Client: A leading power generation company
Business need addressed:
- Time-consuming and error-prone manual processes of examining complex blueprints to identify the stock keeping units (SKUs)
- Long cycle time in providing accurate quotes to contractors
- Missed sales opportunities and poor conversion on sales inquiries
- Lower win rate and decreased revenue
- Pattern-recognition technologies accurately read and decipher large technical blueprints
- Machine learning and natural language processing (NLP) algorithms parse the extracted data and label the SKUs
- Dynamic workflow cuts time to generate quotes for customer inquiries
- Reduced inquiry-to-quote cycle time from 10 days to 1 day, boosting win rates by almost 4x
- 30% reduction in cost of sales support operations
- Quotation and bill-generation time reduced from 3 hours to 10 minutes
Pattern recognition and machine learning represent powerful digital technologies that can emulate, and in some areas even outperform, humans in completing complex, judgement-based expert tasks. The enterprises that are first to identify the right use cases and deploy these technologies at scale will gain significant competitive advantage. This leading power generation company is one of the early adopters of the technology; and is using it to exponentially scale human expertise and domain knowledge, cut cycle time, achieve better win rates, and boost sales.
Industrial manufacturers lose sales due to delays in responding to sales inquiries. Perhaps not surprisingly, they have struggled to scale an inquiry-to-quote process that involves tedious interpretation of large, complex technical blueprints.
A typical inquiry-to-quote cycle in industrial manufacturing companies starts with contractors sending complex engineering blueprints to their distributors for quotes on components such as electrical panels. The process of getting a quote is intricate, as distributors must manually examine each page of the complex blueprint, and read data tables to get a list of parts—contractors prefer using product technical requirements instead of standard part numbers for specifications. For example, contractors would likely mention “20 Amp 3 pole circuit breaker,” not a specific part number. The distributor must then convert these into SKUs, which is another cumbersome exercise, with millions of hours spent on the process.
A single supplier can receive as many as 300,000 quote requests per year and spends an average of three hours to complete each quote. Typically, the turnaround time for a distributor to receive a quote from a supplier is as long as three weeks, on account of the backlog. This delay results in poor conversion and lost sales while wasting expensive time and effort.
A computer vision solution combined machine learning and natural language processing (NLP) capabilities to read, process, analyze, and understand large blueprints. The algorithms were refined through the experience and domain knowledge of the engineers to accurately extract parts data from engineering diagrams and label them as SKUs correctly.
A dynamic workflow was created to automate the assignment of tasks between sales support engineers and machines in the following sequence:
- Sourcing blueprints from the distributor
- Automated decoding of blueprints to generate quick and accurate quotations/bills of material
- Delivering quotations to the distributor alongside value-added recommendations through a digital gateway
Computer vision allows the manufacturer to quickly respond to sales inquiries and also reduce time and effort across sales support operations. The resulting benefits include: (See Figure 1).
- Increased win rates by almost 4x
- Reduced inquiry-to-quote cycle time from 10 days to 1 day
- Lowered cost of sales support operations by more than 30%
- Reduced quotation and bill-generation time from 3 hours to 10 minutes
The solution created additional impact in the form of increased partner satisfaction across the value chain. It improved data accuracy for future use while freeing up engineers to focus on value-adding activities.
Figure 1. Reimagining the inquiry-to-quote process
Use cases indicate that computer vision is maturing differently in different industries. A great example of the application of computer vision is a major retail chain setting up counter-less stores without point of sale (POS) systems. Similarly, computer vision can be integrated with other advanced technologies to provide innovative solutions to solve even bigger problems, such as banking fraud. The results have already begun to show; at the same time, many possibilities have yet to be explored. A discovery process that involves human-centered design thinking and a practical implementation approach that leverages the unique strengths of humans and machines to complement each other, leads to better outcomes, at scale, than either could produce alone. This is the key to realizing the full potential of such technologies.