Documente Academic
Documente Profesional
Documente Cultură
generally defined as the present value of all future profits obtained from a customer over his or her life of relationship with a firm.
It is similar to the discounted cash flow approach used in finance. However, there are two key differences:
1. CLV is typically defined and estimated at an individual customer or segment level. This allows us to differentiate between customers who are more profitable than others rather than simply examining average profitability. 2. Unlike finance, CLV explicitly incorporates the possibility that a customer may defect to competitors in the future.
Churn rate: % of customers who end their relationship with the co. in the given period. One minus Churn rate gives retention rate Discount rate: Cost of capital used to discount future revenue from a customer. Current interest is usually used for discount rate Retention cost: Money spent to retain existing customer. Includes, promotional costs, loyalty incentives, billing, customer support costs, etc.
CALCULATION OF CLV
Acquisition Year
Second Year
Third Year
Customers
Retention rate Orders per year
1,00,000
60% 1.8
60,000
70% 2.5
42,000
80% 3
Rs. 90
Rs. 1,62,00,000 70% Rs. 1,13,40,000 Rs. 55 Rs. 55,00,000 Rs. 1,68,40,000 Rs. 6,40,000 1 -Rs. 6,40,000 -Rs. 6,40,000 -Rs. 6.4
Rs. 95
Rs. 1,42,50,000 65% Rs. 92,62,500 Rs. 20 Rs. 12,00,000 Rs. 1,04,62,500 Rs. 37,87,500 1.16 Rs. 32,65,086 Rs. 26,25,086 Rs. 26
Rs. 100
Rs. 1,26,00,000 65% Rs. 81,90,000 Rs. 20 Rs. 8,40,000 Rs. 90,30,000 Rs. 35,70,000 1.35 Rs. 26,44,444 Rs. 52,69,531 Rs. 53
where
pt = price paid by a consumer at time t, ct = direct cost of servicing the customer at time t, i = discount rate or cost of capital for the firm, rt = probability of customer repeat buying or being alive at time t, AC = acquisition cost and T = time horizon for estimating CLV
A student (Ph.D/Regular/Exec. MBA) on the day of joining purchases a burger for Rs.30 in white house. Calculate his/her CLV.
2.
3.
Probability Models
Econometric Models
4.
5. 6.
Persistence Models
Computer Science Models Diffusion/Growth Models
1. RFM Model (used extensively in Direct Marketing) Recency: When was the last customer interaction Frequency: How frequent was the customer in its interaction with the company Monetary value of the interactions Recency enables the prediction of future value, while frequency and monetary value enable the estimation of the current value. The combination of the 3 dimensions (RFM) allows the combined analysis of current and future customer value Assumptions: Customers who ordered recently are more likely to order again than those who ordered in a less recent period Customers who ordered frequently are more likely to order again than those who ordered less frequently Customers who ordered a higher monetary value (spent more) are more likely to order again than those who ordered a lower monetary value
RFM: Past purchases of consumers are better predictors of their future purchase behavior than demographics. Given the low response rates in various industries (typically 2% or less), RFM model was developed to marketing products/services for specific customers with the objective to improve response rates. In case, where there is no monetary value attached to an interaction like web visit or a complaint, the analysis may be limited to recency & frequency only The higher the RFM score, the more probable it is a for a customer to respond to a marketing program Customers who scored high on RFM (in orders or service usage events) are higher value customers
RFM models create cells or groups of customers based on three variablesRecency, Frequency, and Monetary value of their prior purchases. The model classifies customers into five groups based on each of these three variables. This gives 5 x 5 x 5 or 125 cells. It is also common to use weights for these cells to create scores for each group.
Mailing or other marketing communication programs are then prioritized based on the scores of different RFM groups.
1.25%
1.08%
0.63%
0.26% 5 4 3 2 1
Recency Quintile
0.00%
Frequency Quintile
1.61%
300
200
100
-100
-200
555
455
355
255
111
RFM Cells clearly show who to mail to, and who to drop
Limitations:
1. Recent purchases in R are given higher score.
It ignores the pacing of a customer's interactions/ the time between each purchase/ patterns of buying behavior
2. Monetary-value component is almost always based on revenue rather than profitability. 3. They are scoring models and do not explicitly provide a Rupee number for customer value.
However, RFM are important past purchase variables that should be good predictors of future purchase behavior of customers.
Limitations: 4. These models predict behavior in the next purchase period only.
However, to estimate CLV, we need to estimate customers purchase behavior not only in Period 2 but also in Periods 3, 4, 5, and so on.
5. These models ignore the fact that consumers past behavior may be a result of firms past marketing activities. 6. RFM Deals With Very Small Numbers
Only a small percentage (such as 2- 5%) of customers respond to the typical offer 95% or more will not respond at all RFM tells you which customers are most likely to be in the responsive 5% Those who respond may not be your most profitable customers
Responsive
Customers RFM
LTV
Not all responsive customers are profitable Not all profitable customers will respond when you write them.
2.
Probability Models
Probability models are based on the assumption that consumers behavior varies across the population according to some probability distribution. For the purposes of computing CLV, the models make predictions about whether an individual will still be an active customer in the future and, if so, what his or her purchasing behavior will be. Like most statistical models, it figures the probability that some future event will occur based on statistical patterns observed either theoretically or empirically in the past.
"event-history modeling
Requires only four variables (RFM approach) 1. The frequency: the number of transactions in the past 2. The recency: time units since last purchase 3. The cohort: time units since first purchase 4. The monetary value: the average profit per transaction
CLVi t 0
h
xi ,t mi ,t (1 d ) t
Where
xi,t = number of transactions yielded by customer i in the period t mi,t = profit per transaction yielded by customer i in the period t d = discount rate h = time horizon of the prediction
There are more and less complex ways to use the event-history model to compute the probability that a customer will keep on purchasing. In its simplest form, the formula is tn Where
n is the number of purchases he made in the entire time period and t is the fraction of the period represented by the time between his first purchase and his last one.
Unlike RFM, this approach is particularly good at predicting how quickly a customer's purchasing activity will drop off, as the probability of their being active in the future drops steeply with time, so it clearly has the potential to prevent heavy overinvestment in profitable but disloyal customers. In practice, several other variables are also taken account of variables: demographics, amount of spending, and type of products purchased.
3.
Econometric Models
Many econometric models share the underlying philosophy of the probability models.
Generally these models study customer acquisition, retention, and expansion (cross-selling or margin) and then combine them to estimate CLV
4.
Persistence Models
Like econometric models of CLV, persistence models focus on modeling the behavior of its components, that is, acquisition, retention, and cross-selling.
When sufficiently long-time series are available, it is possible to treat these components as part of a dynamic system.
Models are used to study how a movement in one variable (say, an acquisition campaign or a customer service improvement) impacts other system variables over time. Example: -to study the impact of advertising, discounting, and product quality on customer Equity and to examine differences in CLV resulting from different customer acquisition methods.
The major contribution of persistence modeling is that it projects the long-run behavior of a variable or a group of variables of interest.
Extremely useful as : In the present context, we may model several known marketing influence mechanisms jointly; that is, each variable is treated as potentially endogenous. For example,
a firms acquisition campaign may be successful and bring in new customers (consumer response). That success may prompt the firm to invest in additional campaigns (performance feedback) and possibly finance these campaigns by diverting funds from other parts of its marketing mix (decision rules).
At the same time, the firms competitors, fearful of a decline in market share, may counter with their own acquisition campaigns (competitive reaction).
Depending on the relative strength of these influence mechanisms, a long-run outcome will emerge that may or may not be favorable to the initiating firm. Similar dynamic systems may be developed to study, for example, the long-run impact of improved customer retention on customer acquisition levels, and many other dynamic relationships among the components of customer equity.
Persistence models can quantify the relative importance of the various influence mechanisms in long-term customer equity development, including customer selection, method of acquisition, word of mouth generation, and competitive reaction. However the demands on the data are high, for example, long time series equal-interval observations.
5.
The vast computer science literature in data mining, machine learning, and nonparametric statistics has generated many approaches that emphasize predictive ability.
These include projection-pursuit models; neural network models; decision tree models; spline-based models such as generalized additive models (GAM), multivariate adaptive regression splines (MARS), classification and regression trees (CART); and support vector machines (SVM).
Many of these approaches may be more suitable to the study of customer churn where we typically have a very large number of variables.
6.
Diffusion/Growth Models
Forecasting the acquisition of future customers is typically achieved in two ways: 1. The first approach uses a disaggregate customer data and builds models that predict the probability of acquiring a particular customer: The models discussed so far. 2. An alternative approach is to use aggregate data and use diffusion or growth models to predict the number of customers a firm is likely to acquire in the future. A model for forecasting the number of new customers at time t:
where , , and are the parameters of the customer growth curve. It is also possible to include marketing mix covariates in this model as suggested in the diffusion literature.
Summary
I. CLV is the long-run profitability of an individual customer. This is useful for customer selection, campaign management, customer segmentation, and customer targeting
II. One can generally find four approaches in the literature. RFM Models Probability Models Econometric Models Persistance Models
In Practice A MIX
III. 1. CLV is a forward-looking metric, unlike other traditional measures (that include past contribution to profit) 2. It helps marketers to adopt the right marketing activities today to increase future profitability 3. It is the only metric that incorporates all the elements of revenue, expense, and customer behavior that drive profitability 4. It enforces the focus on the customer (instead of products) as the driver of profitability