An introduction to Six Sigma for direct marketing
Six Sigma methods, used correctly and thoroughly for continuous improvement of direct marketing sales, can produce remarkable results:

In this article, we will look at how some Six Sigma methods, when adapted correctly to fit direct mail requirements, achieve profit-growth goals quickly. We do this by adapting a few conventional Six Sigma working tools to the complex interaction of 13 simultaneous direct marketing variables.

Method
Our first step is to discover the sales produced by our present direct marketing variables (our control). We also use data to estimate what sales will be next month. In other words, our goal is to find a combination of variables that will produce gains against the control, and to capture variables data that will predict the following month’s results.

We compare our control to the first test direct marketing piece. If the first test wins, it becomes the new control. Then we make the second test piece beat the new control. We then design a third test piece to beat the second, and so forth, with a goal of consecutive growth every two mailings. This is a two-cycle compounding effect, whose data created the above three case successes.
Where traditional Six Sigma uses define, measure, analyze, improve and control (DMAIC), for direct marketing we prefer diagnose, prescribe, test, optimize and execute (DPTOE).

The diagnose and prescribe steps in Six Sigma for direct marketing have unique requirements, because of the language usage practices used in direct marketing. The test, optimize and execute features are similar to the AIC features in DMAIC.

Diagnostic variables consist of the standard statistical tests, with emphasis on three-variable designed experiments. First, evaluative market and direct mail data are studied to establish the diagnostic foundation of the advertiser. This is accomplished with a data analysis to produce five types of basic internal data that are labeled diagnostic variables.

Prescriptive variables focus on eight primary variables unique to the direct mail process. These will be described below. Data are collected, and startup hypotheses to be tested are identified.
Tests are developed that confirm certain diagnostic elements combined with tests of the startup hypotheses.

Optimize works with the information collected to define the economically most significant test data. After analysis, the confirming tests are structured. New tests will start the continuous improvement in the second phase.

Simultaneously, where extension of first data is indicated, the operating budget is structured to incorporate the new findings into the present system as quickly as practical. The purpose of this is to convert first tests into an operating improvement plan for the present program and to start the continuous improvement program indicated by the data for the next step in the ongoing promotion series.

Execute is the implementation phase of all of the above into an ongoing program.

Following are some examples of diagnostic measurements:

Diagnostic variables are important underlying factors that are separately investigated. They are not part of the direct mailing specifications. Here are some examples:
1. Late mail, defined as items mailed less than six weeks before the offer ends, and that are not mailed twice to the same name at three-week intervals. On average, late mail produces 30–50 percent less sales volume.
2. Relative dollar revenue of every mail segment in every list. What is needed is present dollar volume of each product in each segment with one year or more of data, as available.
3. Written marketer statement of the product problems in the more important segments. That is, what are the buyer’s wants and needs (within each important list segment) with respect to the product, presented in greatest possible detail? A long problem list is desirable.
4. Profit or attained margin measurement by

o every important list segment
o every offer that has been used
o list total and list segment, and as a whole for every mailing done in the past year
o Copy and format for every mailing in the past year, as used in envelopes, circulars, order forms, and each other enclosure sample, with test data if available.

5. Rate of growth by each marketing list segment for three years, if available.
6. Pricing and marketing policies for each marketing category.

Diagnostic categories are marketing foundation policies, for the most part. Data analyses of trend and market differences, when acted upon, are readily evaluated and lead almost immediately to profit improvement. In a sense, they are similar to the “Define” category in Six Sigma, except they often define sensitive marketing system issues not limited to a single problem, but instead help define and specify broadly significant policy options.

The prescriptive tools are immediately practical for development of the direct marketing materials themselves.
In DMAIC, the purpose of "measure" is to define the variables to be measured in later work stages. Our direct marketing alternative is P for prescriptive. Many copy (marketing lingo for "what the reader sees") elements cannot be measured directly. One method to test the effectiveness of ad copy is to test all the copy in a control piece as a single unit, and entirely different copy in the test piece. In this case a useful strategy is often first to compare marketing results from copy A to copy B , then in separate tests compare results of changing main headlines or other elements. similar tests are useful in format design, offers and product descriptions.

Many prescriptive variables can be rank ordered or evaluated quantitatively in other ways. For example, every mailing is likely to have letter copy, headline copy, format display, offer options, optimum repetitions and greater or lesser degrees of personalization. These variables can be defined and compared. Our strategy is to standardize tests and controls, and to use experiments to assess three variables per experiment at a time, thereby evaluating six variables in a mailing with two designs. Our long-term probability for discovery of two test wins in six tries in a single mailing is about 80 percent.

Prescriptive variables summary follows; the unique values differ in every single marketing effort.

1. Copy
A common procedure is to write the benefits vividly, document or prove the promise of the benefits, develop the offer to create a desire to buy, add the hook or incentive to act now, then write the headlines for key topics, prove your claims of benefits and create credibility using vivid picture words and a focused display. This is the total copy function. A common strategy is to use this method for broad comparisons with a prior control to find a proposed new piece.

When favorable, your next test can change only the copy or the offer, then test against the earlier successful test. This is compounding growth on copy alone. However, if you do the same thing with each list segment and simultaneously with offers tailored to each list segment, then you will discover the big winner in the test series.

For example: if the market consists of three primary list segments—each with its own buyer fundamentals, like age differences, sex differences, ability and reasons to buy—then three specialized kinds of copy are more effective (often by 200 percent or more than the same appeal averaged over the pooled lists), since it produces larger gains in each test segment than a single piece of copy focused only on the common product attributes. This simultaneous segmentation of copy, list and offer is called “personalization.”

2. Offers consist of three primary factors: The offer itself, the language used to state the offer, and the number of corollary offers that may accompany the primary offer. Examples: (1) Buy one, get a second one free; (2) Buy one, get the second one at half price; (3) Buy two for the price of one.

Note the differences: Numbers 1 and 3 are the same offer in different words. They are not likely to pull equally with all list segments. Learning which segments are better represents potentially large profit differences. Often, restating key words doubles the response.

3. Hooks for a right-now response: Corollary or secondary offers, “Order before Friday, and receive a free subscription to “XYZ magazine.” With the same basic offer, will the secondary offer increase the response enough to more than pay for the offer? Would a different secondary item out-pull the first one tried? We test to find out. When we know, we plug it in for long term use.

4. Complete product descriptions, each tailored to list segments to be used: Where list segments are interested in different product features, they are made more dominant, with more copy to fit that segment’s special interests. This is a form of personalization to the list. We test copy variants and suspected features.

5. List segment past productivity is critical: We compare static versus time-related list segment performance by sales period. “Static” refers to comparative periods, and “time-related” compares extended times such as by month, by season and by year for the past three years. Schedules are modified to compare similar periods.

6. Layout and format design tests: One example tests charts and graphs vs. words.

7. Optimum repetitions, to be discovered in each case: Basic repetitions (one time per segment) compared to two, three and four repetitions for a mailing, to observe the effects on total profits after mail costs for each kind of repetition.

8. Personalization: One outstanding performance caused five programs to increase over non-personalized tests by a factor of eight. For each order produced previously, personalization produced an eight-fold increase. Personalization refers to repetitive use of a mail list name within the same mail piece. With referrals to recently purchased items by the buyer, the mailer offers unique value for that person only, and for a limited time, with respect to selected products, with a high value premium for purchase by a given expiration date. First class postage may be tested against third class. Growth requirement for personalization promotion is no less than four to five times the nonpersonalized promotion to the same list, with at least double the profit margin.

How to beat your control—some working tips
—You’re trying to compare the 13 diagnostic and prescriptive factors in both control and test mailings.
—Start with mail samples, lists and results data. Look at each item in both lists. Compare control to test headline, benefits display, words used, how were key ideas displayed control vs. test. Write down differences. Make note of which look better or worse, in terms of the 13 variables. Write down ideas for change as they come. Organize, rewrite, polish, strengthen it. Review two or three times, then use it.

Cause and effect process charts monitor and strengthen the improvement process, record winners and facilitate new idea development to help growth become continuous
The Ishikawa fishbone, or cause-and-effect, chart is an effective monitoring and idea development device. Note the captions on each fishbone. Note copy, offer, list segment ideas, etc. Each cause produces it own ideas for improvement of each factor. Brainstorming leads to new ideas to be added.

We record and date each idea used, along with the results number, using a suitable metric for that particular fishbone. Every planning session starts with a review of what we have learned since the previous meeting. The meeting is then opened to brainstorming or idea development for new ideas. Appropriate ones are selected, added to the chart, tested, and evaluated at regular planning meetings when new data has been accumulated.

Documentation for source improvement is what Six Sigma brings to direct marketing. If your direct mail–generated leads or sales grow from 20 per 1,000 to 40 per 1,000, you will increase your direct mail profits this year by 222 percent.

See column below, Profit Margin/1000

Six Sigma compounds sales from direct marketing by a continuing inflow of verified new ideas from 13 critical variables.

Three of the 13 more important variables are copy, list segments and offer. The following direct mail experiment was used to generate qualified inquiries from several types of manufacturers, and several types of wholesalers serving five different industries.

Worksheet Data for this DOE

Each test was significantly greater than the control. The total response (all three variables combined) can be seen in the last row of the above table. The combined total produced 96 responses in controls and 237 in tests. This was promising since this was the first test. All three tests outperformed controls by 147 percent on average.

The New Problem: Which variables of our 13 can we now apply to the successful test we just transformed into our new control, so as to compound our growth further?

Some new ways of thinking about this test
1. The main envelope headline supported generic product benefits. A secondary headline on each envelope defined special buyer interests by industry. (e.g., machine tools had a different sub headline than steel makers, who differed from warehousing. Your hypothesis might be: Is personalization by type of industry the likely cause of the superiority? Can we test this idea by new expanded copy based on these differences? Here’s new work for the marketing team.
2. List segments were differentiated in a team meeting, which suggested that since the purchase was expensive, CEO’s might want involvement, therefore lists could separate CEO’s from production managers. This hypothesis is supported.
3. The variable, letter copy, expanded on the envelope promise. List segments were selected by field to agree with the envelope sub headlines. They used the same industry appeals as envelopes. Question: Can we add a new offer for the next mailing to both envelope headlines and with a major display in the letter? Will that increase response further?

Discussion and action
The designed experiments of Six Sigma tell us that some 13 direct marketing variables, combined and measured correctly, are the primary causes of response. Available hard buying evidence separates more effective from less effective methods precisely. For 100 years applied sales methods were driven by what vendors chose to sell, rather than on how marketers needed to sell. Today, marketers can achieve predictable continuous growth at lower total costs, more quickly and easily, when they use all the relevant, available, objective facts. You can almost choose your growth rate.

Summary of the key analytic process from the data—the next test step
To beat your best test results, combine the primary findings of the winning variables with relevant new ideas. This becomes the new test to compare to the best-yet control. This is an ongoing process, whose goal is to retain all the best factors of the winning variables. A practical strategy is to combine them with new ideas from the last three variables used in a prior test, then add the most appropriate of the 10 as yet unused variables. Essentially a recombination of the best-pulling elements added to some entirely new variables. Use three-way tests only.

How can we recover from past sins of omission?
The problem of continuously increasing returns is a scientific one, which can be solved by widespread use of applied science methods. It cannot be solved with “pretend” scientific sales talk. Numbers, measurement and statistical analysis are musts in the marketing mix, no matter which genius writes the copy. Word people are indeed crucial, but they are far more effective when directed by complex, interacting, verifiable facts. That means designed experiments developed to solve specific marketing problems. Two-way split comparisons are useful, but nowhere adequate to manage multiple interacting variables, which is precisely what Six Sigma for direct marketing does for business marketing. With it, response increases. You can predict from relevant data what the response will be, and over a mailing series you will see continuous improvement. For example, for an ASQ chapter, we mailed 1,600 members two mailings per quarter for 12 quarters, starting with 38 registrations per quarter, and by the 12th quarter we had reached 350 registrations per quarter. Note: Except for a limited turnover of members, we sold the same 1,600 people 10 times as many courses as they had ever bought before. This is continuous growth, equivalent to beating your current control 10 times in 12 consecutive quarters. Now you can finally say goodbye to the one or two percent returns most firms experience and see profits grow as never before.