Refer a friend programs: are they worth it?

September 6, 2010 By 2 Comments

Refer a friend programs: are they worth it?  Yes, as it happens they are and some very recent research gives us the numbers to back it up.

Referral Programs and Customer Value was recently published by Schmitt, Skiera and Van den Boulte.  In this excellent paper, rigorous analysis was applied to a topic of discussion by marketers the world over: do ‘refer a friend’ programs make money?

Their findings: customers acquired through paid customer referral programs have a higher retention rate and higher initial contribution margin than other customers.  In other words, yes they do work.

In this post from last week “Using social effects to improve customer retention”, we looked at how client referrals can work in reverse.  Losing a client increases attrition for the people already in their social network.  Now we have proof that the positive reverse is also true.

What this means is that the effort invested into so called stimulated word of mouth (WOM) programs generate real and valuable client referrals.

The entire paper (in truth, perhaps not the statistical methods section) is interesting because of the completeness with which the authors address the question of successful referral programs.  For instance they discuss why customers acquired through client referrals are a better match (and higher margin) for the organisation than the average acquired customer:

  1. Reciprocity – the person performing the customer referral feels that they owe (because a payment is involved) the company a good customer.
  2. Triadic balance – where the propensity of two people to feel the same about an object (in this case company) make the friend in “refer a friend” more likely to like the company.
  3. Homophily – people have friends that are similar to themselves.  If I like, or am a good customer, for this company so will be my friend.

Getting back to the source data, the authors were able to access substantial (approximately 10,000) customer records and watch their progress over a 33 month period from acquistion.  This allowed them to review not just the initial margin but also look for differences in retention rate and long term margin rates.

For practical marketers the key outcomes from this paper include:

  1. Higher gross margin levels: customer referrals showed an initial 25% higher contribution margin than non customer referrals.  However, the difference reduced over time to be similar at the 29 month mark.Customer-Referral-Programs-Margin
  2. Better retention rates: customers acquired through referral programs have a higher retention rate and that difference does not reduce.Customer-Referral-Programs-Retention-Rate
  3. Client referrals generate higher value customers: taking (1) and (2) together, customers obtained through successful referral programs are of higher overall value than other customers.  For the authors that difference was 25%, i.e. customer referrals were 25% more valuable than non-customer referrals.
  4. Customer referral programs are a good investment: the net increase in value more than pays for the cost of the reward.  This is of course subject to reasonable reward costs.  In the case of this study the reward cost was 25 Euros, which was more than covered by the 25% increase in value.
  5. Abuse costs less than the increase in value: The costs of progam abuse and other negative side effects of customer referral programs were smaller than the increase in value.

In many practical ways the authors have made our lives easier as marketers by proving the value of customer referral programs.  Some of the debate can now cease and we can focus instead in creating successful referral programs for our companies.

By Adam Ramshaw

Filed Under: Customer Analytics, Customer Loyalty Management Tagged With: , ,

Word of Mouth Programs – Making Money

April 5, 2010 By Leave a Comment

Another thought-provoking piece of research from one of our ‘must read’ organisations, the Marketing Science Institute: Sources of Social Value in Word-of-mouth Programs.

Using computer agent models based on 12 real world communities (several provided by Lithium), Libai, Muller and Peres focussed on the impact Word of Mouth (WOM) has on the value (discounted cash flow) generated in a new product introduction.

Their particular interest was on the WOM generated by ‘seeding’ programs, where organisations give new products to a portion of the customer population in order to invoke WOM. I believe the principles discussed in their paper probably apply to any WOM initiative, not just seeding programs.

The focus on hard dollar returns is notable and useful for marketers considering WOM programs (and certainly plays to our bias for ‘do’ marketing). The authors call these hard dollar impacts changes in ‘social value’ for the customer population.

Two ways that WOM can increase the value of the social network, i.e. make you money (versus new product diffusion with no WOM) were examined;

  1. Acquisition of new customers who would not have bought without the WOM, and
  2. Acceleration of purchases from customers who would have bought anyway, just later. In this case, the added value is the time value of getting money earlier.

They also contrasted WOM programs that randomly select customers with Influentials programs that specifically target customers with large numbers of social links.

The findings, in plain language, were;

  1. Competition amplifies WOM effects. With 1 brand, i.e. a monopoly, WOM increased value by 17% for random customer selection and 27% for influentials. Adding a competitor jumped these numbers to 80% and 107%!
  2. The stronger the brand relative to competition the less value delivered by WOM; stronger brands benefit less.
  3. You can select too many customers: Random customer selection WOM programs including 20% of the target population deliver the maximum social value

    Do not try to recruit all customers in launch WOM programs

  4. Compared to random customer selection, Influentials’ effects decrease faster as you increase the number of customers in the program
  5. Most of the value from WOM can be achieved by random customer selection. Targeting Influentials can increase the return; an additional 33% in these networks
  6. If you target Influentials, acceleration drives the majority of value, not acquisition

This is good news for our start-up customers introducing new products, using social media, in highly competitive markets…

Filed Under: Case Studies and Statisitics, Customer Analytics, Customer Feedback Process Tagged With: , , ,

Net Promoter Score (NPS) and service delivery styles

March 18, 2010 By 7 Comments

Transactional Net Promoter Score (NPS), where you ask customers to indicate their willingness to recommend your processes rather than their overall relationship with your organisation inevitably leads you to review your service delivery mechanisms – and perhaps it even leads you into the exciting world of Customer Experience Management, CEM.

Service delivery strategies are often focussed on the Taylorite-stopwatch, as most organisation assume that for customers, faster and more efficient equals better service experience. While, we sometimes wish this were true, it is not.

I read one of my career ‘epiphany’ books way back in 1995, as a young relationship marketer. Written by Barbara Gutek, The Dynamics of Service introduced me to the idea that there are two distinct types of customer service delivery. Gutek described them in terms of the experience each delivers to the customer; the Relationship and the Encounter. Her follow up book in 2000, Brave New Service Strategy is easier to obtain if you are interested.

Relationship service delivery, is typically associated with ‘Practices’ of physicians, lawyers, accountants, even (ahem) consultants. On the other hand, Encounter service delivery, is associated with ‘fast’ – food, banking, ticketing, on-line quotes. The differences between the two are pretty fundamental for customer experience designers.

These customer service delivery strategies are opposites on many dimensions

 

I suspect that the general move to service encounters through automation and even outsourcing is driven by the difference highlighted on the last row of this table… a distinctly un-customer centric motive.

But as we are talking about measuring customer satisfaction with service, it is important to consider their perspective. For customers, the distinction is really ‘does the service provider know me as an individual or not and vice versa?’ In a McDonald’s encounter you do not really care who hands you the hamburger and, frankly, they don’t care who takes it. Try the same approach as a lawyer or GP and see how your business goes.

So when is it important (for the service experience to be good) that I know the provider and believe they know me?

  • When the transaction is really important to me. We just sold our house, the relationship with our realtor had to be personal
  • When I am inexperienced in the transaction
  • When it appears complicated and the risk of failure or embarrassment is high
  • When the consequences of failure could be dire

And when is knowing the provider not just unnecessary but a frustration in an experience that I just want to be fast and efficient?

  • When the transaction is routine – like withdrawing cash  - we want ATMs not teller queues
  • When I am expert and experienced at the transaction
  • When the perceived risk of failure is low
  • When the consequences are well understood and predictable

But here is the rub. At the extremes, it is easy to see which style of service delivery will be preferred by customers; cash withdrawals versus divorce settlements. But in the very large grey area in between it is not so clearcut.

And we, as service providers, cannot easily tell which will be preferred by only examining the transaction; a great service experience also depends on the characteristics of the individual customer.

An illustration; one of the early telephone stock trading companies found that to be sucessful they had to recognise novice and expert traders (by their trading patterns, not their value), and

  1. Have an experienced broker answer the calls of ‘indecisives’, to reassure and answer questions. This led to high levels of customer satisfaction, loyalty and increasing trading volumes.
  2. Removed the human interaction from experienced day traders who wanted to get on with the transaction, leaving interaction to the IVR or web site, with brokers available on request. Get out of their way was the name of the game. This led to high levels of customer satisfaction, loyalty and increasing trading volumes.

Some customers transitioned from 1 to 2, some never did.

We sometimes forget that a similar evolution occurred in banking when ATMs were introduced, with many customers persisting with tellers until fees drove them, begrudgingly, out of the branch. Banks are now, belatedly and expensively, trying to tempt them back into the branch because selling and buying personal banking products are still human interactions.

Encounters are generally cheaper for the service provider, but they may not be more profitable, if the lifetime value of your customers is the measure you use to gauge success.

Net Promoter, Net Promoter Score and NPS are registered trademarks of Bain & Company, Inc., Satmetrix Systems, Inc., and Fred Reichheld

read
Filed Under: Customer Analytics, Net Promoter Score Tagged With: , , ,

Life Time Value for “Do Leaders”

November 26, 2009 By 2 Comments

We prefer to use the concept of a 3 x 3 ‘Value Map’ to guide our marketing strategy and execution, as we discussed in this previous post on customer value . The concept of ‘captured customer value’ is pretty straightforward for most clients, but the ‘uncaptured value’ can get people to scratching their heads.

So before we move onto the promised discussion of  ’marketing allowable’ for each Value Map cell, I’d like to briefly discuss how to calculate ‘Uncaptured customer value’.

In financial services markets, the example given in the initial post is a good approach. With (typically) contract based products and services, comparing the product holdings (and therefore typical income) of a customer with ‘best customers’ like her, allows you to determine what products she could reasonably buy from you. She either does not have these accounts (for example) or has them with a competitor, but these are uncaptured potential value for you for this customer.

In markets where no contractual relationship exists between you and your customer, it is a little harder to calculate ‘uncaptured value’, ironically because it is harder to determine when a customer has left you – stopped buying your type of product or has switched to buying from your competitor.

In financial services, a customer typically leaves you, ‘attrites’ or ‘churns’, by cancelling the account or service they have with you (leaving the more common ‘silent attrition’ where customers simply stop using your services without explicitly cancelling, to one side for a moment). In retail they buy less over time then disappear.

Sometimes they even come back!

This plays havoc with calculations of uncaptured value unless you set some consistent hurdles / triggers and focus on relative values, trends, consistency and not fall for the illusion of precision. In a world of limited marketing resources, the objective is to spend them on the right customers in the right way to optimise returns. For this purpose, relative measures between groups of like customers are enough to start.

In the spirit of ‘Do Leadership’ we set some informed but arbitrary hurdles that we tune and refine as we learn more about our customers. One of the most important of these hurdles is the behavioural measure that indicates when a customer has left, churned, attrited.

Why is it important to have a measure of churn? Because very few organisations have a long term, data driven view of their customer’s life cycle. That’s why debates about Lifetime value (LTV) usually start with the question ‘but how long is our customer lifetime?’.

If you know the churn rate in a particular group of customers, you can take a snapshot in time (just as a balance sheet does) and say ‘at this moment, these customers have a lifetime of x (weeks/months/years) based on a current churn rate of y.’ If I also know the captured value for this group of customers for this period I can calculate the net present value of their future cash flows.

Then, just as you compare balances sheets over time to see increases in company value, you can do the same with this predicted cash flow to determine if you are making things better of worse, customer value wise. But in the meantime I can rank customer groups by their uncaptured, future value. And populate the Value Map to drive marketing execution.

If I can measure churn.

The formula for calculating the NPV of future cash flow for a segment of customers is not scary;

NPV of customer future cash flows = M(R/(1+i-R))

  • M is the captured value (e.g. margin) from this customer group this period
  • R is the retention rate – the inverse of churn
  • i is the discount rate used by your CFO to price money internally

So intuitively, you can increase the uncaptured value of a group of customers; by selling each one more, getting more of them and by keeping them buying longer.

The challenge to determining that a retail customer has gone is caused by the fact that customers shop at different intervals, they have different ‘rhythms’ to how long they rest between buying. So a large monthly shopper has not left you when she is absent for 2 weeks, but a daily shopper most likely has. Different latencies should be taken into account when calculating whether a customer has gone.

There are 2 ways to approach these individual latency differences when setting hurdles for churn Y/N decisions;

  • Score individual customers on the likelihood they have churned. A simple ‘event-history’ calculation may be a good start. In its simplest form, the scoring formula is t to the power of n, where ‘n’ is the number of purchases made in the period (say 12 months) and ‘t’ is the fraction of the period represented by the time between her first purchase and her last one.

An example (from Reinhartz & Kumar 2002) ; Smith has made 4 purchases, the last in month 8, so n is 4 and t is 8/12 or 0.6667. Smith’s probability of still being active with you is (0.6667) to the 4th or 0.198. There is a 20% probability that Smith will keep on purchasing.

Jones also made her last purchase in month 8 so her t is 0.6667 as well, but as she bought only twice her probability is (0.6667) squared or 0.444, nearly 45%. Jones is more than twice as likely as Smith to remain an active customer.

You can then set an arbitrary hurdle and stick with it; let’s say that all customers with less than 10% probability of continuing with us are deemed to have churned at the point we calculate uncaptured value.

  • The second approach is simpler and takes advantage of the fact that in retail, the Recency, Frequency and Monetary attributes of RFM are not independent. Take the High, Medium, Low groups of customers on the ‘Captured Value’ axis of the Value Map and calculate the average time between purchases for customers in each group. List the average and the standard deviation. The high value group will typically have notably shorter times between purchases (latency) than the other groups.

Then set an arbitrary threshold that says, customers in this group who have not shopped for {average days between purchases plus 1 or 2 standard deviations, (use your judgement)} have churned.

This approach may lead to some apparent anomalies, for example I have a client with an empty High-High cell in their Value Map as high captured value customers churn quickly – but that is important to know!

Value Map

Calculate churn then future cashflow

These examples fit high transaction, non-contractual environments best, but they are effective. What do you think?

Filed Under: Customer Analytics Tagged With: ,