Accelerate data transformation
Reverse costs
Drive revenue growth

This paper will explore how data analytics can quickly transform decision making and deliver cost savings that maintain performance and optimise growth, when time and money are in short supply.

No business has been immune to Covid-19 and one of the greatest challenges, as the threat levels decrease and life starts to return to some sense of normality, is how to choose the right strategy. Informed decision making is key, especially when both budgets and time are in short supply. Against a backdrop of unprecedented disruption, the most critical next goal is to survive and then getting to a new kind of 'business as usual'.


A critical challenge for businesses will be how to successfully exit the Covid-19 pandemic. We are already seeing with clients how a key enabler of success has been using the insight provided from data analytics, to maintain performance
whilst taking cost out of the business at pace. 


It is no secret that data-driven decision making can transform the way a business performs. We have reduced operational costs by over £108m across our client base, by applying the smart use of analytics and putting the data to work. Now is
the time for business to truly leverage data analytics to move quickly from survival mode to some sort of 'business as usual'.

In this white paper we present our experience with a number of strategies, which we know can be implemented at speed without costing the earth that put data to work and make the most out of restricted budgets and resources.  We touch on some of the key principles to successfully use data at pace and consider a number of areas where we believe businesses today should tackle first, to both mitigate the threat of Covid as well as making the most of the opportunities it presents.

Conversion Rate Optimisation

Diminished marketing budgets must now work really hard to drive awareness and brand visibility after the extended lockdown and associated disruption. Therefore, it is critical that websites are optimised to take customers from browsing to purchase as seamlessly as possible.

 

Data analytics can quickly identify bottlenecks in the online customer journey and identify where potential customers are not being converted. Identifying the customer journeys (through products and pages) falling below the website average for conversion will prioritise immediate fixes and the development of new features on the website to improve conversion. 

 

Using the same analytics, triggers and events for the customer journey can be tested to improve the conversion rate. The same predictive response approach to following up abandoned baskets allows prioritisation of those most likely to be influenced to convert with outbound activity.

Using online conversion optimisation, we worked with a leading retail business and identified £30m of incremental revenue based on inefficiencies on the website. 

Featured Case Study

Journey analytics mapping for an online retailer

+ £1.8m incremental sales from light browsers

Detailed journey analytics mapped the movement of users through the website and investigated drivers behind key areas of high traffic or high drop-out. This uncovered a number of high potential ranges that were being heavily browsed but with very light promotional activity. Changes to the promotional focus of these areas towards the experience, occasion and needs of customers were implemented and tested.

The Optimisation Journey

Whilst the process of improving conversions requires thorough planning it can be undertaken in manageable chunks and still deliver the benefits. In order to maximise the potential impact we recommend carrying out as detailed analysis of the omni-channel behaviour of your customers that you can to start with.


This needs to build as complete a picture as possible to fully understand the end to end journey to conversion starting with: how did the customer arrive at the site?; what did they do first/search for?; what route to the product did they take?; what was in their basket if and when they abandoned; through to: how, what and when did they finally make a purchase?

01/

Discover

Analysis of the customer journey from lead sources, web visitor behaviours and transactions, to understand customer needs and pin-point where drop off is happening.

02/

Action

Identification and sizing of opportunities.  Translation of findings into actions across product ranging, pricing and website design.

03/

Grow

Continuous measurement of conversions through the funnel and A/B testing of  changes across the journey.

This discovery analysis should sit at the heart of your brief, test hypotheses and planned programme of optimisations. It will enable you to identify and size where the biggest opportunities lie and what the drivers to increase conversion might be and crucially provide the data points to test and learn possible solutions. 


It will also deliver the understanding of What, Why, How and When of what is happening across the customer journey which will drive the actions that will deliver the best results.
 

  • What? - The customer behaviour that has been captured will provide the data to tailor the message and experience for that customer’s needs. 

  • Why? - Identifying the drivers of abandoned visits and transactions informs improvements on product, price and site experience.

  • How? - Understanding the type of customer will define how to communicate and interact with them: e.g. call, chatbot, email, text.

  • When? - Knowing the length of and stage of different journeys will indicate the optimum time and trigger point to target/re-target a lead. 

It is then down to making changes to the customer experience, be it site features or navigation, product placement, pricing or even promotional message and testing these to see which work.

Delivering Results

Test Result - Messaging
Customers who engaged with an ‘abandoned basket’ email had a higher propensity to convert.

Tests for different content and format were put in place to improve click through rate %.
 


+5% Click Through Rate
Test Result - Upsell
Customers diverted to online self-serve travel booking management solution to reduce pressure on call centre (7% decrease in calls).

Upsell options, such as excursions,
targeted during the post booking management process through the platform.

+6.5% Incremental Up-sales
Test Result - Experience
Online speciality retailer was
seeing a large number of heavily browsed ranges where there was very light promotional activity.

Re-focus of promotions around experience, occasion and needs of customers, was undertaken.


+1.8m Incremental Sales
driving revenue growth unsplash.jpg

Predictive Response Modelling

With a large proportion of the employees still on furlough or slowly returning to work, it is likely that there may be significant resource constraints for marketing teams to target all customers through direct marketing or call centre activities.

 

Predictive response modelling enables the business to understand the most likely response from customers and which of these to prioritise based on their commercial potential. Using predictive response modelling, a leisure business with 200+ clubs around the UK reduced its mailing quantity by 50% resulting in £2.3m in annual cost savings with no impact to revenue.

 

Integrating these results with customer segments can help teams sort through the customer base and figure out the best groups to pursue. This means they can then focus on devising more tailored strategies for these customers and where outbound contact is needed, selecting manageable volumes to go after.


Expect a number of benefits from good response modelling aside from decreasing marketing costs; increased response rates; more effective acquisition; and the knock on benefits of increased revenue through identifying and attracting the right people.

This can be achieved through the following five steps:

Predictive Modelling for your CRM

  1. Variable creation:  Construct variables from historical order, transaction and campaign data of both responders and non-responders that describe their previous behaviours.  These may include looking at days since last purchase, number of times mailed, number of past purchases etc.
     

  2. Modelling: Using these variables begin the predictive modelling process by defining what a 'response' means for your business or campaign and use this to assign a score for a customer’s likelihood to respond positively to your latest campaigns and offers.
     

  3. Prioritisation: With these scores then create a list that ranks customers or prospects based on who is most likely to buy, subscribe, or respond to an offer.
     

  4. Action: Use this list to prioritise customer volumes where costs or resource constraints are an issue such as: who to send expensive catalogues; how to get the most out of a restricted mailing or advertising budget; which sales calls have the best chance of closing for your agents.
     

  5. Further optimisation:  Use the learnings to further optimise your efforts above by using the models to predict which customers are most likely to buy anyway without being marketed to.

Featured Case Study

Loyalty scheme for a pharmaceuticals brand

Pharmacy
+13-14% increased offer redemption rate

A leading health and beauty brand, wanted to revitalise their loyalty scheme and drive up response rates to targeted offers and upsell promotions. Targeted campaigns were developed in conjunction with predictive response models to deliver the most relevant product offer campaign to each customer.

Considerations

AI and machine learning models play an increasingly significant role in the business decisions of companies today.  As the use of these techniques grow in business it is important to consider some key issues around governance of data and AI. In particular with predictive modelling be cognisant of two important issues:


1. How do you ensure models are delivering ethical and explainable outcomes? Historical data can contribute to the creation of unintentional bias that unwittingly leads to your business treating people differently or unfairly. Read more of our insights on bias in machine learning here. 


2. The build and maintenance process of the AI models themselves is an additional important consideration. Consider the safety net you have in place to monitor the performance and degradation over time of your models. 

Customer Churn Modelling

It’s generally accepted that it costs more to acquire a customer than to retain one. Managing churn appropriately is a huge
growth driver for any business. Using churn modelling it is possible to identify the key drivers and behaviour changes that are pre-cursors to a customer leaving. 


Models can be created that continuously learn using historical customer data to identify the triggers and behaviour changes that indicate potential churn. This can spot and score the risk of potential churn far more early than traditional recency and value models and this can have a big impacts on costs.


The process requires an initial piece of discovery analysis to properly identify how churn is happening, in which customer
groups and what stage of the customer lifecycle. Ideally all of this can happen using data, but sometimes it is not straight forward to integrate customer events across marketing, sales and customer services. If this is the case we recommend quickly probing for potential issues from the teams at the customer coal face and then validating them in the data.

Image by Alexander Kovacs

Once it is clear where the problems lie the data scientists can begin putting data to work to identify the features of interest and figure out the right type of models to build that will score individual customers based on their propensity to churn. They then test and validate the model with training data set before deploying and testing the models in real life.

 

In practical terms the outputs can be quite straightforward. This could be a customer list with scores out of 100 that indicate likelihood to churn alongside a series of key behavioural metrics that will help inform the reason why. The rest is then up to the business to put the scores to use in marketing and customer service responses.

Featured Case Study

Reducing churn in online retail

+£9.6m increased revenue through reduced churn

This multi-channel retailer knew that many online customers were not returning and was looking for solutions to proactively identify those most likely to churn.  We ran over 40 different models (varying customer features, historical data, segment and
predictive model used) to identify the models that best
explain the propensity to churn. Changes to targeted product placement, pricing and delivery offers were implemented across the website.

Using churn modelling we supported a specialist credit card business (with 50% of market), to reduce its churn by 80% through a tailored churn communication strategy based on data analytics.  In all business cases, the dynamic model enables the business to focus on the highest need cases first;
 

  • By spotting potential churn early the business can react early with responses that demonstrate great customer care and 
    costs less than the final rescue attempts used when churn is spotted too late.
     

  • The learnings from these models can feed directly into redesigning the customer experience, emphasising the moments that matter and stripping out costly initiatives that do not add any long-term customer value.
     

  • Freeing up capacity in call centres to deal with disgruntled customers or where time and resource are wasted on customers who are going to leave anyway. 
     

  • Efficiently monitoring and predicting churn not only reduce cost, but to identify opportunities to broaden sales and upsell new products and services.

To understand how to implement data analytics solutions in your business, contact our team of strategy experts.  You may also be interested in reading more of our experts insights.