Here are five interesting facts we discovered while analyzing sales transactions of B2B companies.

Are you interested in learning from your sales data? Artificial intelligence makes sales teams truly data-driven, while offering exciting insights. Learn what other companies have discovered with us now!

How did we get to these interesting facts? Using our Predictive Sales-AI software. Our AI-based predictive sales software analyzes sales transaction data and generates accurate customer behaviour predictions (cross-selling potential, pricing, and churn risks).

We have trained our algorithms with about 50 terabytes of ERP sales transactions over the last few years. 50 terabytes of ERP sales transactions is a huge amount of data. Remember that sales transactions are not images nor audio files. Only a few companies worldwide have access to such an amount of data.

So, the five Fun Facts all came from customer cases which, we would like to share them with you. Several customer cases have confirmed the facts – but we would not go so far as to say that they are transferable to everything. These are still anecdotical facts and not representative statistics.

1. Not all additional data add value for more precise forecasts.

At the beginning of an individual customer case, we include all the data that a B2B company has available in our analyses. Unaggregated sales transactions are the first prerequisite – nothing works without them.

However, suppose a company has a maintained CRM system with additional information about its customers (e.g., complaints or how often there is a contact) or competitors. In that case, we feed our algorithms with that data as well.

However, the goal here is not “a lot helps a lot”, but to find out “what helps a lot?”. Regular data gathering and structuring are very costly. Therefore, it is precious to know what data do I need at all for a precise result?

Because a good horse only jumps as high as it should.

In cases like this, we have found through our software that certain information adds value and other information does not.

Here is an example from a customer case that involved predicting customer churn:

We had included various criteria for customer emails in the software. However, only the fact whether there was email traffic (in a certain period) – yes or no – had a significant impact on the prediction. All other information, such as the content or length of the emails, was superfluous.

Funny, isn’t it?

2. Customers who complain are less likely to churn.

Not just one, but several customer analyses demonstrated this fact. Our software has detected a pattern that customers are more likely to be loyal when they complain. We were astonished by this finding.

However, we have found out that several studies on customer satisfaction already discovered this fact(long before us). Ok we are not the first – but still pleased 😉.

Our personal takeaway is: treat the customer that complains as one of your most loyal customers and predict the churn probability of those who do not.

3. AI Technology does not solve ethically critical questions.

It happened, that algorithms discovered a correlation between employee ID (salesperson) and cross-selling potential, pricing, or customer churn. Are you surprised?

Cross-selling and pricing are relatively innocuous. But what if a correlation occurs between customer churn and an employee ID? In other words, customers assigned to Key Account Manager Ronny Arias are more likely to stop buying from your company.

First, correlation does not equal causation! That is, just because there is this correlation, you absolutely cannot say, “BECAUSE employee XY was on, customer Z left.”

You can still investigate the reasons behind that correlation.

And this is where the ethics start: Do you want to? Do you want to go up to your employee and call them on it? It entails a rat’s tail of measures. And finally, the works council would have to be involved.

Perhaps it is advisable after all to ID the employees and not make this correlation visible.

The bottom line is that regardless of how powerful your AI sales software is, successful executives clear to way for ethical issues during implementation of an AI Software.

4. Pareto principle is still relevant.

The Pareto Principle – or also called the 80/20 Rule:

“80% of the results can be done with 20% of the total effort.”

In terms of our customers, it is more like, “80% of the revenue is made with 20% of the customers.”

Especially with customers in B2B industrial distribution, this Pareto principle is valid (in our experience!). When our customers start with AI, of course, there is always an “as-is” analysis. So, we have noticed that some of our customers come quite close to this distribution. Not precisely 80/20, of course, but in any case, a clear minority of customers is responsible for a clear majority of sales.

What does this mean for sales planning? In our opinion, key account managers should focus 100% on the 20% of customers eager to generate sales.

However, they should not neglect the rest. Helpful tools and technologies such as eCommerce, email automation, and AI-based predictive analytics software can get the most out of the remaining (low-value) 80% of customers.

So, in terms of sales, you could say modern technologies make the job 80% easier.

5. Human total forecasts top – individual forecasts flop.

One application of our software is demand forecasting. Total sales, but also more detailed individual sales forecasts per customer and product. So: “how much will be sold of product XY?” Or “how much will customer XY buy?”.

Most B2B companies make forecasts themselves by asking their salespeople, gut feeling, and through various Excel projections.

If a customer has made such efforts, we are happy to offer, comparing their forecasts with our software for the coming quarter.

We call our clients’ forecasts “human forecasts” and our “software forecasts” for simplicity. Which predictions are closer to reality?

Here it turned out that the total “human” sales forecasts are not extremely bad. The difference with our software forecasts was not significant. That shows that good intuition and experience are often right. Salespeople were, however, totally inaccurate at the customer – product level. The aggregate for the human errors just compensated themselves.

The software wins by faris in the more detailed individual forecasts. The deeper it goes into detail, the more “human hits” went outside of range.


Five Fun Facts from our Software – Conclusion

That was a little insight into our software findings. In our opinion, data-driven – or better – data-enabled sales is the future.

Insights from your data allows the sales teams to prioritize, act and thus save time and money.