Monday, September 17, 2012

Database Scoring using RFV simplified


Database scoring involves selecting the most important customers of your total customer list.
Typically, database scoring is done so that a limited marketing budget can deliver maximum benefits.
A very common technique is using the RFV to score databases.
This is a short story to show how it works:

There was once a very successful businessman who owned a number of sari shops.  He had 3 sons called, Master R, Master F and Master V.  Very bright youngsters too!
One day, the sons approached the father asking to be taught the business of sari shops.  The business man was wise and knew that the only way to learn was by doing.  And that the most important lesson to be learned was customer relationships.
So he asked his sons to come the next morning to one particular store.  There he gave each of them a present.  And he asked them to give it to any one customer from this shop who they felt was most deserving to be given a present.  He gave them a month to do the job and said that they would all meet again in one month in that very store at the same time.

1)    Master R was in a hurry.  Why do tomorrow what you can do today?  He saw a lady paying for a sari at counter and immediately gave her a gift. “Thank you ma’am,” he said.  “We’re so happy you visited us today – please come again soon.  Here is a small gift to keep us in your thoughts and come back soon.”

2)    Master F was more patient.  He had spent some time at this same store before and knew of a college teacher who visited every Saturday.  Every Saturday, she bought a sari and gave it to her best student of the week (she taught at a girl’s college) the following Monday.  The saris she bought were not expensive since they were for students and the college teacher wasn’t particularly rich.  But the customer was a frequent shopper so Master F waited for Saturday and gave her the gift.  “Thanks for being our regular customer.  Please do keep coming.  Here’s a little something for you”.

3)    Master V was a little more like his father.  He waited for an entire month.  Then he pulled up the details of the customer who had the highest bill.  He realized that she had bought 4 saris at one go!  He drove to her house – a mansion quite a distance away from the shop.  Upon entering the mansion he learned that she was a rich industrialist’s wife with three teen-age daughters.  She had bought them all saris for Durga Puja.  So he gave her gift, thanking her for the business.

The three sons met their father at the end of the month.  He asked about their gifting experiences.
Master R, with his usual impatience jumped in.  “I realized that the sooner I gave the present to a recent customer, the higher the chances of that person coming again.  So I gave the gift to the first customer I saw buying a sari, hoping she could come again soon out of gratitude.  In fact, she did, only yesterday!  So I believe I’ve invested the present most profitably.”

Master F laughed at R’s simplicity – he explained how he waited for a week for the teacher who came in regularly and gave her the gift to keep her loyalty and frequency of visits.  And in fact, she did come again on all the following Saturdays too!

Master V found this all absurd.  He explained how he had identified the most valuable customer of the entire month and had given her the gift only yesterday.  If she comes again, she would indeed be profitable.  He showed every one the value of her bill to prove his point.

All 3 began arguing.  F and V felt that R was being impulsive.  Why the rush?  R argued that time is money.  And old customer does not even think of us as much as a recent customer does.  So the sooner he got the gift to work for him, the better. 
R and V felt that F was wasting money.  The teacher would have come anyway.  Why bother with a gift?  F argued that keeping a loyal customer was far more cost efficient than trying to create a loyal customer.
R and F scoffed at V.  “You really think that woman will want more saris after buying so many the last time?  She also lives so far away!” they said.  V argued back, “She likes our stuff.  She has spent so much with us.  One more visit and she’ll give us far more value than either of your customers.”

Their father intervened, “Let me get this straight.  R: You think the most recent customer is the most likely to buy again if you give her a gift.  F: You think the most frequent customer is likely to buy again if you give her a gift.  And V: You think a customer with highest historical value is most likely to buy again if you give her a gift.  What if I say you’re all right?”

The boys look puzzled. 

The father then pulled out a long list in which he had meticulously recorded actions of every one of his customer with their dates of visits and purchase details for the last 3 years.  He then gave the boys 10 presents again and said, “Why don’t you take a day more and try again? Tell me whom you would give these 10 presents to.  We’ll meet tomorrow on this.”

This time, the boys knew they were on to something.  Otherwise their father wouldn’t have parted with this list, which he updated every night and kept locked up in safe.  After some fighting and quibbling, the boys prioritized all customers and agreed on 6, who were:
Very Frequent (came in at least once a week) + Very Recent (all of them had come to one of their sari shops the previous day) + High Value (all of them had purchased over ten saris each in the last one year)
There was no dispute there – these 6 customers needed to be given presents.  But there was some trouble deciding on the last 4 gifts.  They had to decide between:
  • High Value customers (those that had purchased 20 saris last year but not come even once this year? 
  • Highly Frequent customers (those coming every day but buying very cheap, low margin saris)?
  • Very Recent customers (all that came in this morning – hoping that they will come in soon again if they receive the presents)?
The next morning, they went to their father with this dilemma and apologized for not being able to decide on the balance 4 gifts.  “Great!” said the father.  “You’re now ready for database scoring”.  The first six customers basically have full scores on parameters of Recency, Frequency and Value.  The balance, you’ll now give less than perfect scores since not all of them score as well on all three parameters.

“But how should we decide which customer is more important?” asked the boys sheepishly.  “An infrequent customer with high value or a very frequent customer with low value or a very recent, high value 1st time customer?” he continued.

“You test!” said the father.  “You test by giving presents to those coming in once a week, and to those buying 5 saris at one go and to first-time customers.”  See who comes back to buy more.  Then you you’ll know what works for you and which parameter to value more after scoring.”
The boys were mixed with exhilaration and confusion.  They now realized the value scoring the list (database) their father had given them testing to refine their marketing strategy.  They were excited about building their own learning and experiences into scoring.

Learning:  All other things being equal
  1. A Recent customer is more likely to respond to a marketing stimulus than a less recent customer
  2. A more Frequent customer is more likely to respond to a marketing stimulus than a less frequent customer
  3. A customer having given historically higher Value of business is more likely to respond to marketing stimulus than a customer having given less value to the business
Database scoring using RFV involves scoring a customer record on each of these 3 parameters (say classifying a customer as "High", "Medium" or "Low" on each) and then experimenting with various combinations to learn which of these segments offer highest opportunity for ROI on marketing stimulus.

8 comments:

  1. Very very interesting! Typical dilemma of a mktg manager. I would put one caveat though - there will be diff requirements depending on the category you operate from. It will be different for commodities compared to electronics to apparel. For example - in the learning section, point 1 - the behavior of the customer will be different in the case of grocery versus electronics

    ReplyDelete
    Replies
    1. So true - thanks for making that point Vejay. In Grocery or movie rentals, - frequency and recency tend to be very strong predictors while in categories such as electronics, value has a critical role. appreciate the comments.

      Delete
  2. Nicely explained Ashish. The style is simple, engaging and to the point. Saying it again - maybe you should consider writing a book on DM concepts.

    ReplyDelete
  3. Excellent One Ashish! However, as indicated by S Vejay Anand, I was expecting some more insights on the relative importance between R, F and V. Or atleast, how would they differ across categories.

    But no doubt a very thought-provoking piece...conveyed in least jargonised, who moved my cheese kind of way! Good going!

    ReplyDelete
  4. Interesting. Very much engaging till the very last word. Easy to digest, easier to explain ! You ought to pen down such things in a book sirjee. Will be of crucial help to a naive to catch up with crucial buzz words in the market.

    ReplyDelete
  5. Very well written Ashish. Like the story telling technique employed. May be you can do a follow up piece analyzing spend patterns- for e.g whether/what is the correlation between frequency and av. spend, variation in spend among various frequency buckets, etc. Would be really helpful to marketers

    ReplyDelete
  6. Hi, Very good use of the story telling technique to make the concept stick. However for a potential marketer the dilemma would continue as pointed out by some of the other comments as to what should be the relative weights for the three parameters. As you have explained this could vary depending on the category and the class of customer you are serving. In my experience as far as Indian customers go in most categories it may be prudent to assign lesser weights to recency as compared to frequency and value. Look forward to more such examples.

    ReplyDelete