Are the Cooperative Databases Good Business Partners for Catalogers?
The buzz at the NEMOA directXchange Spring Conference earlier this month was a scathing critique of the cooperative databases. A lot of people were shaking their heads and asking, "Were the co-op database criticisms simply factually wrong?"
Why have the co-op databases become the dominant source of prospecting names? Because transactional data is very powerful. Not only does transactional data work to provide universes of profitable prospecting names, the co-ops' data have the potential to provide the infrastructure for names for the world of e-commerce. Note the recent acquisition of Datalogix by Oracle, a purchase which was fueled by the strategic value of the data and Datalogix's ability to tell both traditional catalogers and online retailers which households are buying with both precision and deep universes. Here are some of the most common complaints you hear about the co-ops:
Lack of Transparency
The co-op databases are a black box. Catalogers don't have visibility to what's making up their models and how can model response rates be improved.
The simple key to any relationship is communication. Most catalogers’ experiences with the co-ops are that communication about results and specific challenges are the starting point. The co-ops welcome a dialogue, as they're in the business of providing names that prove profitable and will be used repeatedly. The co-ops have a vested interest in improving model performance. Best practice is to talk with the co-ops and share your results, making sure the co-ops understand your business and its needs.
The co-op databases want relevant client input. They can and do give directional guidance on who is in your model and what the top variables look like. Catalogers can improve their own results by supplying regular housefile updates and full mail results, as well as communicating why consumers buy from them — i.e., things that are important and can't be seen through just transactional data.
A Mediocre Pool of Names
What's correct in this assessment is that the co-op databases feel it's in the mutual best interest of catalogers and the co-ops to mail all names that respond above breakeven. Unless told otherwise, a co-op's goal is to provide the maximum number of names that meet acceptable criteria.
When a catalog needs to improve response rates or operate with a conservative prospecting strategy, communication about business needs is the key to getting better performance. Is this simply a frustration that with a 1 percent response rate, 99 percent don't respond? And if you could only get rid of those 99 percent who don't respond, then profits would improve. I think this is a frustration with the daunting economics of a prospecting business model built on a 1 percent response rate rather than the co-ops either withholding good names or wanting to provide "bad" names.
The key to getting better model performance is to use analytics rather than intuition to challenge model performance. If your intuition tells you that the catalog should have a demographic overlay, test the premise to see if your intuition is backed up by actual results in the mail. Analytics beats intuition because you can test and prove your intuition.
Same Segment Sizes ... And They're Too Large
The co-ops are actually pretty responsive if a cataloger prefers to take names in smaller segment sizes. Taking smaller segment sizes is particularly useful in trying to pinpoint where a model performance falls below breakeven.
It's not very useful to break a prospecting universe of 500,000 names into segment sizes of 5,000 each because you create a problem trying to read results from 100 different segments.
Fewer Names Flowing Into the Co-Ops
Actually, the number of households and transactions flowing into the co-op databases is at an all-time high. The co-ops are also signing up pure-play retailers and new clients that are contributing data.
Lack Flexibility and Push Catalogers to ‘Let the Models Do the Work’
The basic proven models are a result of modeling a company's buyers and using that data to develop models of most likely buyers. Then the models are tested in the mail. It's smart for catalogers to question the proven models and look for ways to improve them. While the co-ops’ methodology to develop and roll out models is proven and established, it's always useful to push the co-ops on the need to expand with new models, deeper universes and improved performance.
Demographic Overlays Are a Valuable Pre-Select That's Overlooked by Co-Ops
In some cases demographic overlays can prove useful in either finding pockets of response or to suppress demographics that won't work. The key is to test and see if the hypothesis about a demographic overlay improves response. Often the model incorporates the variables that you intuitively feel are important. I've personally tried over the years to make demographic overlays work to improve response rates with limited success. Demographic overlays can be tested as post-modeling tags. Age, for example, can be used for lower segment improvements.
Co-Ops Encourage Catalogers to Mail the Same Households Too Frequently in a Season
This is easily testable and should be tested. Simply segment "new" vs. "previous" names from the same model and measure whether follow-up mailings in the same season prove that "new" names" outrespond "previously mailed names." My experience is that mailing the best prospect names multiple times in a season is a proven best practice — better than mailing good prospecting names only once in a season.
Managing a database of all the prospecting names mailed and the lifetime frequency of contacts can only be managed at the merge/purge service bureau level and not by the co-op databases because an individual co-op is typically not the sole provider of names.
Co-ops Are Actively Selling Your Transactional Data
The transactional data is being sold, with the exception of Wiland, which has a policy to not sell its data. Does it help or harm catalogers that their data is being sold? The long-term issue is what happens if online retailers learn how to convert over a significant portion of a co-op's buyer database, but never have to reciprocate and flow their own buyer files into the database? The co-op's response is that they're actively signing up pure-play retailers and insisting pure-plays contribute their own data.
Co-Ops Are Losing Out to Amazon and its Bigger, Richer Pool of Data
Catalogers are finding that if they're Amazon.com affiliates, they need to be very careful before they drop Amazon affiliate buyer data into their buyer files. Amazon affiliate buyers purchase completely differently — always worse and usually much worse — than a company's own buyers. Why? Because the Amazon affiliate buyer bought from Amazon; you were really just serving as the shipper of the product. So an Amazon affiliate buyer is loyal to both Amazon and buying at the lowest price, and has little loyalty or even connection to you the affiliate, and therefore will respond much differently than your own buyers in the same RFM segment.
Response Rates Are Flat
Well, is the cup half full or half empty? Another way of describing flat would be to say that response rates overall are stable. Stable response rates sure beat the wild ride response rates took during the 2008-2009 recession. Should both catalogers and the cooperative databases work together to improve response rates? Yes, of course.
The grab bag of tactics ranging from smaller segment size to lower pricing for fringe names to demographic overlays to only mailing a prospect a single time in a season to capping the number of times a prospect could be mailed don't have the potential to individually or collectively "move the needle" for catalogs.
At NEMOA, I kept waiting for a solution to the criticism of the cooperative databases. But alas, no real set of solutions was put forward. I would say that, again, communication between catalogers and co-op databases is the key along with a commitment to relentless testing.