Customer extension in E-commerce

Customer extension has the aim of increasing the lifetime value of the customer to the company by encouraging cross-sales, for example an Egg credit card customer may be offered the option of a loan or a deposit account. When a customer returns to a web site this

is an opportunity for cross-selling and such offers can be communicated. Direct e-mail is also an excellent way for informing a customer about other company products and it is also useful in encouraging repeat visits by publicizing new content or promotions. E-mail is vitally important to achieving online CRM since the web site is a pull medium which the customer will only be exposed to if they decide to visit the web site and they are unlikely to do this unless there is some stimulus. However, e-mail is a push medium where the cus­tomer can be reminded about current promotions and offers and why they should visit the web site. This is why it is so important to capture the customer’s e-mail address at the acqui­sition stage. The use of direct e-mail to communicate promotions to customers presents a dilemma to the online marketer since although it is potentially powerful in achieving new orders, as explained in the section on retention, if it is used too frequently or e-mail is unso­licited then it may achieve the opposite to the desired effect and the customer will be lost.

Many companies are now only proactively marketing to favoured customers. Seth Godin (1999) says ‘Focus on share of customer, not market share – fire 70 per cent of your customers and watch your profits go up! One UK financial services provider has analysed characteristics of high-churn-rate customers, and when a new prospect fitting this profile contacts the call centre they are actively discouraged. Using these techniques it is possible to increase share of customer.

1. Advanced online segmentation and targeting techniques

The most sophisticated segmentation and targeting (see Chapter 8 for an introduction) schemes for extension are often used by e-retailers, who have detailed customer profiling information and purchase history data as they seek to increase customer lifetime value through encouraging increased use of online services over time. However, the general prin­ciples of this approach can also be used by other types of companies online. The segmentation and targeting approach used by e-retailers is based on five main elements which in effect are layered on top of each other. The number of options used, and so the sophistication of the approach, will depend on resources available, technology capabilities and opportunities afforded by the list:

• Identify customer lifecycle groups. Figure 9.20 illustrates this approach. As visitors use online services they can potentially pass through seven or more stages. Once companies have defined these groups and set up the customer relationship management infrastructure to categorize customers in this way, they can then deliver targeted messages, either by person­alized on-site messaging or through e-mails that are triggered automatically due to different rules. First-time visitors can be identified by whether they have a cookie placed on their PC. Once visitors are then registered, they can be tracked through the remaining stages. One particularly important group is customers who have purchased one or more times. For many e-retailers, encouraging customers to move from the first purchase to the second purchase and then on to the third purchase is a key challenge. Specific promotions can be used to encourage further purchases. Similarly, once customers become inactive, i.e. they have not purchased for a defined period such as 3 months, they become inactive and further follow-ups are required.
• Identify customer profile characteristics. This is a traditional segmentation based on the type of customer. For B2C e-retailers it will include age, sex and geography. For B2B companies, it will include size of company and the industry sector or application they operate in.
• Identify behaviour in response and purchase. As customers progress through the lifecycle shown in Figure 9.20, though analysis of the database, they will be able to build up a detailed response and purchase history which considers the details of recency, frequency, monetary value and category of products purchased. This approach, which is known as ‘RFM or FRAC analysis’, is reviewed below. See Case Study 9.1 for how Tesco target their online customers.
• Identify multi-channel behaviour (channel preference). Regardless of the enthusiasm of the company for online channels, some customers will prefer using online channels and others will prefer traditional channels. This will, to an extent, be indicated by RFM and response analysis since customers with a preference for online channels will be more responsive and will make more purchases online. Customers that prefer online channels can be targeted mainly by online communications such as e-mail, while customers who prefer traditional channels can be targeted by traditional communications such as direct mail or phone. This is ‘right-channelling’, which was introduced in Chapter 5.

• Tone and style preference. In a similar manner to channel preference, customers will respond differently to different types of message. Some may like a more rational appeal, in which case a detailed e-mail explaining the benefits of the offer may work best. Others will prefer an emotional appeal based on images and with warmer, less formal copy. Sophisticated companies will test for this in customers or infer it using profile character­istics and response behaviour and then develop different creative treatments accordingly. Companies that use polls can potentially use this to infer style preferences.

2. Sense, Respond, Adjust – delivering relevant e-communications through monitoring customer behaviour

To be able to identify customers in the categories of value, growth, responsiveness or defec­tion risk we need to characterize them using information about them which indicates their purchase and campaign-response behaviour. This is because the past and current actual behaviour is often the best predictor of future behaviour. We can then seek to influence this future behaviour.

Digital marketing enables marketers to create a cycle of:

• Monitoring customer actions or behaviours and then …
• Reacting with appropriate messages and offers to encourage desired behaviours
• Monitoring response to these messages and continuing with additional communications and monitoring.

Or, if you prefer, simply:

Sense → Respond → Adjust

The sensing is done through using technology to monitor visits to particular content on a web site or clicking on particular links in an e-mail. Purchase history can also be monitored, but since purchase information is often stored in a legacy sales system it is important to inte­grate this with systems used for communicating with customers. The response can be done through messages on-site, or in e-mail and then adjustment occurs through further sensing and responding.

This ‘Sense and Respond’ technique has traditionally been completed by catalogue retailers such as Argos or Littlewoods Index using a technique known as ‘RFM analysis’. This technique tends to be little known outside retail circles, but e-CRM gives great potential to applying it in a range of techniques since we can use it not only to analyse purchase history, but also visit or log-in frequency to a site or online service and response rates to e-mail communications.

2.1. Recency, Frequency, Monetary value (RFM) analysis

RFM is sometimes known as ‘FRAC’, which stands for: Frequency, Recency, Amount, (obvi­ously equivalent to monetary value), Category (types of product purchased – not included within RFM). We will now give an overview of how RFM approaches can be applied, with special reference to online marketing. We will also look at the related concepts of latency and hurdle rates.

Recency

This is the Recency of customer action, e.g. purchase, site visit, account access, e-mail response, e.g. 3 months ago. Novo (2004) stresses the importance of recency when he says:

Recency, or the number of days that have gone by since a customer completed an action (purchase, log-in, download, etc.) is the most powerful predictor of the customer repeating an action … Recency is why you receive another catalogue from the company shortly after you make your first purchase from them.

Online applications of analysis of recency include: monitoring through time to identify vul­nerable customers and scoring customers to preferentially target more responsive customers for cost savings.

Frequency

Frequency is the number of times an action is completed in the period of a customer action, e.g. purchase, visit, e-mail response, e.g. five purchases per year, five visits per month, five log-ins per week, five e-mail opens per month, five e-mail clicks per year. Online applications of this analysis include combining with Recency for RF targeting.

Monetary value

The Monetary value of purchase(s) can be measured in different ways, e.g. average order value of £50, total annual purchase value of £5,000. Generally, customers with higher monetary values tend to have a higher loyalty and potential future value since they have purchased more items historically. One example application would be to exclude these customers from special promotions if their RF scores suggested they were actively purchasing. Frequency is often a proxy for monetary value per year since the more products purchased, the higher the overall monetary value. It is possible then to simplify analysis by just using Recency and Frequency. Monetary value can also skew the analysis for high-value initial purchases.

Latency

Latency is related to Frequency, being the average time between customer events in the cus­tomer lifecycle. Examples include the average time between web-site visits, second and third purchase and e-mail clickthroughs. Online applications of latency include putting in place triggers that alert companies to customer behaviour outside the norm, for example increased interest or disinterest, then managing this behaviour using e-communications or traditional communications. For example, a B2B or B2C organization with a long interval between purchases would find that if the average latency increased for a particular customer, then they may be investigating an additional purchase (their recency and frequency would likely increase also). E-mails, phone calls or direct mail could then be used to target this person with relevant offers according to what they were searching for.

Hurdle rate

According to Novo (2004), hurdle rate refers to the percentage of customers in a group (such as in a segment or on a list) who have completed an action. It is a useful concept, although the terminology doesn’t really describe its application. Its value is that it can be used to compare the engagement of different groups or to set targets to increase engagement with online channels as the examples below show:

• 20% of customers have visited in the past 6 months
• 5% of customers have made three or more purchases in the year
• 60% of registrants have logged on to the system in the year
• 30% have clicked through on e-mail in the year.

Grouping customers into different RFM categories

In the examples above, each division for Recency, Frequency and Monetary value is placed in an arbitrary position to place a roughly equal number of customers in each group. This approach is also useful since the marketer can set thresholds of value relevant to their under­standing of their customers.

RFM analysis involves two techniques for grouping customers:

1. Statistical RFM analysis. This involves placing an equal number of customers in each RFM category using quintiles of 20% (10 deciles can also be used for larger databases) as shown in Figure 9.21. The figure also shows one application of RFM with a view to using communi­cations channels more effectively. Lower-cost e-communications can be used to communicate with customers who use online services more frequently since they prefer these channels and more expensive communications can be used for customers who seem to prefer traditional channels. This process is sometimes known as ‘right-channelling’ or ‘right-touching’.
2. Arbitrary divisions of customer database. This approach is also useful since the marketer can set thresholds of value relevant to their understanding of their customers.

For example, RFM analysis can be applied for targeting using e-mail according to how a customer interacts with an e-commerce site. Values could be assigned to each customer as follows:

Recency:

• – Over 12 months
• – Within last 12 months
• – Within last 6 months
• – Within last 3 months
• – Within last 1 month

Frequency:

• – More than once every 6 months
• – Every 6 months
• – Every 3 months
• – Every 2 months
• – Monthly

Monetary value:

• – Less than £10 2-£10-50
• – £50-£100
• – £100-£200
• – More than £200

Simplified versions of this analysis can be created to make it more manageable, for example a theatre group uses these nine categories for its direct marketing:

Oncers (attended theatre once)

• Recent oncers           attended <12 months
• Rusty oncers             attended >12, <36 months
• Very rusty oncers      attended 36+ months

Twicers:

• Recent twicer          attended < 12 months
• Rusty twicer            attended >12, < 36 months
• Very rusty twicer     attended in 36+ months

2+ subscribers:

• Current subscribers         booked 2+ events in current season
• Recent                             booked 2+ last season
• Very rusty                         booked 2+ more than a season ago

Product recommendations and propensity modelling

Propensity modelling is one name given to the approach of evaluating customer character­istics and behaviour, in particular previous products or services purchased, and then making recommendations for the next suitable product. However, it is best known as ‘recommend­ing the “next best product” to existing customers’.

A related acquisition approach is to target potential customers with similar characteristics through renting direct mail or e-mail lists or advertising online in similar locations.

The following recommendations are based on those in van Duyne et al. (2002).

• Create automatic product relationships (i.e. next best product). A low-tech approach to this is for each product, to group together products, previously purchased together. Then for each product, rank product by number of times purchased together to find relationships.
• Cordon off and minimize the ‘real estate’ devoted to related products. An area of screen should be reserved for ‘next-best product prompts’ for up-selling and cross-selling. However, if these can be made part of the current product they may be more effective.
• Use familiar ‘trigger words’. This is familiar from using other sites such as Amazon. Such phrases include:

‘Relatedproducts’, ‘Yourrecommendations’, ‘Similar’, ‘Customers who bought…’, ‘Top 3 related products’.

• Editorialize about related products,e. within copy about a product.
• Allow quick purchase of related products.
• Sell related product during checkout. And also on post-transaction pages, i.e. after one item has been added to basket or purchased.

Note that techniques do not necessarily require an expensive recommendations engine except for very large sites.

Source: Dave Chaffey (2010), E-Business and E-Commerce Management: Strategy, Implementation and Practice, Prentice Hall (4th Edition).