CRM Blog!

  

May 2017 - Rating NG - Not Geeky
 
Recently I’ve been observing how retail businesses are enhancing their in-store experience with digital technology.  I’ve seen gyms really take off in this area, trailing behind fitness apps which record so much data about your activities.  After joining a Pure gym you receive a code which is your key to enter the 24 hour gym – Pure gym does not have a reception by the way, they don’t need to!  I have to admit it does feel a bit like you are on another planet, on the dark side as you enter your code and one side of glass pod opens to let you in and the door on the other side slowly lets you into the gym.  I wonder when the pod will also be taking my weight, and other biometrics!   What I love about the code system is that it is a great data capture tool - Pure Gym know exactly when I go into the gym and how long each workout is.  This “time in gym” data is useful for email engagement and I often get emails helping me to get my mojo back when I haven’t been for a while!  Good to see more companies using recency and frequency data in this way, in a controlled way so its not annoying!
 
I had another great digital experience at a small fitness studio called Tribeca.  They give you a digital band which is also my locker key and I use it to tap onto a device next to each fitness studio which registers me for a class and shows on a digital screen that I have checked into class.  All my payment details are held on the system and the studio operates a cashless system.  When I finish my class I can put my name down for a smoothie and payment is taken automatically from my card.  It’s a great experience and it all works “smoothly”.
 
At both Puregym and Tribeca however, I believe there is one small flaw in the system with regards to security.  There is no way of checking that it really is me in the gym – the system relies on a code being entered or a device being tapped.
 
If you have any other examples of great omni-channel experiences, I would love to hear from you!  There are also some great examples of e-commerce companies starting to have retail and pop up stores – eg Amazon opened a bookshop in Seattle last year.
 

 

November 2016 - Rating NG - Not Geeky
Earlier this year, whilst in New Zealand I went to see Danny Bhoy, the Scottish Indian comedian, performing in Auckland.  Danny is extremely sharp about the business world and I love his simplistic, punchy observations.
 
His theme this year was “Tick the Box”, referring to the complex world of data protection and the various ways companies opt customers in to their marketing communications.  At the time I was working on exactly this with a client so the message was even more poignant.
 
As marketers, we do make things complicated for consumers, and even for ourselves.  As a data marketer my job depends on having more customer data to report, analyse, model and target the most appropriate products and services.  I have proven in my past roles that data driven marketing drives significant incremental revenue for businesses, as well as significant ROI.  The beauty of data marketing as well of course is that it’s easier to prove the benefits than brand marketing.  So with all this in mind, we owe it to our customers to make the “value exchange” (ie data for us and relevant marketing messages/offer for our customers) as simple and compelling as possible.
 
I really love that small businesses like hairdressers are using data more and more to build their databases and market to their customers.   But Danny made the crucial point that we all need to keep it simple, and he hit the nail on the head with this next message. 
 
He gave the example of walking into a hairdresser and being asked to give his email address straight away.  “Why do you want my email address? I just want to get my hair cut” he said.   He was told that the email was needed to set up an alert to remind him when to get his hair cut.  “But I have a mirror” Danny said!!! This is so true and this is where privacy policies  (simply and correctly worded) are really important to explain why data is being collected, and how it will be used.
 
But back to the Danny at the hairdresser - there were 10 questions on the form he had to fill out.  According to Danny though, in his mind there were only really 2 questions that needed to be asked:
1) Do you have hair
2) Do you want it cut
 
Humour aside, the key message is clearly that that we make things too complicated for our customers.  The message from the General Data Protection Regulation (GDPR) is exactly the same - keep it simple enough for customer to understand. 

 

Feb 2015

Why don't more businesses have a special number for customers to call in and leave compliments?  Wouldn't it be a great metric to measure and track and aim to improve?

We constantly track and measure negative metrics (churn, complaints etc).  It would be great to turn the focus and energy round in the business to positive aspects and metrics to aim for.

Just a thought!

 

April 2014

(It's been a while since I last blogged...a symptom of being busy with CRM plans for clients!)

5 Rules for effective CRM - Rating NG - Not Geeky

 

  • Prioritise your business objectives – is it customer retention or revenue growth?  You will need different marketing programs for each.
  • Know your data - what data is available and already captured v what data do you need to start capturing and/or purchase?
  • Know what works and what doesn’t, what products customers like and dislike, which channels get higher response rates, which offers drive higher conversion rates?
  • If you don’t know the above, start testing to optimise your campaign results.
  • Golden Rule of CRM - One size does not fit all.  Build a strategy which optimises products, channels, offers, content etc for the best customer targets.
 
December 2014 
 
#12daysofCRM - Rating - NG - Not Geeky
 
On the first day of Christmas
My CRM Guru gave to me
One perfect set of data
 
On the second day of Christmas
My CRM Guru gave to me
Two triggered offers
And a perfect set of data
 
On the third day of Christmas
My CRM Guru gave to me
Three subject lines
Two triggered offers
And a perfect set of data
 
On the fourth day of Christmas
My CRM Guru gave to me
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the fifth day of Christmas
My CRM Guru gave to me
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the sixth day of Christmas
My CRM Guru gave to me
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the seventh day of Christmas
My CRM Guru gave to me
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the eighth day of Christmas
My CRM Guru gave to me
Eight multi variate tests
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the ninth day of Christmas
My CRM Guru gave to me
Nine data insights
Eight multi variate tests
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the tenth day of Christmas
My CRM Guru gave to me
Ten tips on loyalty
Nine data insights
Eight multi variate tests
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the eleventh day of Christmas
My CRM Guru gave to me
Eleven customers tweeting
Ten tips on loyalty
Nine data insights
Eight multi variate tests
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
On the twelfth day of Christmas
My CRM Guru gave to me
Twelve segment strategies
Eleven customers tweeting
Ten tips on loyalty
Nine data insights
Eight multi variate tests
Seven signs of churning
Six geeks and coding
Five Golden rules
Four calls to action
Three subject lines
Two triggered offers
And a perfect set of data
 
 
 
 
11 November 2014
 
Predictive Modelling by Suba Venkataraman - Rating - VG - Very Geeky!
 
Predictive modelling has many uses in a variety of industry sectors.  Here we’re going to focus on the application of predictive modelling in a marketing environment, with the objective of maximising cumulative value generated by customers whilst minimising marketing spend.
Predictive modelling is often useful in marketing as a classification method – essentially to prioritise which customers to target for different campaigns and/or products – for example retention teams often want to know which customers are most likely to leave (churn) in order to target them with the richest offers, or product teams want to know which customers are most likely to take up a new product.
If we only had a small number of customers, we could work out on a piece of paper who are the best prospects to target for a new product.   Below is a simple example of some customer records – let’s just say this is what the data could look like for a mobile telco:
 
 

Let’s say the objective is to decide which customers to target for marketing a new high value mobile phone tariff offering ‘unlimited’ data, voice, and SMS. 

If this was our entire customer base, we could easily apply some logical rules to decide who to target, based on analysis of previous behaviour.  For example, if we know that customers in the 18-35 age range are most likely to take this type of product, this would form one of the prioritisation rules.  We may also know that customers are most likely to respond to new products if they are within the first year of their tenure – so this would be another rule.  Then we could apply usage based criteria for choosing the ‘best’ customers, e.g. the highest users of the key services – data, voice, and SMS would be the prime targets as they are most likely to reap the benefits of an unlimited allowance. So in the above example, customer no. 9 would be prioritised above others as they fit into the 18-35 age range, have a tenure of only 2 months, and their usage across all services is higher than average.  

However, with thousands or millions of customer records, we need to use statistical techniques in order to classify and prioritise customers in this way.  This would provide a probability or propensity score for each customers likelihood to take the product, and allows us to cherry pick the best prospects for a campaign or product.

 
 

Things to remember when building a predictive model:

1.      Exploratory data analysis (EDA) is essential in order to get the model building off in the right direction.  EDA provides the opportunity to test simple business hypotheses and hunches before launching into rigorous model building.  Never assume anything - always run sense checks on what you think you already know about the data, and investigate patterns and anomalies before you start selecting data, particularly large volumes of it.

 

2.     Understand the business context fully to make the data and model relevant and actionable.  For example, highly transactional businesses (such as retail/FMCG or Pay As You Go mobile), will require data from a short time frame, down to the hour or even minute/second. Whereas businesses built around a longer time frame (such as insurance or mobile phone contracts) will require more historic data that can span months. 

 

3.     Further to the above point, use the right type of model for the problem at hand:

                i.         Classification problems such as the mobile telecoms example above are best dealt with by logistic regression or decision trees.

              ii.         Estimation problems such as predicting the future value of a customer are best dealt with by regression or neural networks.

             iii.         Problems that involve classing customers into groups or segments are best tackled by clustering.

             iv.         Association problems such as understanding purchasing behaviour are best dealt with by market basket or sequence analysis

 

4.     Model selection is key – you may have tried several multivariate models (logistic regression, decision trees etc.) to model customer behaviour and generate customer propensity scores.  The final selection of which model to use then essentially depends on 2 factors:

                i.         Predictive power of models: 

The first factor for model selection is overall predictive power that the model has in comparison to other models. For many classification problems, area under receiver operating curve (ROC) is possibly the best way to assess predictive power of models.

 
 
In the above chart example, ROC is displayed for artificial neural networks, logistic regression, and a CART decision tree. Notice, the perfect model curve (in green) here is with 100% predictive power, and random model (in red) represents prediction through a flip of coins.   Notice that decision trees are often the least powerful, however they are still very useful and popular because of their highly intuitive and easy to explain solutions.  So the final model you select may not necessarily be the most predictive – you need to use business knowledge and intuition to make the final choice.
 
              ii.         Business and operations integration
This is as important, if not more so, than the model itself.  After all, if you can’t operationalise or apply a model, it’s going to be redundant.
a)    Consistent availability of data – only include data which is readily available and will continue to be so for the foreseeable future.
 
b)   The model should be simple to refresh/calibrate to keep up with market/data changes – you don’t want to be re-building the model every so often
 
c)    Ensure it is possible to put into production, i.e. apply the model to your customer database
 
d)   Get business buy-in – ensuring that the stakeholders are supporting and using the model to generate value for your business.

 

End note:  The modern challenge of creating effective predictive models is making them as current or ‘real time’ as possible.  With the trend of instantaneous messaging and communication via social media and other online channels, businesses need to strive to keep up with the constant 2 way flow of information.  Traditional predictive models use data from the past to predict what will happen in the future, however these days, with so much near real time data, the time to predict is becoming shorter, using much more recent data and therefore more predictive of future behaviour.  Forward thinking businesses need to seek out the right tools to give them the power to exploit this data.     

 

3 Nov 2014

CRM and loyalty through global experiences - Rating NG - Not Geeky
 
I recently went to New York from London for a weekend and I enjoyed several global CRM experiences.
 
Mobile technology and global brands make it possible for consumers to enjoy their favourite brands wherever they go. 
Here are some of my experiences of loyalty and global CRM:
 
Amex
I paid for my flight to NY with Amex so I didn’t have to ring them specially and tell them that I was going abroad!  Amex emailed me the day after I arrived in New York, giving me more information about help abroad in case I lost my card. 
At the bottom of the email was the option to connect my card to foursquare to get special offers when checking into places. 
 
This is a great way for customers to get special offers, as there are no voucher codes to enter, nothing to carry and great for Amex to learn more about its customers’ lifestyles.
 
An advancement of this would be for Amex to learn from customers’ data in foursquare and make recommendations directly rather than customers having to search for offers when they go somewhere.  It would be great to receive a push notification when you enter a café, telling you that you have special offer on their coffee if you buy a pastry for example!
 
You could argue that with Foursquare businesses are rewarding offer riders, not the most loyal customers, but if the experience is good, it could attract customers on a regular basis and it may prompt them to spend a bit more than they normally would – incremental volume and incremental revenue!
 
Starbucks
I keep a bit of credit on my Starbucks app and I was able to use its mobile payment for my drink in New York! Pretty cool, I didn’t get any rewards on the international purchase which is a shame, but the mobile payment was pretty slick!  Apparently 90 percent of all mobile payments in 2013 were made at Starbucks!!!
 
Turns out I’m not the only one impressed with Starbucks mobile payments….
 
Starbucks has 8m active members using its loyalty program!  What fantastic data they are now getting, learning about their customers who travel and use Starbucks around the globe.
 
Strava
First of all, let me just say that I love Strava!!!!  Unlike some apps, Strava lets you log your runs/bike rides without a network connection when abroad.  I always take my running shoes with me when I travel and I really enjoyed my autumnal runs in New York! I think I deserve double kudos from Strava for my international runs, but nevertheless was happy for those runs to have counted towards my fitness plans. 
 
Strava, it would be so cool to offer me special offers on US branded sports goods whilst I was there!  You knew where I was, you could have given me a list of stores to go to nearby to pick up new running shoes for example!
 
Virgin
Nothing to shout about here!
When I used to travel to San Fran every few months for EA, I quickly earned a Virgin gold card, but I’ve recently been downgraded to silver.  I’d have loved to still use the lounge, even for a small fee.  That would be a great delightful surprise if they offered that to regular Virgin travellers.
 
The Sunday Times
Isn’t it great how you can read your favourite national papers abroad now with the iPad!  I wish News International offered local NY content to me whilst in New York (they could have given me the option) and local ads would be really cool!
 

 

With mobile technology, every app should be thinking about how to adapt content, offers and segmented targeting f