Vol. 5, No. 10; October 2005 Printer Friendly PDF version

THE BUSINESS VALUE OF CUSTOMER DATA: PRIORITIZING DECISIONS

by Gabriele Piccoli, Senior Consultant, Cutter Consortium; Assistant Professor of Information Systems at the School of Hotel Administration at Cornell University

INTRODUCTION

Cutter's survey on customer data indicates that more than half (59%) of the 106 participating organizations are currently attempting to extract value from their customer data; the majority of remaining organizations have minimal efforts in place (22%) or are in the planning stages (16%), while 3% of organizations have no plan to use customer data (see Graph 1 in the Survey Data section beginning on page 27). This insight shows that organizations clearly recognize the business value of customer data. While there is no doubt that customer data has business value for most firms, implementing a formal strategy is difficult. Our survey reveals that just over one-third of respondents (36%) have a formal strategy in place for using customer data (see Graph 2).

Extracting value from business data requires an attentive cost-benefit analysis. Such cost-benefit analysis, particularly when focusing on customer data, is not straightforward because it calls for us to make a number of assumptions about the uncertain gains the use of the data will yield and for an estimation of costs associated with collecting, storing, processing, and distributing this customer data. Our respondents find that showing acceptable ROI on projects that use customer data for analytical purposes is the primary challenge to getting these projects approved (43%), closely followed by inertia and resistance to change (42%). (See Graph 3.)

While discerning how to optimally extract value from customer data is not simple, I believe there is significant value in engaging in the analysis. When performing this analysis, it has been my experience that it pays to go from the general to the specific. This approach allows us to first assess what class of initiatives fits with the firm's environment and is aligned to the firm's strategy and positioning. Once we have an appreciation for the general category of appropriate initiatives, we can turn our attention to prioritizing amongst the set of all the possible uses of the data, using data already available or to be captured.

In this article, I follow much the same progression. My objective here is to offer some thought-provoking ways to think about your organization and how you can use customer data. These models are grounded in information systems research and a number of case analyses. Yet they are simplifications of reality and are not intended to be exhaustive or "silver bullet" solutions. Instead they are designed to bring discipline to the analysis of a very open-ended and creative process.

INDUSTRY ANALYSIS

The point of departure for the industry analysis is the recognition that your firm specializes in the sale of a given set of goods and services. The characteristics of these products and services, barring change of product mix or significant innovation, are fixed. It is therefore within the constraints of these characteristics of "what the firm does" that we need to look for the degrees of freedom available to you when it comes to crafting a customer data strategy.

Let's look at a brief example. During the heyday of the customer relationship management (CRM) solution, personalization was all the rage. Many large consulting firms were proposing personalization and customization solutions to all kinds of clients. One firm I recently worked with (a cruise line) called this the "big bang" personalization approach. The consulting firm the cruise line had engaged was pressing it to "go personal." "Because you can collect so much data about your customers' habits and preferences," the consultancy reasoned, "you can create deep profiles of your returning customers. Once you know that Joe Cruiser likes Corona beer, you can have a few chilled bottles waiting for him in his cabin when he embarks. It is hot at the port of Miami, and he is tired from the embarkation process, but as soon as he checks in, he can unwind with his favorite drink. He'd love you for that and will never want to cruise with another line!"

Those among us who have recently cruised know that there is no cash onboard a modern cruise ship. Rather, every cruiser has a card that serves as an identification document when leaving or returning to the ship at various ports of call; functions as a room key; and, most importantly, acts as a debit card to pay for any one of the hundreds of onboard services. It is certainly true that since the introduction of these cards, cruise lines have the ability to unobtrusively collect huge amounts of individual-level behavioral and preference data -- but is personalization really the best way to use this customer data? As I discuss here, there are other strategies that can be enacted with the use of customer data -- personalization is just one of them. How do you then decide how your own organization should use the customer data it is able to collect?

Repurchase and Customizability: The Dimensions of Decision Making

The industry analysis I propose here helps answer this question by looking at two fundamental variables: the theoretical repurchase frequency in the industry and the degree of customizability of the product or service being offered.

The dimension of theoretical repurchase frequency represents the regularity with which the average customer acquires goods and services offered by the firms in the industry or segment of interest. This measure is concerned with the potential for high repurchase frequency, not with the actual repurchase rates any specific firm is experiencing -- hence the use of the term theoretical repurchase frequency. If a firm in an industry characterized by high theoretical repurchase frequency has very few returning customers, this is a signal that the company is missing an opportunity. Imagine going to a fast-food restaurant in your neighborhood and finding it dirty and painfully slow in service. If this state of affairs is not quickly rectified, you most likely will not return to the same store. But will you stop patronizing fast-food restaurants in general? Probably not; you'll just shift your demand to one that does an acceptable job. The key point here is that theoretical repurchase frequency is a function of the industry the firm is in and the characteristics of the product or services it offers. It is not determined by how well your firm fulfills the value proposition to its customers. Car manufacturing and real estate are typical examples of industries characterized by relatively low repurchase frequency. At the extreme end of this spectrum is a product sold in the industry in which I am a member: the Master of Business Administration (MBA) product has a theoretical repurchase rate of zero -- once a student has obtained one MBA, she will likely never enroll in the same or a competing program again. How good a job we did and how satisfied she was with the experience is irrelevant -- people generally don't need two MBAs. Coffee shops and grocery stores are at the other end of the spectrum; they enjoy high theoretical repurchase frequency, as do pizza delivery businesses (particularly those who serve my onetime MBA customers).

Degree of customizability represents the extent to which the product or service your firm offers can be tailored to the specific needs and requirements of individual customers or a segment of customers. This dimension is a function of the complexity of the product or service being offered. Gasoline, as most commodities, is an example of a product with a very low degree of customizability. Airline service, creative attempts by innovative firms like Virgin notwithstanding, represents another example of a service with limited degree of customizability. After all, what an airline offers to most of its customers is a seat on a plane that gets them from point A to point B. Of course, this simplistic view of airline service can be challenged if we define the product being offered in terms of transportation, or reliable transportation. Yet the success of low-cost airlines seems to suggest that price is a key driver of the customer airline purchase decision, thus leading us to conclude that differentiation is very hard in this commodity-like industry. If some debate can be had about the degree of customizability of airline service, it is quite clear that there are industries that offer highly customizable products and services. Large resorts and destination spas are a perfect example. Canyon Ranch, the leading luxury destination spa in the US for example, offers upward of 230 paid spa and health and healing services ranging from the most exotic massage and therapeutic services to medical procedures and lab tests worthy of the most advanced hospitals in the country. Canyon Ranch also offers countless free fitness classes, wellness and self-improvement lectures, and outdoor activities. It is this staggering array of options that allows every Canyon Ranch guest to "experience a different Canyon Ranch" as the firm is fond of saying. Moreover, each service can further be tailored to the unique needs of the guest by changing lighting, music, scent, intensity, and so on. Cruise lines represent another example of an industry that sells highly customizable products.

Customer Data Strategies

When taking into account the characteristics of the industry your firm is immersed in and the product and services that it offers, the decision matrix depicted in Figure 1 can be used to find a matching customer data strategy.

Figure 1 -- Decision matrix.

The matrix identifies four quadrants. While your firm may or may not fit neatly into one quadrant, the matrix will help you evaluate the advantages and disadvantages of each general strategy and, more importantly, the natural fit of each of the four approaches to your firm's characteristics.

Personalization

A typical service personalization or product customization strategy is most appropriate for firms competing in industries characterized by both a high theoretical repurchase frequency and a high degree of customizability. Under these conditions, the potential is there to collect significant individual-level data because of the repeated interactions the firm has with its returning customers. Moreover, because of the high degree of customization, management has many opportunities to use this information to tailor the product or service to the specific needs -- learned or inferred -- of the returning customers. Thus, the firm can use the information to modify its operations and differentiate its product or services. Event planning is a good example of an industry that fits in this quadrant -- particularly those firms that work closely with customers that need organization of many recurrent events (e.g., large investment banks). Another example is large IT vendors catering to business customers with complex business and IT requirements.

Rewards Strategy

A rewards strategy is predicated on the notion that the firm's product and service will be purchased frequently but that these products are fairly standardized and therefore it is difficult for the organization's managers to tailor them to specific customer requests. Under these circumstances, the firm can use customer data to evaluate the profitability of each customer -- actual and potential -- and then use this information to reward behavior in an effort to increase customer loyalty or boost share of wallet (i.e., to ensure that customers consolidate their purchase behavior in the industry by sourcing from the company rather than competitors). The firm can also use the individual-level data collected to generate accurate reports and improve its operations (e.g., grocery stores performing basket analyses). Note that this means understanding customer profitability as well as their propensity to repurchase without incentive -- a strategy much more complex and sophisticated than the "buy nine coffee cups and receive the 10th free" that many firms seem to settle for. The passenger air transportation industry is a classic example for this quadrant.

Acquisition Strategy

Much conventional thinking about strategies based on customer data seems to imply that when an industry has little potential for repurchase (i.e., low theoretical repurchase frequency), customer data is not worth using. This could not be further from the truth. Even in the face of low theoretical repurchase frequency, a firm in an industry with a high degree of customization may benefit from an acquisition strategy. Following this approach, the firm collects exhaustive data about its current customers in an effort to profile them and develop predictive models to identify and attract new profitable customers while avoiding nonprofitable and marginal ones. The availability of such deep business intelligence becomes all the more important during slow periods when marketing budgets get slashed and efficiency in attracting new profitable customers becomes paramount. A good example of an industry that falls in this quadrant is the wedding reception business -- an industry that offers highly customizable products but typically enjoys low repurchase frequency.

Low Potential

When a firm is in an industry characterized by low theoretical repurchase frequency and relatively low degree of customizability, there seems to be little potential for crafting a strategy around customer data. This is because very little data will likely be generated, and managers' hands are tied in respect to what they can do with it. A chain of budget or limited-service tourist hotels in an exclusive fly-in destination (e.g., Hawaii, Fiji) offers an apt example. Midscale hotels in these locations are generally a "window on an experience" rather than the experience itself and their value proposition is to offer guests an affordable opportunity to experience a great location. Because of the time commitment and cost of reaching these destinations, repurchase is relatively infrequent. Thus, there is little opportunity to enact any of the three strategies discussed above. Under these conditions, the firm is better off focusing on efficiency and low prices and avoiding the cost of collection, management, and analysis of customer data.

Though we can argue that every business likely has both recurrent and rare customers and thus is difficult to unequivocally fit into a quadrant, the use of the theoretical repurchase frequency dimensions is important because, under resource constraint, a business cannot be all things to all people and the various programs generally have high fixed costs. Therefore, the potential for payoff needs to be looked at over the entire customer base (or a large proportion of it) so that the fixed cost can be spread amongst a large number of customers. In other words, there is no point in doing significant personalization (creating the systems and processes to training of people, and so on) unless the personalization strategy can yield a return by being valued by a large number of customers or a limited percentage of customers who account for the majority of the firm's profits.

Remember my cruise line example earlier? When analyzed through the lens of the matrix, it becomes clear that a personalization strategy, though intuitively appealing, is probably not optimal. The cruise product is typically characterized by low repurchase frequency, for as inexpensive as cruises have become, they remain a fairly pricey vacation option, and the cycle of repurchase is typically long (e.g., a honeymoon cruise followed by a five- or 10-year anniversary cruise). How likely is the firm to be able to profit from the (considerable) investment in a personalization strategy? Would it not be better to focus on an acquisition strategy designed to attract profitable first-time cruisers, based on what the firm learns by analyzing and clustering the profile of its past cruisers?

Like the cruise line in the example, many of our respondents have already moved beyond a sole focus on personalization and are using the gamut of customer data strategies available.

How Feasible Are These Initiatives?

The strategic initiatives described above are of course predicated on a firm's ability to capture the necessary customer data. The results of our survey show that more than half of the organizations we queried (61%) find the current degree of detail of their customer data somewhat or not at all acceptable given their intended strategy for its use (see Graph 4). Moreover, just over one-fifth of the respondents (21%) feel that their organization is effective at leveraging the available customer detail into usable analyses (see Graph 5); rather, the majority of respondents (41%) claim a spotty performance and more than one-third (38%) consider their results insufficient. Given these potential difficulties and the cost associated with using customer data, it is imperative to evaluate how difficult it will be to collect, store, process, and distribute the needed data during the planning process. While proposed initiatives must be evaluated individually (see below), it can be useful to think about this dimension early on during the industry analysis to evaluate how amenable the firm's operations are to the implementation of customer data strategies.

The expanded model shown in Figure 2 directly acknowledges that different industries, because of the general norms about how business is conducted in them, offer different potential for data capture. In other words, the degree to which data collection can be done easily varies dramatically by industry and is an important early consideration.

Figure 2 -- Industry analysis: data capture.

Degree of Unobtrusive Data Capture

Despite being a mouthful, the degree of unobtrusive data capture is a largely intuitive concept. It indicates the extent to which, in the normal course of business, customer data is collected and stored in a readily usable format. For example, we are all painfully familiar with the "Get paid for your opinions!" e-mails that clog our in-boxes or the guest satisfaction surveys we rarely fill out. We largely consider both to be a disruption and a waste of our time. Conversely, we are outraged when our doctor pushes us out of the door quickly without gathering much information about our medical history, allergies, and symptoms before making a diagnosis. The doctor's office is a place where "getting surveyed" is not a waste of time, but rather something we expect. The difference, of course, is in customer expectations as to what the encounter with the firm should be like -- the norms within the industry. While it is OK for a hospital to ask us for our Social Security number, we would be outraged if the coffee shop down the street did.

This simple example shows that some firms are highly constrained when it comes to gathering customer data and may have no better way to obtain it than to pay a representative sample of customers to take time to respond to a survey. Other organizations, however, have more data than they can ever hope to use. An early analysis of practices in your industry can be illuminating. Imagine, for example, a fine dining restaurant in an industry with high repurchase frequency and a relatively high degree of customizability of the experience. A personalization strategy is quite suitable for such an establishment, yet much of the data needed to carry it out is generated in fleeting customer-waiter exchanges that are difficult to capture and codify for easy storage and retrieval. Add to this the typically high turnover in the food service industry, and it becomes clear why, for as much patronage as we pay our favorite restaurants, we generally don't receive a commensurate degree of personal service. Compare the difficulty a restaurant has with collecting and storing its customer data in a readily usable format to the relative simplicity of the same task at an online retailer, such as Amazon.com. Granted, the potential depth of the relationship is much lower, but the ease with which Amazon can collect, store, and process the data you provide is much greater.

While the degree of unobtrusive data capture for a firm is largely given at any point in time, technology improvements and innovation may pay off here if you are willing to shoulder the cost of changing people's habits. For example, while much of the information about customers' gambling behavior in casinos was traditionally left to busy and fallible casino hosts and pit bosses, the advent of electronic slot machines ushered in a new era. Casino executives realized that a modern slot machine is, in essence, a digital computer, and as we know quite well, a computer records all its transactions with great speed and accuracy. Once customers were convinced to use magnetic strip cards -- and assured that the card would not jinx them -- tying these transactions to individual customers was a relatively small step. Today the natural course of business in the casino industry is such that a company can have an accurate, real-time picture of each of its customer's slot-playing behavior -- a fact that has allowed operators like Harrah's Entertainment to wring huge value from this customer data. The move of cruise lines away from cash onboard to the use of cards introduced a similar highly accurate data collection infrastructure. Ironically, the elimination of cash on cruise ships was driven by a typical transaction-oriented focus: to control theft and increase the efficiency of onboard operations. Yet its main benefit in the long run may very well be its unobtrusive data collection potential.

THE INFORMATION SYSTEMS CYCLE ¹

The cruise industry experience described above is more typical than one would think and reminds us of the importance of understanding the full information systems (IS) cycle. With so many IT innovations being justified by their transaction processing and automation contributions (i.e., efficiency), we see myriad instances of firms that fail to capture the full benefit of the data these transactional systems generate. An important insight then is that the data generated by transaction processing systems (TPS) during the natural course of business often has value beyond the completion of the business transaction it was designed to process. Figure 3 illustrates how transactional data, when stored with some attention to future uses, can become the raw material for the development of valuable business intelligence and knowledge about business operations.

Figure 3 -- The information systems (IS) cycle.

Take, for example, a grocery store. As customers come and go, a wealth of transactional customer data is generated. This information is necessary for the store to manage its daily operations -- in other words, to handle the present. Customers need to find shelves stocked with popular items. Managers must optimize inventory in order to minimize working capital and spoilage of perishable items. Correct billing, taking into account promotions and discounts, must be generated at checkout. All these instances describe events that occur during the normal course of business in a grocery store. When customers leave, and the present has been handled, all the data generated that is not subject to legal requirements may be thrown away. Alternatively, the store can create data repositories that catalog and hold this information, now historical in nature. This act of "remembering the past" is only meaningful, of course, if the data can be used at a later stage to prepare for the future -- for example, to gain beneficial intelligence through analysis of the data.

The IS cycle is a simple, intuitive model. Yet it shows how the cost of customer data initiatives is a function of how difficult it is to go from handling the present to effectively preparing for the future. This cost must be offset by the benefits associated with preparing for the future. An analysis of your firm and the customer data it collects -- and could collect -- offers the opportunity to prioritize amongst countless available initiatives.

PRIORITIZING STRATEGIC INITIATIVES

Given the wealth of information available to the modern business, it can be extremely confusing to figure out where to start. Yet the importance of execution cannot be underestimated. To aid in this decision, a simple prioritization matrix can be created based on two dimensions: revenue lift potential and availability of the needed information (see Figure 4). The first dimension provides an assessment of the financial benefits associated with the initiative. The second dimension provides an assessment of the immediacy with which the initiative can be implemented and of the costs associated with it -- the higher the availability of the needed data, the cheaper the successful implementation of the initiative.

Figure 4 -- Initiative prioritization matrix.

When initiatives are mapped to each of the four quadrants (see case study below), it becomes clear what can be quickly implemented, maybe as proof-of-concept or to gain support from other executives. It also becomes clear which initiatives are resource-intensive and require a much higher level of organizational commitment. I have identified the following four principal categories: imperatives, quick wins, tradeoffs, and losing causes.

Imperative

In this quadrant fall projects that have high revenue lift potential and rely on readily available information. Because of the ready availability of the needed information, these initiatives are usually quickly implemented at a relatively low cost. They therefore deliver an appealing cost-benefit ratio, and not implementing them is tantamount to leaving money on the table.

Quick Wins

In this quadrant fall projects that, while not having vast revenue lift potential, can be readily implemented based on immediately available information. I call these initiatives "quick wins" because they can be implemented easily, therefore not requiring significant resources and a demanding approval cycle.

Our survey data indicates how difficult it can be to show ROI and overcome inertia for projects that use customer data. Thus, imperatives and quick win initiatives can be used as proof-of-concept to gain momentum and to build credibility with other executives. This credibility can then be put to use when making the case for tradeoff initiatives.

Tradeoffs

In this quadrant fall projects that have significant revenue lift potential but rely on information that is not readily available. Consequently, they tend to be quite costly. This may be because the information is not easy to capture, is not in a readily usable format -- customer preference at a restaurant, for example -- or because the initiative requires the pooling of substantial information from multiple sources requiring substantial integration. I call these initiatives "tradeoffs" because they require substantial cost-benefit analysis and a rigorous approval cycle before the allocation of the needed resources can be justified.

Losing Causes

In this quadrant fall projects that have little revenue lift potential and rely on information that is not readily available. These are projects that, though they may have appeared to be good ideas in principle, generally offer little bang for the buck. Initiatives that fall in this category should not be implemented unless the cost associated with making the needed data available can be justified and assigned to other projects with positive ROI. In other words, these initiatives should be shelved until a change in circumstances moves them to the quick wins quadrant.

Using the prioritization matrix requires creativity and knowledge of the specific company context. I illustrate its use with a rather straightforward case study.

CASE STUDY: ONLINE DISTRIBUTION AT THE INDEPENDENT

The scenario we present is set in the lodging industry and draws on a set of real experiences and data from an existing independent hotel we'll call The Independent. The customer data of interest refers to prospective travelers' online booking behavior. As many observers have noted, the e-commerce revolution has created huge potential for unobtrusive data collection since all e-commerce transactions are computer-mediated and therefore easily recorded and archived. In very few other situations can transactions be tracked with such precision, comprehensiveness, and depth.

The Independent is an upscale property in New York City; it has no chain affiliation but is a strong brand with good name recognition. Management at The Independent was very forward-looking and saw the potential offered by online distribution very early. It created its first Web site in the mid-1990s and developed a channel strategy to improve its online exposure. What The Independent did not realize early on, however, was how interesting the footprints left by its online customers could be.

The IS Cycle

The IS cycle can help us quickly map the data The Independent has available and offer early insight into how it could be used. Table 1 presents a brief inventory of the data typically generated by online shoppers.

The information described in Table 1 is, of course, necessary to enable online customers to browse through The Independent's offers and to enable them to complete online transactions. If appropriately organized and stored, however, this data can also prove very powerful for later analysis. Examples of the type of analysis that can be performed on this data are reported in Table 2.

Table 1 -- Handling the Present: TPS Data

Table 2 -- Preparing for the Future: Data Analysis

The analysis shown in Table 2 allows The Independent to better prepare for the future and to be more efficient in the deployment of its resources. For example, it would allow the firm to keep track of the sources of business that consistently provide qualified travelers who end up booking -- rather than simply browsing -- through the hotel's Web site. When faced with a tightening of the marketing budget, The Independent would have concrete data upon which to decide how to optimize its online advertisement mix.

When additional data -- such as demographic and geographic customer profiles, guest history data, and booking statistics from the other channels of distribution -- is appended to that obtained directly by way of the interaction, further analysis becomes possible. The Independent can, for example, better evaluate the geographic and demographic characteristics of its online booking population. This analysis would further enable the firm to select the appropriate channels to reach this audience. Discrepancies between the profile of the overall customer base (i.e., travelers who stay at The Independent) and the profile of the online customer base would highlight opportunity segments to be moved to the online channel. ²

It is crucial to note how all of the data compiled in Table 1 is recorded automatically and unobtrusively by the Web server and the booking engine: the customer does not have to do anything for this data to be recorded. Of course, the primary reason for this data generation is transaction processing (i.e., to allow the customer to receive the Web pages it wants to see and allow her to look for availability and make reservations). Yet, as with most transactional data, carefully feeding it through the IS cycle offers the potential to create value by repackaging and analyzing it.

Prioritizing amongst the wealth of possible uses of the data collected and stored becomes important. Applying the two dimensions of the prioritization matrix offers a structured approach to this process.

Imperative: Real-Time Channel Monitoring

With the number of travelers who book online steadily increasing, it is crucial for hotels to monitor their different channels of distribution in real time. This monitoring can be done using no additional data than what a hotel like The Independent has ready access to (i.e., availability: high) and has a potentially significant impact if rates across channels fall out of alignment (i.e., revenue lift potential: high).

The Independent, for example, was enjoying consistent Web site traffic and an equally consistent level of interest from Web site visitors shopping for rates and availability. This consistent flow of interested shoppers is shown by a flat visitor-to-look ratio, where the first variable is the number of unique visitors and the second is the number of unique availability searches by each visitor. With a consistent visitor-to-look ratio, it becomes important to monitor the firm's ability to convert these visitors into paying customers -- the look-to-book ratio. In the case of The Independent, though, a post hoc analysis revealed that, over time, the look-to-book ratio had steadily eroded well below the typical 6%-8% the hotel had grown accustomed to.

If The Independent had been tracking the performance of the proprietary direct channel (i.e., its Web site) in real time -- something it easily could have done with the available information -- it would have quickly detected the discrepancy. This simple analysis should have tipped management off to the fact that, while the marketing efforts of The Independent were steadily driving traffic to the Web site, some other channel was probably underpricing the hotel. As a consequence, the people coming to the Web site were not booking there because they were able to find a better deal elsewhere. Had it been armed with this timely information, The Independent could have reversed this trend quickly.

A follow-on analysis reveals that the above explanation of the falling look-to-book ratio is accurate (see Figure 5). While bookings posted by The Independent's own engine (i.e., its Web site) and by Internet retailers had closely tracked during the third and fourth quarters of 2003, a trend reversal occurred in early 2004. Specifically, during the second quarter of 2004 and beyond, the bookings made directly through The Independent's Web site had fallen by half -- even though the visitor-to-look ratio had remained unchanged.

Figure 5 -- Bookings by channel of distribution.

Moreover, since The Independent has access to average daily rate (ADR) figures by channel, it was able to compute exactly how much not catching the dip in the look-to-book ratio in a timely fashion cost the firm. Assuming that the channel mix for the bookings placed in the first quarter of 2004 remained unchanged during the second quarter, the cost of not detecting the drop can be easily computed. In this case, not using readily available data cost the firm thousands of dollars per month in lost revenue (see Figure 6).

Figure 6 -- Average daily rate (ADR) by channel of distribution.

Quick Win: Referring Site and Keyword ROI

With online travel becoming more and more important as a source of business for hotels like The Independent, prioritizing online spending can help a firm be more efficient in its allocation of marketing resources. Because the booking engine behind The Independent's Web site automatically tracks referring sites (e.g., search engines, partners' sites) and any keyword used in searches that lead to the Web site, availability of the needed data is high. But with only 2%-5% of business currently being booked through the proprietary Web site, the revenue lift potential for The Independent, at least in the short term, is relatively low. Note that the information obtained through this analysis is different from a click-through ratio produced by the referring site. Click-through ratios, while valuable, provide no insight as to what the customer did once he reached the firm's Web site. This further analysis is crucial because when the pay-per-click revenue model is used, the buyer pays the same for a visitor that leaves the site immediately as for one that ends up making a purchase.

Nonetheless, with the cost of a keywords marketing strategy easily in the several thousands of dollars, computing an ROI on referral sources and keywords can be useful knowledge and would set up the firm to better target online customers as they become an increasingly important segment of the overall customer base.

The first result The Independent obtained from this analysis concerns referring sites and keywords that had extremely low visitor-to-look ratios. The immediate conclusion is that these sites and keywords were producing unqualified traffic or mismatched customers who, once they reached the site, left without shopping. A second possible result from this analysis is a ranking of referring sites and keywords by look-to-book ratio. Such ranking allows management to quickly identify the combination of referring site and keywords that is producing the most qualified traffic -- that is, those with the highest conversion ratio. Finally, computing the average rate and average booking lead time for each site and keyword enables management to see beyond traffic and conversion statistics and to factor in the booked rate these customers pay. This analysis is important because the highest look-to-book ratio may be produced by keywords that signal deals on distressed inventory; these keywords, however, may have a lower overall ROI than keywords that produce lower conversion ratios but a higher average rate (e.g., referring sites and keywords that capture a highly targeted segment of the customer base with low price sensitivity).

Tradeoffs: Dynamic Pricing

One of the key idiosyncrasies of the lodging industry, and of the hospitality industry at large, is that customers are often willing to provide a firm with significant personal information in order to receive better, more tailored service. Harnessing this personal information has remained a fairly elusive target for the industry to date. Yet significant efforts have been made, generally under the umbrella of CRM, to better know customers, discriminate profitable from unprofitable ones, and generally offer differentiated service to more valuable customers.

An initiative that The Independent has begun considering relies on being able to dynamically price its inventory, keeping into account not only typical revenue management variables (e.g., booking dates, length of stay) but also the identity of the prospective guest. This initiative requires the ability to track individual customers -- for example, asking customers to input their frequent guest ID number prior to querying the search engine -- and to be able to pass this number through to the engine. From a technology standpoint, these are both nontrivial tasks (i.e., availability: low). This initiative has elements of revenue management (e.g., optimally pricing a room) as well as marketing and operational effectiveness (e.g., loyalty, better customer targeting, customer satisfaction) and therefore has the potential for significant revenue lift.

While this initiative may have significant appeal at face value, it requires a major set of operational changes, impacting revenue management, operations, marketing, and so on. Moreover, because of the low current availability of the necessary data and the need for technology upgrades, its implementation is not immediate or risk-free. Finally, computing an ROI associated with its implementation is a nontrivial feat. It follows then that the initiative is likely to be met with significant resistance.

CONCLUSIONS

As our survey reveals, most organizations realize that customer data has significant value. How to extract this value from the heaps of data available, though, is a process not without challenge, as our respondents confirm. Yet with the amount of customer data proliferating, and with customers becoming increasingly more discriminating, the ability to harness the value of customer data is becoming a crucial skill. As you hone these skills, you will find the framework outlined by my colleague Ken Collier in the next article to be a comprehensive guide to how to select high-potential initiatives and how to ensure that the initiative is implemented successfully.

The purpose of my contribution was to stimulate your thinking and to offer analytical tools that may help you leverage the customer data you currently have. Successfully using the models I present here is as much an art as a science, requiring creativity and, perhaps most importantly, an inquiring attitude and a sense of possibility. I hope that these models will energize you to engage in this pursuit as, I believe, the rewards can be great for those who endeavor in this journey.

NOTES

¹The idea of the information systems cycle was first formalized by Professor Rick Watson of the University of Georgia. I have borrowed and adapted it here.

²Note that moving any portion of the customer base from an offline or an intermediated online channel to the branded Web site is a positive result since the online direct channel has the lowest variable cost per transaction.