Advanced Analytics Consulting
Customer affinity forecasts
Predictive analytics helps with early recognition and understanding of critical changes in customer and market behavior. Future events can then be forecast much more precisely, allowing for efficient consideration of potential follow-up actions.
Our unique, prize-winning solutions are designed to make predictive analytics as quick and simple as possible for our clients to use. Train, score and validate your models in a self-learning, dynamic and automated closed-loop application that ensures long-term forecasting quality: should a model fall short of your pre-defined threshold values, the application calculates a new and better model on your behalf, without consuming undue computing resources.
Geared toward sustainability, our solutions specifically provide efficient and re-usable predictive analytics processes that can be permanently integrated into business procedures.
Uplift Modeling / Incremental Response Modeling
Identify customers that exhibit a desired behavior upon contact
In campaigns, you need to contact customers in a very targeted way; indeed, they’re often scored on how the recipients reacted to them. For a realistic assessment of the profitability of a campaign, it’s crucial to know how customers might have behaved even if they hadn’t been included – only then can the opportunity costs of a given campaign be quantified. That’s exactly where uplift modeling comes into play: after drawing up two comparable customer groups in a random sampling, just one of them is contacted. With our proprietary and award-winning uplift solution, you can identify characteristics that will greatly enhance your response rates upon contact. This continually saves on advertising budget, as you can see which customers will probably purchase even without contact, and you can further identify customers who might react negatively to your advertisements (that is, the ones with higher propensity to churn).
Identify customers with similar needs and address them specifically
Your customers are unique and you’ve already collected a lot of information in order to understand them. Knowing their basic needs and categorizing them according to those needs helps you control your campaigns: you can be sure that they receive the right offers, at the right time and via the right channels – thereby significantly improving your cross-channel contact and service strategy!
We tap into our extensive cross-industry expertise to help you choose a segmentation approach that works for you and your customers; together we develop measures for approaching your segments in a targeted way that meets their needs. Using ex post simulations, you can discover retrospectively how the volume distribution and needs structure of your customers have developed over the longer term, from which you can extrapolate future changes – and react accordingly.
Customer Value Modeling
Basic prerequisites for increasing customer value
When calculating customer value, it’s important to distinguish between the various analytical approaches:
Customer lifetime value (CLV) is an approach in which one-time customer acquisition and ongoing retention costs are compared to the profit expected over the customer’s entire lifecycle. The basis for this is a detailed calculation of the customer profitability margin.
Customer scoring makes it possible to forecast the turnover or profitability margin of your customers. Using data mining methods, you can forecast relevant performance indicators over a specific time period (e.g. 3, 6, or 12 months) by customer. Our intelligent ‘sliding window’ helps you find the forecast period that works best for you.
Gleaning information from unstructured texts
Using text mining algorithms, information can be extracted and categorized from unstructured text data such as emails to the service center, social media posts, blogs and comments. Sentiment analysis makes it possible to automate classification of large numbers of text documents by mood and tonality and, when appropriate, to forward them via email. Our tried-and-tested text analysis even facilitates the efficient assignment of texts to various business processes, products and/or business units.
Our expertise doesn’t stop at pure analysis and processing of the text file. We also make sure that your newly-won insights can be integrated into your private CRM activities, e.g. by implementing trigger campaigns or through optimizing service processes – for the good of your customers.
Forecasting time series data
Forecasting models can be useful wherever structural predictions of time series (e.g. call volume in the call center) aren’t possible because the requisite data are unavailable, because channel contexts are unknown, or because modeling thereof would be too complex. In such cases, time series analysis methods are required to calculate an accurate prognosis.
With our dynamic forecasting application you can rely on precise predictions to formulate plans to:
•Optimize your minimum required capacities (e.g. call center staffing)
•Process incoming correspondence efficiently from the back office
•Take timely action against the risk of price changes
This gives you a decisive advantage over your competition!
The customer’s view: from first contact through to purchase
When making a purchase decision, every customer goes through a set of phases, coming in contact with a range of touchpoints that can exert various influences on this decision. In order to control your channels efficiently, you need to know to what extent they contribute to successful decisions. Unfortunately, the most common attribution models (First Click, Last Click, etc.) are only marginally helpful for determining and assessing the interplay between touchpoints. Our data-grounded approach considers not just the converted customer journeys but all of them. This makes it possible to recognize the touchpoint combinations that have a disproportional effect on conversions.
Effective allocation of your advertising budget is decisive for achieving your goals. So, it’s important to ask how the budget should be divided up in order to maximize the relevant target values. In our innovative approach, we use methods from non-linear programming to calculate what allocations to your channels can be expected to produce the highest returns.
Big Data Analytics
Analytics solutions for making even voluminous, unstructured data usable
Each and every second, tremendous volumes of data are generated by clickstreams, social networks, purchase transactions, market analyses and sensors; these data must be quickly and intelligently assimilated. But the topic of Big Data Analytics only really becomes pertinent when standard business intelligence tools can no longer efficiently store and process all the relevant, mission-critical data or make them available for marketing, customer management and sales.
New technologies for storage and processing of these data (Hadoop) or in-memory databases for real-time analysis (e.g. Exasol) help you deal efficiently with the special characteristics of big data.
Our Big Data Analytics solution: we meet the new challenges head-on with a unique portfolio of solutions that make existing business intelligence environments ‘big data ready’.
- Provision of innovative warehousing concepts for real-time or batch storage of structured and unstructured data
- Efficient parallel processing of this information (e.g. cleansing, aggregation, enriching) with MapReduce
- Provision, distribution and linking of big data content with existing dispositive data and provision of information via in-memory databases for real-time campaigning and real-time analytics
Credit Limit Optimization
Maximizing customer billings while minimizing risk
More and more, customers prefer to be invoiced for their orders. For the organization, however, this poses substantial risk of loss should the invoices not be paid in a timely manner – or go unpaid. To protect against just this kind of risk, it’s smart to impose a reasonable credit limit starting with the order process.
However, whereas very stringent controlling minimizes the risk of debt collections, it also cuts off a relevant part of the demand. On the other hand, too-generous credit limits that allow servicing of the maximum demand can lead to unpaid receivables.
Exactly the right individual balance between these two extremes can be found for the various customer segments by separately forecasting the dimensions ‘risk’ and ‘potential’ in a multivariate model, taking other relevant influencing factors into consideration. Finally, an optimal credit limit is produced by mathematically combining the risk and potential dimensions while taking the degree of risk aversion into account.
Next Best Offer
Optimized recommendations for your customers
In order to present a visitor highly relevant offers with good chances of acceptance, the following factors must be considered:.
- Customer lifecycle phase
- Customer behavior over the course of the contract
- Preferences (e.g. contact channels or product use)
- The context within which the customer contact occurs (request, contract modification, complaint, etc.)
Analysis of available data and factors influencing product bookings in the customer history can produce an exact profile from which recommendations can be drawn. Next best offers, calculated by us, can be integrated into your operative systems, facilitating cross-domain decision making.
Data collected in a wide variety of channels (CRM and call center systems, online and/or self-service portals, campaign management systems and email marketing applications) are used repeatedly to keep optimizing the results.