Coogan is the founder and president of Catalog Marketing Economics, a consulting firm focused on catalog circulation planning.