Using Machine Learning for pricing in B2B sales
Sales are the bread and butter of every business.?
Everybody is familiar with the Byzantine ways in which plane ticket and hotel prices fluctuate. Demand-adjusted pricing was again brought into the spotlight with the price calculations of Uber, Bolt and other ride-sharing apps. What do they know, or need to know, in order to determine that those are the "right" prices? How can a business manufacturing waterproof widgets determine what the right price is for them? How can one even define the right price? As the above examples show, the right price is the one that clients are willing to pay for the product, which at the same time maximizes profit for the seller. And often, different clients value the same purchase very differently.
Even within the limits on collecting, storing and processing data given by the GDPR, companies nowadays have a wealth of data from their clients. Should the client be another business, and therefore not subjected to the stringent European privacy legislation, also numerous other sources are available, such as number of employees, turnover, and others.
The large amount of factors involved can be overwhelming for the specialists deciding on how to correctly price products or services. Inner knowledge in the relevant departments, personal experience, anecdotal evidence, gut feelings, personal biases and huge Excel spreadsheets are not foolproof. The large amounts of data might make the formulas overly complex and overwhelming.
Clients regularly need highly customized products and prices for their requests. Let's think about a low-cost airline and a billionaire. They might end up with the very same Boeing 737, but their motivation for purchasing it, as well as the technicalities of the purchase and the specifications of the cabin will be vastly different. If the price, given the requirements, is not "just right", the client might end up buying the equivalent model from the competition (in this case, Airbus A320), or a different model altogether, or just deferring the purchase for another time. One must also remember that, same as the vendor, the potential client also has a cornucopia of data, and might be more than happy to use it and not show their cards in a negotiation. Both seller and buyer find themselves in a technological arms race towards what they can consider the acceptable price.
The complex question is: How can a seller decide what the optimal price point is in order to make a winning bid?
领英推荐
?Luckily, given the overwhelming amounts of data available, machine learning algorithms can ingest it and offer a recommendation. Some companies work on AI-based pricing solutions which they sell to their own clients. A quick online search will reveal several such tools available for purchase and customization. Others are developing such solutions in-house or with help from their own external collaborators or contractors. AI and "data-driven organization" are therefore in this case much more than just a buzzword or an empty but catchy corporate slogan. It is a real solution to a multitude of real problems, it can bring genuine added value. Client segmentation, client churn, discount offers, sale leads predictions are all tasks which can be solved using machine learning. Corporate leadership is also onboard: surveys in the past 1-2 years show that most high-level executives are sure that AI will significantly change ways to do business - and in numerous cases, it already does! In their 2019 work "Automating the B2B Salesperson Pricing Decisions AI Human-Machine Hybrid Approach" Yael Karlinsky-Shichor (Northeastern University) and Oded Netzer (Columbia University) quantified the improvement in profits from using AI - and there are numerous other papers or conference talks on the topic! Therefore, it is definitely time for businesses to get onboard. The shift from a "one-size-fits-all" paradigm augmented by some simple and static rules towards a personalized pricing is already happening.
No human can be completely free of biases and preconceptions. Machines, on the other hand, are great with data and patterns, even less obvious ones. Therefore, as already mentioned, past experiences and "known wisdom" - while being a great starting point in exploration for a solution - can cloud human judgement.
?What sort of data is relevant? Common wisdom has that the size of the purchase, the length of the client relationship, the type of customization, and the strength of the competition are generally the most important factors in setting the price. Are there also other fundamental ones? One might think about history of past sales, location, type of client, product mix, socio-economic and political climate... Sellers have this data, it can be leveraged at any time. But, even data which is completely independent of the seller, the buyer or the sold product can have an impact. Beer sales correlate well with the weather and with large public events such as fairs, rock concerts, or big football matches. Event calendars and weather forecasts are easily available online, and such data can be ingested into a model. Let's put the reader in the shoes of a beer seller. Will this information be used in different ways for the pricing strategy if they are from a big mainstream brand vs a small local artisanal microbrewery? Each will need its own custom-made pricing AI. And a second thought exercise for the reader: what external factors can influence the price of, say, smartwatches, reinforced concrete girders, or a sales consultant's hourly rate? How might local laws influence the cost of oversized transports in different areas? This overwhelming variety in totally different types and sources of data, often requiring different domain knowledge to even consider, adds another layer of complexity. Some models have the capability to quantify the importance of specific factors, both when taken in isolation, as well as the interaction between them.
An AI's role is not to replace specialists in pricing and sales. In B2B sales one is not only in competition with other vendors, but one needs to convince the potential client that the purchase is genuinely worth it for them - and will bring some sort of value for the organization, their clients, employees, or other stakeholders. AI can also react well to changes in circumstances, when suddenly changing data will quickly lead to a change in the suggested price. The question is not only "Should I buy waterproof widgets from vendor A or vendor B", but also "Do I have a use for waterproof widgets?". Sale specialists' role can migrate towards customized communication with the client, building and maintaining interpersonal relationships, and allow an AI to "get their hands dirty" with the heavy behind the scenes computations. It will also allow for faster price offers and a more proactive communication with the prospective client. Therefore, sales specialists using a pricing AI are not "obeying to the future robot overlords" as some might fear, but are becoming empowered by having an additional powerful tool in their arsenal.