Increase your profits with predictive analytics and statistical customer profiling and segmentation for intelligent customer targeting
You have a nagging feeling you could be doing more with your customer base to increase sales and, more importantly, profits. You may consider cutting margins to improve sales. But how long is this sustainable for? What alternatives are there?
This is where customer analysis, profiling and segmentation come into their own. Customer analysis is all about taking advantage of an often under-valued resource – the data held in your company systems. Re-evaluating your customers based on their value, trading history and sales potential, and then dividing them into groups for different sales and marketing treatment, will improve customer targeting and result in dramatically improved profits.
You’ll be able to stop ploughing money into visiting and communicating with unprofitable customers and unlikely prospects. Customer analysis and customer insights will also enable you to identify and target new prospects from within your internal customer database, and to source external lists matching the profile of your more profitable and higher potential customers to gain a higher response.
Some of the analysis techniques we use include CHAID and decision tree to improve campaign targeting through response modelling; gravity modelling to identify the likely business gain from specific retail outlets or media location sites and their catchment areas; cluster and factor analysis to determine customer profiles and create more granular segmentation; basket analysis to identify cross-sell opportunities; behavioural pattern analysis to identify who to contact, when and why.
We give you the answers to these key customer targeting questions:
Define market and customer development strategies
- What is the size of my market? What is my share?
- Who are our most profitable customers? How do they behave? What do they look like? How valuable are they to us?
- What are the best types of customer to target? Do we want premium buyers or those looking for value?
- What products should we focus on? Which offers will be most appealing?
- Which channels to market should we use – field sales, agents, distributors, outlets, telesales, e-tailers, direct mail/e-mail? Which will give us the highest return on investment?
- What is the lifetime value of each of my customers? How do my actions affect this lifetime value?
- How can I improve loyalty and retain more customers?
- What offer will reactive dormant customers?
- Which customers might be thinking of defecting? How can I identify then re-engage them?
- How effective are my communications? Which media should I use? How often?
Customer segmentation and profiling
Your sales and marketing database would usually contain records of all your customers together with all of the transactions they have made with your organisation over time. This highly valuable resource would typically be enhanced with marketing response, sales activity, survey, suppression and prospect data to provide you with the deepest customer understanding possible. However, left incomplete your customer vision will be blurred! So data enhancement is an obvious precursor.
The saying that ‘the whole is greater than the sum of the parts’ is never truer than in data-driven sales and marketing, as when viewed in its entirety, this information will help to maximise the potential of your business and uncover otherwise hidden sales opportunities that will drive increased profitability.
However, before embarking on segmentation and profiling there needs to be a sales and marketing benefit and hence it is sensible to first define a customer development road map. This could be as straightforward as…
- Recruit (acquire / win) ‘X’ new customers within ‘Y’ months with an average order value of ‘Z’ for ‘ABC’ products or services
- Improve retention of high-value customers by ‘Y’ % over the next ‘Z’ months
- Increase average order value by ‘Z’ %, margin by ‘Y’ % and average order frequency by ‘X’ % through cross or up selling
- Reactivate ‘X’ inactive / lapsed high-value customers at a rate of ‘Z’ per month
- Launch new product ‘A’ to customers that have previously purchased ‘B’ or a combination of ‘C’, ‘D’ and ‘E’ or who have a profile of ‘F’
… and then develop for each of these an appropriate segmentation scheme coupled with offers that you can establish and test by studying past purchase and behaviour patterns.
Understanding the profiles and behaviour of your customers
Tech4T examine your customers by looking at the information you hold about each individual. We can also profile at a household or business level by rolling up the data from the individuals within each. Variables we typically statically model can include lifetime value, spend patterns, product purchase patterns and behaviour, together with attributes. Personal attributes could include: age, gender, lifestyle, leisure interests, occupation, income, accommodation type, geo-demographics… Company attributes could include: number of employees, turnover, type of business, job position, job responsibility…
These variables need to be understood both individually and in terms of how they interact – affect and impact – upon one another. We also match external data to further increase your customer understanding, validate the information you hold and add additional dimensions to aid your targeting and help you visualise who buys what, when, why, how and how often.
Once we understand your customer profiles we use this and other database information to statistically segment individuals sharing similar characteristics. There are two main approaches to segmentation. The first uses techniques such as cluster analysis to create natural groups by measuring how closely attributes correlate. You can then examine the attributes for each group and describe them to paint a picture of your customer segment – average age 50, minimum age 45, maximum age 55, male, spend average £500 per year, etc.
The second approach we take to segmentation uses predictive modelling (such as regression analysis or neural networks) or response modelling (such as decision tree analysis including CHAID). These techniques identify which of your customers are most likely to respond to your offers, which generate higher levels of sales, who are more likely to result in bad debt, etc. The different types of customer can be related to recency, frequency and value of purchases, products or service purchased, payment method, offer, etc., plus their profile data. Clearly the actual variables used will be specific to your industry sector.
Usually modelling results in scores being derived – likelihood to buy, likelihood to respond, likelihood to lapse, likelihood to be your champion, etc. – and these stored against each customer record on your database. The scores are then used to group and rank individuals within segments and can also include other predictive factors such as future profitability, in total or by product. An additional segmentation approach is to use Recency, Frequency and Value patterns.
Segmentation, predictive and response modelling will enable you to make informed sales planning and marketing decisions to get more from your field resource and promotional spend and help prevent waste – in terms of both money and time – by not contacting people who are unlikely to respond to your offer. And you will be able to invest appropriately in sales activities and marketing to the individuals who are more likely to make repeat purchases.
Customer Lifetime Value Analysis (CLV or LTV)
Customer lifetime value analysis is extremely important when deciding how much sales effort and marketing spend to apply to a customer in order to maximise their profitability. CLV or LTV as its often known by will help you define improved segmentation and plan more cost-effective sales and marketing strategies.
There appears to be a consensus that computing lifetime value is a straightforward process and people often cannot understand why they can never find a simple solution. Indeed, we are often asked to supply software with a ‘button’ to automatically show lifetime value and there is disappointment that this is not possible until we explain why.
The CLV process requires significant data integration, transactional analysis and testing but is something Tech4T can undertake for you in a very cost-effective way. Alternately, software such as FastStats can be used to calculate the lifetime values once the calculations have been defined.
The way customer lifetime value is computed usually follows two approaches:
- Customer Lifetime Value (CLV) – delivers profitability of customers to date and is the sum of the gross profit from all historic purchases for each individual customer. It’s always the ‘net present value’ rather than the value over the expected lifetime
- Predicted Customer lifetime Value (PCLV) – uses predictive analysis of prior transaction history combined with various behavioural indicators to forecast the lifetime value of each individual customer (or donor, member, business, etc.). This predicted value will become more accurate with every subsequent purchase and interaction.
Customer Lifetime Value (CLV)
There are two different approaches as to how this is calculated – one is based on gross margin to the date the calculation is run.
For example, if the gross margin is the same across all your sales transactions, then a formula could look like this:
Customer Lifetime Value = (Sales transaction one + sales transaction two… + all other sales transactions to date) * the average gross margin. If every transaction had a different gross margin, then this would need to be pre-computed for each transaction prior to being aggregated.
The second approach is more complex and delivers a value closer to what the customer is really worth. It works on net profit and takes account of all costs associated with that customer. So in addition to the calculation above, we would include for example the cost-to-serve by the field force (visit cost * number of visits per annum) and/or by a customer service or support team, the cost of handling any cancellations, returns or refunds, the cost to acquire the customer in the first instance and a proportion of the marketing, data, sales and marketing support cost including any investment in specialist analysis tools.
Predicted Customer Lifetime Value (PCLV)
Where we need to predict the total customer worth over their ‘lifetime’, we use predictive algorithms. Here the PCLV would take account of the probability of both the net present income and the sum of all likely future revenues from a customer, minus all costs associated with that customer.
There are two approaches:
*Simple – Gross margin over average lifetime
CLV = ((AMT * AOV) * AGM) * ALT
AMT = Average monthly transactions
AOV = Average order value
AGM =Average gross margin
ALT = Average customer lifetime (in months). This number would come from separate analysis
Here we use the output from the *simple approach above as an input to a more complex process that would typically include the monthly retention rate and any monthly discounts, plus could include estimates for ongoing net profit contribution. We would tailor our computation to your specific industry sector as this will greatly improve accuracy. One important factor is that we would include the likely ongoing costs to your business for retaining each particular customer.
As we all know not every individual is the same, some are a lot more valuable than others and the more data gathered about an individual enables you to determine, with increasing accuracy, what sort of individual (or business) they are likely to be – High, Medium, Low value etc. How valuable customers really are will become a key driver to field force planning and marketing to ensure cost-to-serve remains relative to profitability and retention.
Sales Targeting Insights – a unique pattern detection process that identifies customers in need of attention to steer field sales and supporting marketing activity planning
Once we have profiled and segmented your customers, the next step is to look for how things change over time. This is more complex and requires a thorough understanding of statistical data analysis and often extensive computer processing in order to determine the dynamics of each of your customers and the segments they belong to – and yes an individual (or business) can belong to more than one segment!
We interpret your customers’ trading activities, seasonal trends, purchasing behaviour, etc., to identify new sales opportunities and customers at risk of defection. We then create scores for each customer (or prospect) based on their current trading activity or their likelihood to take a particular course of action – buy or donate for example.
By comparing scores that are recalculated on a regular basis over time, we determine the significance in any change and then suggest the next action to take. This will alert sales and marketing staff as to who to contact, why and when. A clear benefit here is that if a customer who has been regularly spending £x per month suddenly reduces the amount or gradually slows their trading with you, this could indicate a likelihood that they will lapse. Knowing this before they do, enables you to contact them, understand the cause, calculate the effect and take appropriate action.
We can direct the sales and marketing teams to…
- Accounts that are exhibiting signs of churn – likely to defect to competitors, and the impact of such loss
- Who is more likely to purchase what and in what priority should they be contacted
- Which accounts exhibit significant change across their trading patterns compared to what could be expected
- Key strategic recommendations by customer for retention, reactivation, cross and up-selling
- A breakdown of how customers have moved across segments and the financial impact to your business
- Geographic breakdown showing the locations of customers needing ‘attention’
- Contact action files for each member of your sales team
- If Territory Runner is deployed, we can update this system to reflect the suggested contact strategy
We can also reclassify your customer segmentation system to reflect the changes in trading we have identified and adjust their predicted sales potential.
This is a complex process which will deliver significant help with sales and marketing activity planning and increase sales.
Turning customer analysis into sales and marketing action
Whether you have a specific marketing or sales objective to achieve, or want to gain a deeper holistic view of your customers to improve your target marketing and sales planning, Tech4T have the specialist data integration, statistical and predictive modelling, sales forecasting and spatial analysis skills and technologies (including FastStats, SPSS) needed.
We first give you a consolidated and duplicate-free view of your customers – often called a single customer view. We then provide the insights you need to retain and grow existing customers, and understand where best to apply resources to expand your market share and maximise customer sales potential.
Our work is delivered in a format suitable for loading into your own systems or loaded into our interactive geo-targeting Territory Runner system. We can also deliver standalone spreadsheets and .pdf maps.
Tech4T’s true skill, the know-how to translate business questions into a practical, actionable solution that delivers answers, fast, will add real value to your business.
Contact us to see how we can help you exploit your data and use customer analysis to generate effective and actionable customer insights to drive growth in your organisation.