Session 2 - How to assess the value in external datasets and answer the question Which data should I buy to enhance my marketing? In the first edition on applying data analysis in the B2B environment we explored the area of profiling and market penetration, and the use of indexes and z-scores to describe and compare data populations. In the next instalment we will look at what data can be used to enhance your understanding of your customers and market, and how to evaluate the different data elements that are available. However being able to decide on what data you require harks back to the understanding you have now potentially reached relating to your market and the customers within it.

B2B Information

nevitably marketing into a B2B environment requires definition & understanding of the decision making unit (DMU), and associated business requirements. Essentially like any sales an understanding of what are the key triggers or environment that will cause your customer or prospect to buy. Do you know what these pieces of information are? This type of information might be related to product or equipment type, feature or age, or a service consumed, with associated information such as contract date, replacement cycle etc. Real trigger data is extremely powerful in that it will allow your sales effort to be directed at a particular time and in a particular way. Quite often this type of data is built up on the customers you have, and sourcing this data for prospects is trickier. We would all love to know when a prospects contract is up or will require more of our product. So you may consider a collection programme of data, which can be instigated through telequalifying a prospect list. This might be viable depending on the volumes in your market and the readiness of your prospects to release this information. However youre prospective audience maybe too large or too reticent to reveal this information, so lets take a step back and look at some alternative ways to prospect based on evaluating which data to append to your database to support prospecting activity. There are a wide range of variables that are externally available for instance through Marketscans business database, Megabase, and these could be added to your existing or prospect records. But obviously you want to ensure that the data youre adding is going to be useful and give you a return over not adding it.

So how do you identify which variables to add?

Firstly you should get a representative test file of data appended with a wide range of available data. (This is usually a good exercise to also audit the current state of your data, and see where you may have gaps, duplicates or issues). Then secondly you need to analyse the variables and design some business situations in which you might use the data for instance in database selections, segmentation of your base to send different offers etc. Really think through what you use your data for selections, business information, segmentation, territory planning. Obviously a new piece of data should add to the refinement of these activities and help you gain more accuracy or efficiencies on your marketing. In the first edition, we saw how profiling revealed the type of customer, using common variables such as SIC code, or no of employees (as an indicator of size of business). This type of analysis is descriptive in that it looks at one variable against another or against a base of data. However we know that life is more complicated than that it is combination of data that will often work best in describing or identifying your best customers or prospects. So you should look at doing further frequency analysis of your customer transactional data in conjunction with externally available data. So examine the value of past orders, timing or orders, type or product etc. together with some of the external data available (SIC, premise type, employees nos etc.). So for instance analysis of a business gift/stationery suppliers customers, might reveal a difference in seasonality of buying for different products by different businesses, as defined by a grouped SIC code within turnover or no of employee bands. In other words we might know that larger business are better for buying certain products within certain SIC codes and smaller businesses better for buying different products. Understanding this in itself may reveal key pieces of data. Beyond this, there are various statistical techniques, which help reveal where combinations of variables are useful, and I will cover one or two of these below. Cross-tabulation is relatively simple, and will allow you to examine one variable against another e.g. job title of purchaser vs turnover. However this may or may not be relevant! So you may want to understand if there is a genuine association between these two-variables before deciding they are worth going out and buying. A test of association between 2 variable is the chi-squared test and guess what - you can do this in excel! You can find our more by clicking on this link. Using a chi-squared test on a whole range of combinations of variables may at the very least be time-consuming. There is a solution! The chi-squared test is the concept behind a technique called CHAID (Chi-Squared Automatic Indicator Detector). CHAID is a useful technique in that it will help reveal which combinations of variables is more useful and would show you the uplift. The CHAID algorithm is found within software packages such as SPSS, and will work on categorical or banded data. The key here would be to look at a recent activity that you sent to a selection from your database and understand if the sales response has some particular cluster in terms of the underlying variables. See CHAID example. For example, of the organisations that responded to our activity how did the underlying information selection information differ from the mailed base. In our CHAID tree we could find that CHAID clusters together information such as certain SICs, premise types (schools & hospitals) and turnover in one node or group as being highly responsive (n times more than our overall response rate), while a different group may be less responsive (n times less than our overall response rate). By looking at the analysis with and without the new variable you will be able to see how much uplift this is adding. One final thing before we move off the topic of identifying variables - beware of correlation! Assuming that you have identified 4 or 5 interesting variables you might want to buy, stop and consider in fact if they are similar or highly correlated. Are they explaining the same uplift and therefore do you need to buy all of them? Consider indicators of business size for instance both turnover and no of employees may indicate a big medium or small business and you may only need one or the other. However at another level they may give you different sub-messaging, and there will be of course some small businesses in terms of employees with high turnovers. However correlation is another topic, and another way of assessing datas importance.

So is the external data good business value?

The next stage once youve identified your information is to examine the business cost vs return. You can do this by applying some of your own business metrics (product margin, response rates, marketing costs) and linking these to an assessment on what the data will add. For instance, in the case of a mailing activity, would a new variable help you target better and get in more sales. Quite often you end up using data you have bought in a different way to that you first imagined, but it is clearly worth regularly evaluating the data you are buying and licensing to ensure that it still continues to add to your business. Summary - How to assess the value of the data you are adding

A quick summary of what to do:

• Assess how you currently use your data
• Audit your current dataset
• Identify gaps & opportunities
• Append a range of externally available data on a test basis
• Assess how you will use the data
• Do some analysis to: Look at the data which adds to you Look at the financial cost vs return of different variables
• Buy, use & evaluate.

If you would like independent help on assessing which data to add, Talking Numbers www.talkingnumbers.com offers this service and can help you evaluate the range of available data. Please contact Julian Foxon, David Dipple or Nigel Magson.