Demographic data is tremendously useful for marketing.  Knowing the demographic profiles of your customers can improve your ability to market to them and allow you to prospect for new customers with similar profiles.  But demographics only give you a very general view of your customer.  Knowing somebody is of a certain age, income, or gender provides some insight into their behavior because on average they are more likely to behave in a certain way.  But if we really want to know what they are likely to do we can gain much more insight by using behavioral data – purchase history, help desk calls, marketing campaign responses, and any other activity that a customer may have with your company.

We were working with a client recently that had been using our consumer database for prospecting.  They had some great success with the prospecting model that we built and started looking at other ways that they could use our data to solve some of their business problems.

They asked us to build an attrition model so that they could proactively engage customers that were most likely to churn.  This is where we found the limits of the data we were using.  We were only able to get a small amount of lift when trying to predict attrition using the appended demographic data.  From working on previous projects in predicting churn we knew that behavioral data would be much more predictive.  Unfortunately this client’s marketing department did not have the data available from their operational systems.  We are now working with them to gather the internal data that they need in order for their retention project to be successful.

Of course there is always a flip side to anything, including knowing more about your customer.  Target received some unwanted publicity earlier this year for being a bit too good at predicting their customers’ behavior:

http://www.colbertnation.com/the-colbert-report-videos/408981/february-22-2012/the-word—surrender-to-a-buyer-power

What is the state of your data?  Does it bring value?

The buzz in the industry is “Big Data”.  What do you do with it?  How do you manage it?  What is its value?  Whether you plan to use the data in a batch environment or real-time or both, you need to decide that ahead of time to make the most of your data management process.  To really understand where the value is in data and where you stand in deriving the value you can use the grid below to assess you situation and what you should do next.

8 States of Data

The cost of getting to the states of ‘Distributed’, ‘Analyzed’, and ‘Prediction’ are relative to the scope, complexity and control of the data with control being the key piece in the cost equation.  If you outsource or use existing structures your cost of control goes down while if you want to own the process and data the cost goes up with the hardware, software and resources necessary to pull it off.  Generally all but the largest rely on external support.

There is a thirst for knowledge spanning across industries.  Data is exploding with social, web, transactional and other data sets becoming more readily available.  To acquire the knowledge demanded by upper management, companies must have access to analytics.  But what kind of analytics?

A quick Google search provides 112,000,000 hits on the term “analytics”.  Of course, number 1 is Google analytics, but you also have Strategy Analytics, Business Analytics, Predictive Analytics, and on and on.  It seems every agency, printer, software vendor, and data processor claims to have “analytics”.  Recently, our President of Analytics was visiting an agency and was told by the Managing Director that they provided their clients with analytics.  Explaining further, he said they provided campaign analysis and reports to their clients that helped analyze the success of campaigns.  Our President explained that he was leading several marketing database projects which included aggregating and linking data from many different databases, adding external data and building predictive models and segmentation strategies to affect campaign results.  The Managing Director responded, “Oh, we don’t have your kind of analytics”.

Why do I point this out?  I think there is a lot of confusion out there, probably led by Google’s unfortunate product title Google Analytics.  I use Google “Analytics”.  A better term would be Google Reporting.  Providing tables, charts and graphs on website activity and Adwords is not analytics in my world.  When I review my web results with Google Analytics, I’m left with two feelings.  Either, “Oh, that’s nice—my website traffic is up” or “What the??? Why is that doing this and how do I change it”.  I then download all of my Google Analytics information and send it to our Analytics department.  They come back with predictions and strategy—basically answers to my questions.

If you really want to make better decisions, pull all of your data together, add external data, and have analytics applied to give you a clearer view of your situation and to help you develop the right strategy moving forward.

A few years ago, Financial Institutions were relying on ‘free checking’ programs to bring them new customers, which in turn usually provided more overdraft associated fees.  Well, those days are mostly over with the new laws and regulations in place. In those days some of the most effective types of mailing were all about large quantities.  Most people like the word ‘Free’ and since everyone’s accounts were alike, the banks were hoping that by mailing the masses, someone would be looking for a new bank or upset with their bank on the day they received the mailer.

Today we are seeing the ‘free checking’ accounts going away.  ‘Free checking’ was short lived as a bank product. People tend to forget that ‘free checking’ has only been around 10-15 years.  Before that almost all the banks had a minimum or a monthly fee.  Now banks need to be more selective than ever.  They need to market their best products to who they consider their best prospects.  What they need to ask are who are my best customers and how do I get more?

If you know which customers are best for you, why not clone them?  By finding the similarities within this customer group, a clone profile could be created.  You’d then be marketing your favorable products to your most choice prospects via channels they interact with.  Distance still plays a role, focus in on the best ‘fitting’ prospects for your location.  Now the old practices still hold true for timing, it is everything.  Even when you’ve chosen the best prospect and offered them the right product changing bank relationships is tough so you may not get the results you’d expect.  Always start with the low hanging fruit, the newly moved.   Don’t forget everyone else but new movers are your first priority.  Your prospect mailing to round out the acquisition still happens but now we have lowered our volume to the best prospects, with the right offer and with the right campaign structure, the right time.

The Wall Street Journal’s September 30^th headline screamed “Fixed Mortgage Rates Hit Record Low”. The WSJ article (http://online.wsj.com/article/) went on to say:

“Fixed mortgage rates sank to record lows over the past week following the Federal Reserve’s decision to buy longer-term Treasuries, according to Freddie Mac’s weekly survey. The 30-year fixed-rate mortgage averaged 4.01% for the week ended Thursday, down from 4.09% the previous week and 4.32% last year. Rates on 15-year fixed-rate mortgages averaged 3.28%, down from 3.29% last week and 3.75% a year earlier.”

With rates this low, we are recommending to mortgage marketers our newly updated Prime Refinance Model. This is an in the market model that accurately predicts those homeowners who have an above market interest rate and are likely to respond to a rate/term refinance offer. The Prime Refinance Model is a non-FCRA product, so actual credit data is not used. However, there are credit and behavioral indicators in the model that identify customers that will have higher credit scores. Several large mortgage mailers have already tested the newest model release and are having outstanding results.

What ties these three things together? Economics! I was recently at a conference where John Silva the Chief Economist for Wells Fargo spoke. His main goal was to show you that even the slightest discussion of economics when doing planning for your company will go a long way.

Let’s link them up. Sovereign debt is the debt by the government of a sovereign state. The debt crisis is the fear that the debt will not be paid back, which in turn causes the interest rates on that nation’s debt to go up.

Now, mortgage interest rates here in the US have dipped down again. Why would this happen? As the Wall Street Journal tells us, all the investors with money to park somewhere are avoiding the instability in the Eurozone and are coming to the U.S.  The availability of funds is lowering our rates.

Finally, lower rates are creating a surge in applications yet again. This surge is mainly in refinance but purchase lending has gone up as well. In order to effectively lend you need to be targeting who you will lend to when you communicate with your customers and your prospects. These quick changes don’t give you time to build up a custom model for your targeting. You will need to use models that are already in place to ride this wave. You don’t know how long it will last and you can’t take the time to build custom.

What Altair has done successfully for many lenders is use a shelf model while gathering the data to either tweak the shelf model to perform optimally for your product or build custom. You won’t lose out on the burst in potential lending while you prepare for a sustained marketing effort.

The number of mortgage refinances has risen steadily for the past 6 months averaging over 400,000 monthly refinances in August, September and October. With interest rates at historic lows and home values at depressed levels, many homeowners are deciding to stay in their current homes longer and to lock in a low, long-term rate.

From a low in November 2008 of 100,000 refinances, there has been a sustained level during the past 2 years of at least 300,000 refinances per month. Four times refinances have exceeded 500,000 loans in April, June and July 2009 and again in April 2010.

As reflected in the chart below, the bulk of the activity is in loans under $500,000 with most loans falling in the $101,000 – $200,000 range. This has trended downward significantly, probably as a result of falling home values.

For banks and mortgage companies offering prime loans, there is still a significant opportunity to market prime refinance loans in today’s market. If you need help reaching the homeowners who are likely in the market to refinance in the next 30 to 90 days, you may want to test the Altair Customer Intelligence Prime Refinance model.  The model gives you access to over 18 million prospects who are 2.13 times more likely to refinance than without the model.

Beyond the latest buzz of social media and the explosion of social media listening tools, your customers are always telling you something.  While interesting and potentially informative, social media listening tools aren’t perfected to the point they can deliver real intelligence at the customer level.  At best they tell you how well a product is received by the open market which is helpful for R&D and mass media but what about direct marketing?

If you listen carefully you can influence your customer’s lifecycle.  What is listening for direct marketers?

• Gathering primary information from customer surveys.
• Using transactional data from your interactions.
• Collecting transactional information from outside your company.
• Appending data that informs on needs, attitudes and behaviors.

What do you do with what you’ve heard?  You analyze the data, segment your customers and predict behavior. 

Analysis of the data includes reporting on coverage and relevancy, indexing against peer populations and preparation for modeling. 

Segmentation in the instance of loyalty would break your customers into groups much like this;

• Loyal customer you wish to retain: they could be high transactors, your largest purchasers, own the most products and/or are the most profitable.  You have tremendous amounts of data on this customer; internal and external.
• Promotable customers you want to become more loyal:  they exhibit behaviors of your loyal customers but only sporadically, they use products from other brands that you provide, they closely resemble loyal customers in external analysis.  Your internal data assets coupled with external data allow you to classify them with potential.
• Customers you’ll maintain: they don’t exhibit behaviors of becoming or being loyal but at the same time they aren’t a burden to your ecosystem.  You have very few bits of information from internal gathering but external data tells a good story.
• Customers you’ll let attrite: they come for the loss leader promotions, they use extensive resources with little in return, and the bulk of their transactions are elsewhere.  You have almost no internal data and external data points to loyalty placed elsewhere.

The prediction phase is where you develop the models that will help you communicate to you customers using a method that will motivate them to action.  It answers questions such as what are the offers and incentives that will retain my loyal customers, promote my on the fence customer and maintain my profitable customers.

Your customers are always telling you something, even silence speaks volumes.

There are many types of predictive models used in direct marketing: response models, clone models, cross sell models, revenue models, and many others.  What these models all have in common is that they require a great deal of data in order to build a model that works well.  But it doesn’t matter how much data you have if that data is not good quality data. 

At Altair we build all of our models using our exclusive multi-source consumer data that is built from over 20 sources.  First, we only use records that have been delivery point verified (DPV) insuring the address is the highest quality.  The model would be worthless unless you can contact the consumer.  The data is validated, meaning that we look at what has been provided and make sure it is in the acceptable range of values for that data element.  The data is also verified.  We do this by comparing several incoming sources to each other and if both elements are the same or within acceptable tolerances we know what is being provided can be trusted.  The result is a data file for modeling that has highly deliverable consumers with the broadest array and deepest coverage of data.  The result is we are able to fill in many of the missing values that are the bane of any modeling project.  And the values that are populated on Altair’s files are going to be the cleanest and freshest values available. 

All of this clean and accurate data helps to create models that are more predictive, more stable and give you better results.  Altair’s proprietary multi-source data is one of the reasons that our shelf models and custom models work so well for so many of our customers.  This data is also available for your modeling projects so that you can build the best possible marketing models.

Do you measure the success of your direct marketing with just the response rate? If so, you may not be getting a good measure of success. Especially for financial products, responders often fail to convert into buyers. This can be problematic if you are selecting prospects to mail by using only a response model. A strong likelihood to respond often means financial issues that mean a low likelihood of being approved as low credit-quality consumers seek additional credit.

read more