ERP Data Mining: what we learnt from analysing 100 million of B2B sales transactions.

Data mining is the application of a varied assortment of statistical techniques to ERP datasets. Companies nowadays use data mining to predict outcomes, identify sales trends, prevent customer churn, and dynamically adjust pricing strategies.

Mining enterprise resource planning (ERP) sales data is critical in Business-To-Business, where small improvements in sales efficiency can have a significant impact on results. Mining ERP sales data helps customers to unlock a significant amount of value, discover quick-wins, and to prioritise their sales activities. Automatizing this process is possible today with the help of artificial intelligence (AI).

Our solution integrates artificial intelligence into ERP systems. An AI enabled ERP allows companies to identify cross- and up-selling efficiently, to dynamically adjust prices and to reduce customer churn.

Since we launched our predictive analytics tool, we analysed more than 100 million of sales transactions, integrated a similar number of CRM activities, and created dozens of different data mining models.

Here what we have learnt so far. We listed some of the key takeaways we found over and over across thousands of ERP datasets.

B2B companies can find gold mining across ERP data.

What is ERP data mining? One of the projects with the highest Return on Investments a B2B company can nowadays find. If data is the new gold, ERP data mining is gold extraction.

What we have learnt so far is that most B2B companies still haven’t dug for this gold and its value is buried in their data. We often ask ourselves why this is the case. We can think of three main reasons.

First, since almost 75 % of all enterprises had a stressful time implementing its ERP, they believe mining ERP data will also be expensive. They are wrong. Integrating data mining in an ERP or CRM is nowadays extremely affordable.

Implementing an Enterprise Resource Planning System can take up to two years and cost millions. Thanks to cloud architecture and artificial intelligence technologies, data mining in ERP is exceptionally affordable and low risk in comparison. Most managers ignore this. Sadly, the buzz around Big Data is not helping much to create trust.

Second, since B2B companies (especially in Germany) are doing relatively well (average sales growth above 5 %; historical record export), sales leaders don’t think of improving profits as a priority. What they neglect is the impact predictive analytics can have on sales performance.

Lastly, sales leaders that have maybe already tried using predictive analytics or data mining to beat their competitors ended up using Excel. Excel is a great productivity tool, just a time-intensive data mining and a lousy sales planning tool. There are today better options. Tools that are easy to use and that do not require any previous knowledge in data mining.

Knowing that companies haven’t yet dug for gold in their ERP is not all bad news. Once they start, they make an exciting discovery: mining ERP data is money-wise desirable.

I want to see what your experience in ERP data mining can do for me.

ERP data mining offers today a highly attractive ROI. Just start with these three methods now.

Mining ERP data can offer ROIs over 800 % in three to six months. Using data mining, a sales leader can discover cross- and up-selling, save sales costs, reduce customer churn and dynamically adjust prices. All these activities have a tremendous impact on both the sales bottom line and the profitability of any business.

These are the three “low hanging fruits” predictive analytics techniques in B2B: dynamic pricing, churn discovery and cross-selling.

Pricing analytics is the data mining technique with higher impact in a company profit. We estimate that selectively adjusting prices for about 5 % or detecting customers that are overlapping several discounts can increment earnings by 40 to 50 %.

Predicting customer churn is the second most profitable data mining method in B2B. Increasing costs of customer acquisition force companies to invest heavily in retention. Likewise, costly sales representatives make impossible to manage all buyers the same way. The answer? Predicting customer churn and segmenting sales and marketing retention activities.

Lastly but still relevant, discovering cross- and up-selling can have a tremendous impact on the bottom line. Most of the companies that come to us are already trying to find undetected opportunities, albeit manually, maybe using a data visualisation tool. Commissioning your salespeople with a data visualisation tool for sales analytics is time-consuming and extremely ineffective.

For example, by implementing an automated apriori algorithm for data mining, key account managers get precisely what they need: prioritised product suggestions based on the likelihood that a customer will buy.


Forget artificial intelligence. Classic rules of sales still triumph.

At the beginning of the past century, Vilfredo Pareto, a Paris-born Italian civil engineer, observed that 80 % of the land was in the hands of 20 % of the population.

His study is today commonly identified as the 80/20 rule, named after him as “the Pareto principle”. This principle has countless empirical examples and follows a power-law probability distribution.

The distributions of an extensive variety of biological, physical, and human-made phenomena follow – approximately – the Pareto distribution. These include the sizes of activity patterns of neuronal populations, the foraging mode of some species, the dimensions of solar flares and craters on the moon, the frequencies of words in most languages, frequencies of family names, among many other quantities.

For years sales practitioners and marketers have been talking about the Pareto Rule as a holy grail of sales and marketing.

After mining millions of ERP sales transactions, we came to discover, that Pareto seems to hold in almost all the cases and sales situations.

Fascinatingly, we learnt that not only most companies make 80 % of their revenues with 20 % of their customers, but they make it with just 20 customers.

This finding has vast implications for sales planning and strategy. For once, it means that Key Account Managers should focus on the top 20 accounts that can move the needle. Companies should thoughtfully consider how to serve the rest using e-commerce, e-mail marketing and other sales acceleration solutions.

ERP Data Mining: what we learnt from analysing 100 million of B2B sales transactions – Conclusion:

Mining ERP data is no longer an expensive endeavour. Businesses today use data mining methods to detect sales trends and to predict outcomes. Unfortunately, most of the B2B companies still ignore how to do it and how much of an attractive investment it can be. Companies still believe that mining ERP data is expensive and high-risk. Due to changes in technology, this is no longer the case.

Mining ERP data offers returns of more than 800 %. Most of the projects pay themselves back in less than half a year. Furthermore, companies can achieve this compelling ROI with just three predictive analytics applications.

Sales leaders in B2B can dynamically adjust prices, reduce customer churn and discover cross- and up-selling using data mining. These activities have a tremendous impact on the profitability of most businesses.

We also learnt that, in this new artificial intelligence world, some characteristic rules of B2B sales still apply. For example, after mining millions of sales transactions, we can say that the Pareto rule seems to hold true. Most companies make 80 % of their turnover with just 20 customers.

I would like to start today.


Free eBook for download: How To Get Started With Predictive Sales Analytics – Methods, data and practical ideas

Predictive analytics is the technology that enables a look into the future. What data do you need? How do you get started with predictive analytics? What methods can you use?
Download the free eBook now.

free eBook How to get startet with Predictive Analytics


Predictive analytics – how much data do you really need?

Further Read:

Provost, F. and Fawcett, T. (2013) Data Science For Business: What You Need to Know About Data Mining & Data-Analytic Thinking. O’Reilly.

Grus, J. (2016) Einführung in Data Science: Grundprinzipien der Datenanalyse mit Python. O’Reilly. (in German)

Witten, I. (2016) Data Mining: Practical Machine Learning Tools and Techniques. Morgan Kaufmann Series in Data Management Systems.

Goodfellow, I., Bengio, Y. and Courville, A. (2016) Deep Learning (Adaptive Computation and Machine Learning) – The Mit Press.