Enhanced for Response ... and Profits
If you're experiencing attrition and dropping response rates, or if you have aggressive profitability targets (or both), then you're probably looking for ways to make your housefile work better. The question is: How can you do it? You know your customers' purchasing patterns inside and out and have eked out all the revenue you can with RFM and/or modeling based on house data. What more can you do? Perhaps there are file enhancements that can generate an incremental lift.
Traditional file enhancements fall into several categories, ranked here from least to most expensive:
-Hygiene enhancementsIf you're wasting marketing dollars mailing to people who no longer exist at the addresses you have for them, then you can get some of that money back through better hygiene and move-tracking.
-Geo-demographic overlayUses geo-coding to allow you to append area-level summary data to your housefile. Common options are carrier-route, ZIP5/ZIP9, or census tract/block group data from various vendors.
-Individual-level overlayAdds information on marital status, family composition, age, income and more at the individual level. Also adds household-level information, such as home value, owner/renter flags and more.
-Behavioral dataGenerally acquired through cooperative databases, such as magazine subscriber or catalog purchaser databases.
-Firmographic dataNot to be confused with demographic data (demographic literally means "people-picture"), this is information on companies. If you have a B-to-B list, then SIC, company size and a host of other attributes will be the ones you're looking for.
-Attitudinal dataGained from either your own survey research or from syndicated research sources. Because of the cost of surveying individuals, these data elements tend to cost the most to acquire unless the information is gleaned automatically through the purchasing process.
What returns can you get from hygiene enhancements, you might be wondering? If you do not currently NCOA your file, or regularly deduplicate or otherwise maintain your file, then you could be losing a great deal of money to communications that are being thrown away. An easy way to test hygiene is a First Class postage test for your next mailing. Simply select a random sample of your target audience and mail their packages First Class. The return rate of undeliverable-as-addressed or address-corrected pieces times the number of communications you sent to these customers in the last year will tell you how much you're losing. Alternatively, you can look up a sample of your housefile on the Internet, i.e., do they still live where the record says they do? Is their phone number correct? Have you spelled their names correctly? Call them up and ask. Projected losses from return rates will help to give you the ammunition you will need to justify data cleansing and hygiene.
Determine the ROI of Enhancement
File enhancements cost money. Depending on the size of your housefile and your budget, they could potentially cost more money than they return to you in performance. This is why you should take an ROI-based approach to file enhancement. In that way, you can get those enhancements and have a good chance of making your profitability targets as well. Four simple steps will help you determine whether a new enhancement source is worth the investment:
Step 1. Determine the baseline performance that you have to beat. Imagine your housefile contains 500,000 unique individuals or businesses, and you currently are using purchase behavior and information on the housefile to create your segmentations and/or models for marketing purposes. You need to collect information about all of the outbound programs you've done in the last 12 monthswhat you spent, what your response rates were, and the resulting sales and profit associated with those programs. This is the baseline that your newly enhanced file has to outperform.
Step 2. Extract data to send out for overlay. Extract the promotions and match response to them, then extract the name/address information of the people on the housefile who received those promotions. Deduplicate the list so that you get down to everyone who received at least one promotion last year, and then create a random sample of approximately 100,000 records. This is the file that will be used for the test append.
Step 3. Identify data sources that you wish to test, and negotiate an
append for analytic purposes only. This means that you do not expect to permanently buy the data, only that you wish to test it. TIP: Some of the data compilers will want you to buy their data so much they will append the data to your sample file for free.
Step 4. Analyze the lift that the new source generates by "backtesting," or including the new data in your old campaign selections. When the data append file is returned, you can analyze the impact of demographic, geographic or other data on the quality of your segmentation or modeling efforts. If you build models, then look at the lift you get from adding fields from each source to the models you used over the last year. Any incremental lift you get from adding those data elements must pay for the full data appendand make you a profitother-wise the data isn't worth what it costs. Make a note of which fields on the overlay files actually "hit" in your models or give you lift, because if you're paying for 150 attributes and use 10 of them, then you're going to be analyzing and paying to store a lot more data than you need. Review the list of attributes to see which of them make sense as predictors of attrition, risk or response. Look at each field to see if at least 2 percent of the rows on your file are populated with dataif not, then it will not be a good attribute for modeling and analysis.
Here is a sample scenario. A test append of individual data and a model rebuild proved that adding age and income from a compiled list source to a Christmas campaign raised purchase rates from 1.6 percent to 2 percent. When those figures were plugged into the campaign, the mailer would have gotten an incremental 400 purchasers at $75 each on 100,000 pieces mailed, or $30,000 incremental dollars from the Christmas effort. If these sources were included in Easter, Valentine's Day, and other holiday models, the company would have ended up generating $400,000 in incremental revenue and $200,000 in incremental profit. If the data append costs $75,000 for the entire year ($150/M for the 500,000 record file), then the company would net $125,000, or an extra $1.67 in profit for every dollar invested in the data sources, which is a good return on the data investment.
What if no data source you test generates profits? For the tested source that comes closest to profitability, use your ROI numbers to negotiate with the data owners. If you need to make a certain amount of profit from your investment, then perhaps they will work with you. Also, it may be that your list broker is enhancing your file so that it can broker the file for more dollars per thousand. Ask if you can get the overlay back from it for your own use.
The larger your housefile, the higher the ROI hurdle you face because
enhancement data can be very expensive. It is critical to determine what you can afford and what your profit-ability targets are for each enhancement effort, since it may end up being cheaper to license an entire 100 million-record data source than it would be to append the data to your 50 million customer records. For companies with small housefiles in limited market areas, you may find that geographic enhancements give you almost no lift because there aren't enough unique values at your disposal. If you have no budget for enhancements, perhaps you can get creative. Look to some government sources to find an incremental lift. For example, the Bureau of Labor Statistics (www. bls.gov) and the Census Bureau (www.census.gov) compile data on consumer prices, unemployment rates, spending and a host of other information at the geographic level. It may be that knowing that particular counties in the United States are richer than others or that unemployment is rising in certain cities, will help you either target or avoid areas that cannot afford your product or service. Also, your housefile itself provides some great information; summarize it to the ZIP9 level and create statistics like percent of best customers in the ZIP9 area. This type of information can help you not only with housefile modeling, but can be used in your prospecting efforts as well.