How can Google Cloud support B2B price optimisation and demand forecasting?

How do we provide optimised pricing to a B2B buyer when, as we covered in Part 3, this often has several intrinsic steps and dependencies?

In Part 1, when we explored the opportunities for B2C, we looked at how historical transactional data and other signals are imperative inputs for calculating the price of traditional consumer retail items. In B2B, there is a far greater degree of analysis and pricing science that goes into pricing optimisation. This means developing and deploying a system that enables a B2B retailer to reset price targets in real time at a customer/product level, based on actual facts, as well as developing a dynamic-pricing capability that places as much on people and processes as it does on analytics and technology. It is this additional level of complexity, and the need for simulations, which creates the opportunity for B2B retailers to make use of predictive analytics, and statistical modeling to predict the impact of pricing.

As with B2C, we start with the same baseline for (automating) price optimisation, but there are some key notions to consider in this setting, and capabilities in the Data Cloud that lend themselves well to address these nuances:

• What is the B2B Customer’s Lifetime Value (CLTV) to the B2B retailer, and is a price variation likely to have a positive or negative impact on this longer-term? Whilst CLTV has been an established measure by a number of B2B retailers, and there are a number of techniques to influence this, it does have its challenges i.e. availability of data, performance across different channels, using history to predict future transactions, customers with little/no historical transactions, identifying and amalgamating probable events, etc.. We use the Data Cloud to unify data and AI/ML, and address some of these challenges. The Data Cloud provides the ability to complete lifetime value calculations that span both probabilistic (RFM — Recency, Frequency, Monetary) and ML models, with the latter offering some distinct advantages. The net result can then be visualised, analysed and tracked through a Business Intelligence (BI) tool like Google Cloud's Looker.

• What is the B2B customer’s likely response to a price variation? Unlike B2C where the price may not vary often, a B2B retailer may vary (unit or volume) pricing from order to order. As noted in Part 3, this is often a cause of complaint by B2B customers, and is where the Data Cloud’s AI/ML comes into a world of it’s own by helping B2B retailers simulate possible scenarios and perform propensity modelling i.e. predict and visualise customer propensity to purchase (through Looker). Also, with emails continuing to be the sales channel that accounts for the majority of B2B revenue¹, we can incorporate the combined power of sentiment analysis of previous/recent email communication using the Data Cloud’s Natural Language AI and Translation AI (for globally distributed B2B retailers).

• What is the state of the relationship between the B2B retailer and consumer? If the relationship has been recently strained e.g. last order was cancelled due to a pricing discrepancy, should the outcome of the pricing optimisation be favourable to the customer? Where a B2B retailer is gauging and quantifying customer loyalty through proven feedback metrics such as Net Promoter Score (NPS), Customer Satisfaction (CSAT) and Customer Effort Score (CES), we can use the Data Cloud’s Machine Learning (ML) and the Cloud Natural Language AI to identify areas of the relationship that need attention as shown in this example:

¹ McKinsey: Busting the five biggest B2B e-commerce myths

• What is the risk of the B2B customer churning as a result of a pricing decision? Customer churn is a residual concern for many B2B retailers, and retaining customers in this new digital age requires a different way of thinking and new approaches to technology. It is important to predict at-risk customers and intervene with pro-active support, learning what is driving customers to leave and developing strategies to prevent future churn (similar to the above). One possible approach is to leverage BigQuery ML's predictive capability and extend this with Vertex AI, incorporating B2B customer churn predictions into a Looker dashboard, or directly into a CRM to power targeted interventions such as targeted incentives.

Let us consider the scenario whereby a B2B retailer is presented a particular deal and is trying to identify the optimal pricing strategy. Through the use of the Data Cloud platform, this can be scored against peer groups and factoring multiple different signals into price recommendations e.g. strategy, deal size, customer type, and product type and mix.

Having established an optimal price point between B2B retailer and customer, how do we mature our demand forecasting to address the common B2B enquiry of “where is my stuff?”

Demand forecasting is one of the most difficult parts of B2B supply chain management, especially when you consider the number of factors that can affect it. It is also the most important, as all other pieces of supply chain management are based on predicted demand, including pricing, promotions and logistics optimisation. Many B2B retailers struggle with finding that balance when it comes to inventory, but transparency is an important strategy when building trust.

When we explored demand forecasting for B2C, there were three key influential factors:

• Retailer-customer relationships were fluid and not ‘sticky’ (beyond cases where there may be membership or loyalty programs in use). This meant that a broader range of data signals would be required to derive demand forecasting. Whereas in a B2B setting, these relationships are, more often than not, well established and predictable. Where new relationships are drawn up (often through commercial agreements), it is often possible for client relationship managers to provide demand inputs that are subsequently used to derive demand forecasting predictions. This is a good example where a demand forecasting solution not only produces predictions through a BI dashboard like Google Cloud’s Looker, but also democratises this data insight so that it’s accessible to various stakeholders across different lines of business to influence the modelling that takes place.

For existing B2B relationships, historical transactions can often act as a reliable indicator of future demand/buying patterns - although not exclusively relied upon. We always have to factor an element of probability into our demand forecasting predictions.

In both cases, and similar to what was covered in Part 1and Part 2, data from a Customer Relationship Management (CRM) and/or Enterprise Resource Planning (ERP) system(s) can be ingested into the Data Cloud platform, before subsequently undergoing data pre-processing and used by Vertex Forecasting.

• Pricing for goods were fairly fixed, but could fluctuate based on key events (product launch, seasonality, campaigns or trends, end of line sales). For a B2B retailer that is setting and optimising pricing as part of an initial agreement, this can often be part of committed orders and volumes, or part of call-off orders. This key information can be ingested into the Google Cloud Vertex Forecasting solution, and used in conjunction with historical data, and augmented with current market trends and external market factors to make accurate demand predictions. Similarly, if pricing is renegotiated at any point of the B2B relationship, this may result in a downwards or upwards revision to the demand forecast. 

• Stock was fairly fixed or may not vary often. With B2B, and the significant variation in stock, there’s a great emphasis to not rely solely on historical data. For new stock items, and given the current volatile nature of the global supply chain, we would always look at incorporating as many data signals (or demand drivers) as possible into the Data Cloud platform and Vertex Forecast solution - particularly those that may come from others within the supply chain, or competitor/market information that we are able to glean.
  For regular/fixed stock items, we can use past B2B customer behaviour to make predictions about likely future purchases by B2B customers over time - particularly in verticals or market segments where demand/supply have been predictable. This gives us an element of predictability, but a less than optimal prediction confidence (score).

With the lack of speed of interactions (with suppliers) emerging as the number 1 pain point by many B2B customers, many B2B retailers also need to consider how they facilitate self-service and automation into existing processes. Some 86% of respondents in the same McKinsey B2B survey said they prefer using self-service tools for reordering, rather than talking to a sales representative. And whilst we haven’t covered the role ofpersonalised voice and chat bots in improving the B2B customer experience, or Google Cloud’s Document AI to automate data capture at scale, both are key adjacent technologies that are implicit parts of our combined AI/ML approach. 

It is essential that the B2B retailer is not just empowered with the tooling and relevant insights to effectively apply pricing optimisation and demand forecasting, but to then automate, integrate and scale this across the relevant lines of the business. This allows them to use AI/ML to drive more business impact, and position the business for stronger ROI.

By using the Google Data Cloud, Kin + Carta can help B2B retailers close the data to value gap. We improve and speed up the decision-making process, while providing more granular insights. This accelerates value by breaking down data silos and leverage AI/ML to realise better outcomes, and enables better planning through demand forecasting, Inventory management, assortment planning, and dynamic pricing and promotions.