In a recent article, I described how artificial intelligence (AI) can help sales and marketing to be better integrated and effective. That, though, focuses on initial sales and often has more to do with B2B sales of unique products such as software. A very large issue in sales is how to handle commodity and volume pricing via channels. It is a different type of challenge but is just as important.
Channel sales are complex, especially when multilevel channels can include distributors, wholesalers and retail. Two of the key incentives to get channels to push products are rebates and chargeback programs, ways in which the manufacturer can cover some of the cost or risk of the channels. However, what’s an effective plan? The primary existing model, for most companies, seems to be a spreadsheet near-equivalent of sticking a wet finger in the air and seeing which way the wind is blowing.
Lots of data exists, the challenge is to figure out the right features, and the right level of features to use. “When we first began to look at this problem,” said Chanan Greenberg, GM High Tech, Model N. “We quickly understood that early AI systems were too granular, with tight segmentation that wasn’t adaptable. For instance, looking at individual products wasn’t adaptable. On the other hand, product family was too generic. Testing models shows us that like-products within families provided good analysis.” Why I’m reminded of Goldilocks remains a mystery.
The early revenue management systems also quickly integrated with ERP, as access to orders, pricing, bills of material, and inventory also were necessary to evaluate whether or not a plan would be successful. As usual, the more data you have, when it is chosen wisely, the more accurately a model can be trained. Given the fluctuation in the models, Model N is working to provide confidence level information for their predictions, with those levels increasing the more repeat buys that exist.
Part of that accuracy was also understanding the details of all sales. “We found that half of B2B high tech sales were off the book, were special pricing,” added Mr. Greenberg. Being able to analyze those is critical to help train salespeople about how and when to go off-book.
As data matters, it should be no surprise that revenue management models are first being attempted at firms and in industries where significant historical data exists. Training a model without that data is not a good choice. Given enough data, Model N says it uses XGBoost as a baseline, adds some custom tuning, and runs their neural network to provide channel analysis.
With enough accuracy, it’s not just reporting that AI can address. AI differs from business intelligence (BI) in the speed of predictive analysis and action which can help automate the supply chain. By understanding what has worked in each situation, an AI system can automate many quote negotiations. Standard re-orders can easily be filled by combining AI and robotic process automation (RPA). Analysis of existing inventory and sales in other channels or regions can result in the system suggesting deals based on availability and timing. Rebate and chargeback programs can be instantly updated and more efficiently managed.
One intriguing thing about revenue management focus was in the Model N 2020 State of Revenue report on a survey they ran. As much as I love AI, I regularly join the chorus of folks talking about how it is overhyped. Evidence of that hype is in the numbers. When executives listed the factors causing increased difficulty in revenue management, AI was number one (page 6). Cloud was down in the fourth place. The next set of numbers shows the increasing number of “revenue execution moments” increasing.
Those moments are increasing because of the subscription model expansion driven by the cloud. Customers have more flexibility, so the need to manage those customers increases in frequency. In addition, one key reason AI has gained leverage in the market is because cloud computing provides the performance needed to make deep learning efficient.
Another interesting part of the report is that senior executives are concerned about the impact of regulations on revenue (page 17). The risks expressed are focused directly on the regulation of their own products. Executives need to look further. More and more governments are looking at data privacy regulations, and regulations for transparency in the use of AI and other technologies will follow. Those regulations will impact how companies relate to prospects and customers, as well as how the companies manage their own computing resources.
Artificial intelligence is an important tool, and it will end up making an impact on every business function. However, business must focus on the whole picture. Right now, AI is shiny, and it’s yet again being overhyped. It is of value, it’s mature enough that I don’t think there will be another winter, but the bubble can burst.
Sales channels, especially in commodities industries, can be very complex. AI can provide advantages in providing visibility into channels and in managing to improve channel performance. While earlier work on sales has focused on the sales & marketing cycle for lower volume products, the need to better handle high volume sales through channels is an area now being investigated, and one that can provide value to companies through leveraging artificial intelligence – if it is done with open eyes.
To get more details click here.