Marketing Analytics - Multi-Channel Retailing

Editor's Notes

• #2: Hello! Welcome to Group 4’s presentation of our final project. Group 4 members are: Chris Robinson, Mary Sauer, Adam Schackmuth and James Young. We decided to complete our final paper and project using Dataset 9. We chose this dataset because we felt that it had a lot of rich data that we could derive our model statements and hypotheses from, and we felt it would challenge us throughout the semester. Dataset 9 is from a multichannel gift company with sales of several hundred million dollars per year. The company is a well-known organization that has a network of retail stores, a well-established catalog channel and a website as well.
• #3: Section 1 of our project discusses the overview of the business case and an overview of the marketing data.
• #4: Again, the dataset is from a Multichannel Retailing organization. Throughout this presentation, Group4 Consulting will analyze and model Dataset 9 to determine which channel has the most potential to maximize growth. Our audience can use this analysis to help target marketing efforts and potentially identify customer segments for: Attracting new customers Retaining the best customers Avoiding unprofitable customers
• #5: The business case are questions that we wanted to determine answers for. Our business case was to determine what channel strategic development be focused on to maximize growth; if multiple channels, what the order of prioritization should be. This prioritization would be related to whether or not there is a correlation between a channel and a customer segment and to identify any synergies between the various sales channels. . Some channels may correlate strongly with certain demographics or may push revenue into the other channels.. If this is the case, it should be determined if those demographics and synergies are important to the growth of the company, or if the company would be better off withdrawing from the channel and investing those marketing dollars elsewhere. Product: The Multichannel Gift Company sells food products which are usually purchased as gifts during the Christmas season. While the products are available year-round, the sales peak during the last few months of the year.
• #6: Customers: The customers of the Multichannel Gift Company are loyal to the brand and are very familiar with it. The brand is well-known and respected. Customers usually purchase goods from the Multichannel Gift Company around Christmas time and give the items as gifts. The customer can shop for the products in the retail locations, through catalogues and online. The customers have a variety of personal interests, ranging from travel to fine arts to fashion and camping. External Market: The mail-order catalog market is a mature market that is projected to decline 7.8% annually from 2014 to 2019 due primarily to steadily increasing competition from e-commerce. The industry shows that internet sales offers “greater convenience” than mail order or traditional brick-and-mortar sales, which combined with projected sluggish consumer spending growth over the next five years, does not bode well for traditional mail-order or local retail stores. E-commerce sites also experience lower operating costs, thus making it difficult for brick-and-mortar stores to compete on price. Globally, e-commerce sales now exceed $1.2 trillion, thus making a multi-channel approach necessary in today’s economy .
• #7: Hello – I am Mary Sauer, and I will be discussing Section 2 of our paper, which is the description of the data. We completed our project using Dataset 9.
• #8: Dataset 9 contains four separate files: DMEFExtractSummaryV01, DMEFExtractContactsV01, DMEFExtractLinesV01, and DMEFExtractOrdersV01. I will go into a little bit of detail regarding each file – what is contained in each file, what information we can derive, and what insights we can gain.
• #9: In total, the DMEFExtractSummaryV01 has 101,051 customer records, or cases, that show the customer buying activity along with demographic, psychographic and distance to retail store information. The buying activity is summarized by channel (internet, catalog or retail,) for eight seasons (Spring and Fall 2004 through 2007.) This Customer Summary file contains all of the information used in the regressions and data analysis, as it was the most comprehensive file in the dataset. This file shows us helpful demographic data that could be used to target our customers. After cleaning the data of no responses to AgeCode, IncCode, HomeCode, Dwelling, LenghtRes and OccupCd, we were left with 10,929 cases. This data give us insightful information regarding who our customer is. The statistics and frequencies can be found below, but the data tells us the profile of the customer. The customers are between 45 and 54 years old, the majority makes an annual income of greater than $50,000, with the most making greater than $105,000, and are home owners of a single-family home and have lived there more than 20 years. The customer’s occupation is usually professional or technical, while many are also in administrative management positions. This information can be used to segment the customers into groups and it gives the company the ability to better target market to the customers.
• #10: This slide shows what I just discussed, this is the SPSS output showing the mean, median, mode & standard deviation for each of the demographic variables.
• #11: In looking at sales dollars per purchasing channel, retail, catalog or internet, we are able to see the spending patterns across each season. For retail purchases, the minimum purchase is $1, and the maximum purchase is $2,318. The internet sales dollars were broken out into gift or non-gift purchases for each season of each year. After summing the gift and non-gift variables, we can see the minimum and maximum purchases for each season; the minimum purchase being around $18 and the maximum being $2,518. Similar to the internet sales, the catalog sales were broken out into gift and non-gift purchases. After summing the gift and non-gift purchases for each season, we can see that the minimum purchase is $19 while the maximum is $2,106.
• #12: The file DMEFExtractContactsV01 contains 3,389,239 marketing record contacts, and contains information regarding customer’s contact dates and contact types; catalog or email. This data shows information for each month for the years 2005 through 2007. Because the SPSS output was several pages long, a summary of the findings are included in the graph below. The data shows the months that the contacts peak. This file gives us insightful information regarding the time of year the organization makes the most contacts. The graph clearly shows that November and December are the busiest months. This file also shows us that the majority of the contacts, 70%, are made via email, rather than catalog.
• #13: The file DMEFExtractLinesV01 contains 618,661 line item records. The data shows the order date and dollar amount, as well as if the item was purchased as a gift. It shows data for each month from 2001 through 2007. After cleaning the data, it shows that the majority of the items are purchased for gifts, approximately 90% of the time.
• #14: The file DMEFExtractOrdersV01 contains 241,366 order or store trip records and shows the order date, channel for purchasing (store, internet, phone or mail) and payment method (bank card, cash, check, gift certificate or house account). The file contains information for each month from 2001 through 2007. This data shows us that the preferred purchasing channel is in-store, with phone and internet coming in second and third, respectively. The least preferred purchase channel is the mail. The majority of the customers pay for the items using a bank card, nearly 78%, whereas cash is the second most used payment method at 17.1%. All statistical and frequency information can be found in our written paper for each of the four files discussed.
• #15: Again, here are the tables from the SPSS output that shows the information I just discussed.
• #17: For our research project, we used a multinomial logistic regression because it was best suited to modeling consumer choice. Using this method, we developed two models for determining how consumers who first come into contact with a retailer via the company’s catalog and consumers who first come into contact via the company’s internet site differ from those who first come into contact with the company through their retail stores. We also created a bar chart of the number of sales in each month for each channel in order to determine if there are any seasonal effects in any of the channels and what their magnitude might be.
• #18: Following are the equations for determining if a consumer will be more likely to choose the company’s catalog or internet site for their first channel instead of the brick and mortar storefront.
• #19: The McFadden R-square value for these models is .133.
• #20: For our dependent variable, we used the first channel through which the consumer came into contact with the company. By knowing the differences consumers who first come into contact with the company through the store versus those who first go through the internet site or catalog, the company can tailor the marketing messages presented through those channels to be the most effective. For our independent variables we analyzed the following: Distance away from a store location because it is possible that the distance from a store makes the internet or catalog channels more convenient. Customer Age because consumers of a certain age group may be more or less technologically savvy and comfortable with online shopping, and therefore more or less likely to first conact the company through their website. The month that the customer made contact with the company because the time of year that a person is shopping may make convenience and selection a greater determining factor in choosing a channel, and thus drive up catalog and internet traffic. A customer’s income level may make them place a greater value on the convenience of the internet site or catalog. And we measured whether or not the customers gave the company permission to contact them via email. We believed consumers who chose the internet channel may prefer to have their interactions online and be more likely to give the store permission to contact them via email.
• #21: It was necessary to recode the FirstYYMM data in order to get the number of sales by month. Every data point ending in 01 was recoded as January, every data point ending in 02 was recoded as February, etc., until we had sorted 84 categories down to the 12 months. We had two hypotheses to test: Customers who came through the internet site would be younger than those who came through other channels. Customers who lived farther from retail locations would be more likely to choose the catalog or internet channels. Beyond these two hypotheses, this was a discovery search to find what else might affect channel choice.
• #23: The dependent variable FirstChannel in the nominal regression has the values Ret, Int and Cat indicating Retail, Internet, and Catalog channels respectively. After several calibrations of using several different independent variables, we settled on the following variables causing impact on the dependent variable with Ret (retail as reference): StoreDist = Distance to nearest retail location AgeCode = Codes (1-7) for grouped ages IncCode = Codes (1-9) for group income brackets Email = Y (yes) or N (no) FirstMonth = derived from FirstYYMM the MM part (Jan (01) – Dec (12)) The FirstChannel provides how the customer initially bought from the multichannel company. Understanding the customers’ first purchase may lead to understanding how to market to these customers allowing the company to increase profits and market growth. Additionally, building customer loyalty by segmenting these customers and their buying channels, should encourage future purchases by these customers.
• #24: StoreDist: Significance level indicates the further away or greater the distance the customer lives from a brick-n-mortar store, the more likely they are to make their first time purchase on the Internet. Therefore, Retail first time purchases increase with customers having a shorter distance to a Retail location. Segmenting these customers within retail locations and marketing/advertising with store coupons and flyers; using Internet marketing/advertising to those not within reach of the retail locations; and further segmenting non-internet using customers by use of catalog would make the most sense. AgeCodes: All fell within the 5% significance level, however the age groups from 18-24 years old and 65-74 years old have less effect on the dependent variable. The younger aged most likely do not have the income to spend while the elderly have less impact because they probably do not spend much time on or perhaps never use the Internet.
• #25: IncCode: A couple of the income levels fall out of the 5% significance level. Those in the low incomes levels (under 20K), and those in the higher income level (100K and above), both are above the 5% tolerance level, therefore not rejecting the null hypothesis. It appears the income range of 30K to 99K has a likely effect of making a first time purchase on the Internet. You have to have money to spend money. Email: For the Catalog, analysis does not show an impact and falls out of the 5% significance level. It is significant for the Internet customer where most likely email is a way of communication for billing, order receipt, etc. With technology advances, many more customers have the ability to order on the Internet and long as the customer remains receptive to this channel, it may push down catalog orders. Exceptional customer service drives catalog orders, which usually means the multi-channel company invests in such practice and keeps it as part of the business model.
• #26: FirstMonth: It appears a few of the months fall out of the 5% significance level for both Internet and Catalog. These months include the summer months of Jun, Jul, and Aug and the month of Oct for Internet, which again creates opportunity for first time purchasers in the holiday months to use the channel of their preference. Catalog has the months of Jan, Nov, and Dec as solid months of first time purchases. Internet has the months most effecting first time purchases as: Jan, May, Nov, Dec. These key months should provide the multi-channel company an opportunity to build brand loyalty efforts and encourage return purchases by marketing/advertising to those first time purchasers.
• #27: Graph of Overall Population initial total dollars spent by FirstChannel by FirstMonth This Graph indicates that around the holiday season – November and December – are the peak times of the year where new customers spend the most money. For Catalog and Internet, the lowest spending months are in summer – June, July and August.
• #28: Graph of Overall Population initial customers new by FirstChannel by FirstMonth This graph indicates that around the holiday season, November and December are the peak purchasing months for new customers. For Catalog and Internet, the lowest is summer (June, July, and August).
• #30: Section 5, Summary and Conclusions Group 4 found through its analysis of a multi-channel retailing organization dataset that its revenue of gift baskets is highly seasonal, as evidenced by its mail order and catalog holidays sales peaking at 6 to 8 times higher than any other time of the year. Their third channel, retail storefronts, also has a holiday sales peak in November and December, but the peak is much smaller than the other channels and sales that are much more consistent throughout the year. This organization is most successful selling to middle aged, high income homeowners. With the explosive growth of Internet sales and the overall decline or stagnation of mail order and retail storefronts, this company needs help analyzing sales to maximize the growth and synergies of their three retail channels.
• #31: Group 4 used Multinomial Regression testing using a consumer choice model with heavy reliance upon IBM’s SPSS software tool. Two models were developed to assess first time mail order purchasers and first time E-Commerce purchasers. First Channel Catalog and First Channel Retail were the dependent variables with store distance, customer age, first month of contact, income level and email as independent variables. Two hypotheses were developed regarding affect of distance from retail store on Internet sales and a question as to whether younger demographics prefer the Internet channel.
• #32: Several conclusions were evidenced by the analysis. For instance, the farther away a customer lives from a store, the more likely they are to use the Internet and mail order channel for their first-time purchase. This indicates that physical retail storefronts pull sales away from other channels. It can also be said that the absence of a conveniently located storefront will push sales towards the Internet and mail order channel. Therefore, this model can be used to fine-tune their retail store rollout to properly balance the three channels, depending on their overall strategy. The data demonstrates that despite the overall stagnation of retail, a retail store helps to provide more consistent revenue and subsequent cash-flow throughout the year. We find that the model is significant for middle-aged and middle-income consumers, but has weakness in the younger segment which has less disposable income and in the older demographic, which has lower use of the Internet. We also find that some people just don’t want to use the Internet channel and prefer the traditional storefront experience.
• #33: We therefore conclude that while the November and December sales peak is huge for Internet and Mail Order, it still exists for the retail storefronts, but with a lower impact. We believe that the retail storefronts can be used to smooth-out revenue stream throughout the year. Also, even though mail order and retail is down, the data shows that it cannot be considered out. First time buyers still prefer the company’s traditional mail-order channel, but the Internet is close behind. Retail is a distant third, but still significant. We also believe that some will always prefer the brick-and-mortar experience and simply don’t trust the Internet, especially for the older demographics. The challenge will be slowly converting traditional mail-order clients to Internet consumers of their product, especially for those for which the retail store front is not an option. The Internet channel should be continually improved and heavily marketed to ensure it picks up the customers that migrate from the mail order channel. With more and more options available for consumers today, the company must not simply wait and see what happens with the mail-order versus Internet battle, but continue to market and support both channels to retain these customers, regardless of which channel they come in on.
• #34: Several opportunities were uncovered through our analysis. For instance, this model can be used specifically to calibrate where to locate retail stores to determine where a physical store best maximizes revenue without pulling too much revenue from mail order or catalog. However, another way to see this issue is this model can help to push revenue to the Internet and mail order channel simply through the unavailability of a retail store front. Therefore, the company an use this model to help develop the proper strategy and store rollout. There appears to be an opportunity for this company to go down-market with its products, based upon its heavy reliance on older, higher income professionals. The younger demographic should be targeted with products and marketing methods, particularly through the Internet channel, to ensure a steady supply of new customers moving into the future. We find that while the traditional mail-order channel is still the revenue leader for this company and is therefore not going away quickly, it is clearly a trend that the Internet channel will one day dominate sales, especially as the younger demographic ages with its heavy usage of the Internet. Slow conversion of mail-order consumers to the Internet along with heavy marketing of the Internet channel to bring in increasing numbers of Internet first time buyers appears to be the best way to ensure continued healthy revenue growth into the future.