The Problem
The company observed that their discounts (in dollars) were growing faster than the actual bookings (in dollars). Millions of dollars were thus being left on the table due to their suboptimal discounting practices. Also, there was a wide variation in discounts offered for similar negotiated deals.
The Before state:
- There was no clear guidance to the Sales representative about what an ideal discount for a deal is. The existing guidance available for them before sending quotes for approval is only blanket rules for discounting. This was resulting in discounting variations apart from delays in quote approvals and reaching out to customers.
- A one-size-fits-all pricing approach results in over and underpricing deals, leading to either loss of sales or leaving money on the table.
- Given the amount of infrastructure and efforts generally invested in a software product, customers are usually reluctant to switch to a different one. This stickiness with the software products means that there is a good scope for the companies to earn revenue from these products in the form of term licenses, subscriptions, cross-sell and upsell opportunities. Pricing teams and sales representatives in the B2B software space make indiscriminate discounting decisions to win or retain customers now and extract more revenue from them in the future.
- Traditionally, salespersons played a pivotal role in pricing decisions due to their immense knowledge of the customer and their environment. While sales representatives understand customers better, most of them generally don’t have enough visibility into how similar products are discounted by peers and believe that discounts primarily drive deal conversions. This could result in a significant deal-pricing variability. With many factors involved in the deal context, it is essential to augment sales representatives’ valuable experience/intuition with robust data-driven/analytics-based discount guidance.
The LatentView Solution
- The goal was to develop an intelligent solution that learns from past deals (wins and losses) to understand whether a particular deal has inherently favorable attributes that make it more likely to win or lose and provide deal-specific discounting guidance accordingly.
- The methodology used was a multi-step solution approach that leverages the below three custom modules that we developed:
- Segmentation Module: We used popular clustering algorithms like k-Means Clustering and Spectral Clustering to group similar quotes.
- Classifier Module: Discounting decisions is a complex culmination of several factors including the customer’s relationship, deal type, deal size, product mix and proposed discounts. All these together decide whether a quote is likely to get converted into a booking.
- Optimizer Module: The Optimizer has been designed to maximize revenue from each quote and maximize the chances of winning the quote. This essentially searches for the most optimal discount between the mean transactional discount and the inflection point while maximizing deal win probability and the revenue.
- The most significant differentiating factor and the USP of the project were identifying the factors that dictate a deal’s conversion, predicting the probability of quote conversion, and identifying the optimal discount within the calculated range that ensures that both the above objectives are met. Also, it leverages customer attributes to make predictions rather than focusing only on quote/booking data.
The After state
- The tool is expected to save at least $42 million annually apart from reducing the need for external manual intervention for a majority of quotes.
- The tool is expected to increase the quote conversion by 4.8%.
