Learn four warning signs that may suggest a data model build for direct mail acquisition isn’t worth your time and money.
When done right, data modeling can be a glorious thing. Good models can help identify top prospects, weed out underperforming names, predict pay up, and find look-alike audiences – all to improve your bottom line.
But building a data model isn’t always the best answer, and could result in a lot of wasted time and money with little return. When deciding whether a model makes sense, rely on your list broker for their sound advice and consider these four scenarios before moving forward.
In the spirit of Jeff Foxworthy, You may not want to model if…
1) The list you are looking to model has a universe of less than 1 million names. No matter the type of model you build (Good Customer Match, Mailed-Match, etc.) they historically perform better and have longer lives if the pool of names you start with is larger than 1 million, and the data is more robust. If this is not the case, then you should consider not modeling.
2) You have taken straight selects off the list you are now contemplating modeling and the results are poor. This is a red flag and should lead you to believe that the list in general may not be a good fit, and the model will not lift results enough to make it a quality performer. Results on the straight selects should be encouraging enough that the building of a model will make the list a top line performer. If this is not the case, then you should consider not modeling.
3) There will be an upfront cost to build a model or an unreasonable minimum usage required (i.e. build cost is $1,500 or the minimum usage for year 1 is greater than 50,000). The cost of building the model should be absorbed by list owner, and the minimum should be no more than 50,000, if the model is working. If this is not the case, then you should consider not modeling.
4) The overall price points of the model don’t work. If the base price and the scoring charge are so high that the overall CPM will offset the gain from the model, you should think twice about modeling. The base price on most models should be in the $55-$85/M range and you should never pay more than $25/M in scoring fees. Net name arrangements should also be negotiable on rollout quantities. If this is not the case, then you should consider not modeling.
Additional consideration: You have gone ahead and built a model. The price points are in line and you take your test quantity and mail it. Eight weeks later you read results and it turns out that the model did not provide the lift you were hoping for, and your overall CPM is high enough so the model would not make the next mail plan. Speak to the list owner and the modeler, and share your results. Perhaps there is some tweaking that can be done to the model to aid in performance, or the model may warrant being rebuilt.
Related posts and whitepapers:
How to Mail More Efficiently Using Merge Optimization & Balance Models
What to Know Before You Model: The 4 Most Commonly Used Models for Acquisition
Are you missing the data modeling boat?
Previous Post
Free Resources Every Nonprofit Should Know About