Powering Costing with Artificial Intelligence: The case of Vodafone Procurement

Artificial Intelligence

By Hervé Legenvre, Gavin Hodgson and Govind Khandelwal

The present article describes how the Vodafone Procurement Company (VPC) has adopted Artificial Intelligence technology to boost its ability to perform “Design2Cost” and achieved significant cost optimisation as a result. The adoption of digital technologies including Artificial Intelligence allows the automation of activities performed routinely by humans. This positively impacts the efficiency of these individuals and allows them, in the best case, to focus on more value – adding activities. At the same time, such technologies can be used to augment the work of procurement professionals by allowing teams to deliver more value. Augmentation typically supports complex and collaborative activities that were not be systematically performed before.

How it started

The Vodafone Procurement Company was established in 2008 to serve as Vodafone’s centralised procurement hub. By 2017, VPC had reached a plateau in terms of maturity. Price negotiations were facing limits, leading the organisation to ask: What’s the next step? The company needed further opportunities to eliminate costs, but it also needed healthy suppliers who could invest in R&D. Procurement experts at Vodafone knew that they were most valued by their stakeholders when they were bringing not only savings, but also knowledge, facts, and options to the table. Knowledge on the design and on the detailed costs of the products and services provided by suppliers plays a key role here. All this led to the idea of further investing in the integration of cost analysis capabilities within the procurement team. So-called ‘design to cost’ allows the setting of an objective cost goal for a product by breaking down the product into sub- elements and assessing their respective costs and benefits. This allows for the rethinking of products and services, achieving cost reduction and increasing the value delivered.

So-called ‘design to cost’ allows the setting of an objective cost goal for a product by breaking down the product into sub- elements and assessing their respective costs and benefits.

The Vodafone team performed benchmarks with automotive companies and looked across the telecom industry to assess existing practices. This provided a good basis to start a pilot, the results of which were very promising and a business case was developed. The initial aim presented in the business case was to save 300 million Euros over 5 years, but the stretched goal was to achieve cost reductions of 1 billion Euros within this 5 year period. So far, after three and a half years, over 50% of Vodafone’s global spend has been influenced by the team and hundreds of millions of Euros have been saved, putting Vodafone well on track to achieving its stretched ambition. Over 250 pieces of hardware have been analysed to date and the pipeline is full for the coming six months.

Setting up a Design to Cost Lab

The project, branded Design2Cost, was launched at VPC in a temporary lab located in a meeting room. Following the success of the pilot, it was clear that more space was needed to perform their cost teardowns, so Ninian Wilson, CEO of VPC, said: “Take the boardroom and build your lab in it. This is the best way forward. You need more space!”  The Design2Cost initiative started small with a team of just five people, and today boasts 16 trained Design2Cost experts from a range of industry backgrounds including electrical engineers, mechanical engineers and manufacturing. On the ground in the lab the team focuses on the hardware teardowns with the support of a team in India who do detailed analyses of the findings. For services cost teardowns, VPC’s category managers are trained in cost analysis methodology enabling them to perform their own costing activities with the support of the Design2Cost team who give coaching and validate their results.

The process and how it was developed

Vodafone’s Design2Cost process was developed by creating a series of in-house tools. The team developed a systematic seven-step process and invested in access to many sources of information on costs. From the start, the process was developed as a cumulative learning process.  Right from the beginning they built a database that could be updated and re-used over time, meaning that it is continuously updated. For instance, the data on the cost of labour are updated every quarter. So, if you have a supplier in a specific region in Vietnam, the database gives you the labour for the appropriate skill level, energy and factory floor space costs in that region. Many sources of information contribute to the database and different sources are regularly compared so that it is reliable.

From the start, the process was developed as a cumulative learning process.  Right from the beginning they built a database that could be updated and re-used over time, meaning that it is continuously updated. For instance, the data on the cost of labour are updated every quarter.

The analysis process can be described as follows. Whenever a procurement project above a certain value starts, category managers must confirm they have considered a Design2Cost approach. When appropriate, the Design2Cost team starts by defining the scope and the context. This provides a basic understanding of the category and of the business issues. They then tear down representative pieces of hardware and scan all the components using a proprietary, internally developed system. This scanning equipment takes hundreds of images that are combined into a single ultra-high resolution image. Using a neural network, which is currently patent pending, software then recognises and defines the locations of components of interest. Other internally developed software then takes the images already captured, and markings extracted using Optical Character Recognition, to compare and match these components with the internal database built over the life of the lab. Up to 92% of electronic components on an unknown board are now being automatically recognised in this way. Each item within the hardware is automatically allocated a cost thanks to the lab’s database. If the information is not already in the database, the team creates a new cost model or looks for external cost information, always seeking multiple sources of information to ensure reliability. From this, they perform all the analysis using their software. The final cost model is completed by incorporating the value-adding steps as well as overheads, R&D, and any other third party costs. This allows the team to produce the information kit for category managers and support them as needed. The process is outlined in Figure 1.

Figure 1

The information kit provided at the end of the process is comprehensive, offering a complete overview of the teardown. It includes what they call a ‘clean-sheet’ costing, the lab’s own version of the costing. They also decompose the costs by subassembly, by components and offer detailed views on labour impact. The lab can also offer competitive comparisons by doing a tear down on competitors’ products. They suggest some levers with their cost impact.

Typically, the info kit provides category managers with a list of design optimisation opportunities as levers that can be used to reduce costs. These can run to 20 or more opportunities, but the actionable number depends on the maturity of the product.  This provide options. Then it is down to decision makers who specify the needs and to the supplier to see how this can be taken forward.

To create this process and the system that supports it the team has created a ‘garage innovation’ environment. Orlando Grigoriadisthe lab’s AI specialist described the development: “This started in my kitchen, I was putting together all the elements for scanning hardware together at home. Everything was developed in house, and sometimes really in my own house! This is often the case with machine learning, you can start with some algorithms that are available on open source but then, at some point you need to make it work for you.” Before this the team was using magnifying glasses and manual tools, they were looking at every single component on the board and identifying everything by themselves. Now they are teaching the machine to recognise all components so that everything is automated.  After the picture is assembled, the components are identified, and the data is aggregated automatically.

These developments undertaken by the team have had a tremendous productivity impact on their own work and performance. The time it takes to do a complex costing has been reduced from 30 to 10 days thanks to the tools that use AI and machine learning that were developed as part of the setting up of the lab. Today, the neural network technology developed by the team is registered as a patent.

Figure 2

In this approach the wide access to external data is also key. Figure 2 describes the sources of data used by the team. You need all these sources to be able to do a cost teardown quickly and effectively. When the team does not have access to the cost within its existing database, members of the team start to look at a wide range of external information. By looking at multiple sources, the information is triangulated. The lab has used thousands of sources of data to get information on more than 20 000 components.  The level of accuracy can now exceed 90% thanks to the machine learning algorithms. But the team still plays Sherlock Holmes sometimes, they dig deeper in technical papers to understand some of the unknown costs and they exchange and work with suppliers to better understand some of these. However, this is less and less required as information is accumulated and updated on an ongoing basis.

The benefits to the company

Since the lab was established, the team has performed detailed teardowns and cost analyses on hundreds of complex electronic and electromechanical products and services, including x86 servers, remote radio units, customer premise equipment and various types of deployment, marketing, call centres, IT and other services.

The impact of these activities has been significant: using the lab’s results, category managers have achieved a step-change in the quality of discussion with suppliers. They have scrutinised vendor cost structures on the most detailed level, uncovered hidden margins, and opened-up joint cost reduction and value creation opportunities with suppliers.  Sometimes this has helped uncover inefficiencies that were due to year of specifications that had been piling up on each other. Many examples of impact show double-digit cost improvement.

The impact of these activities has been significant: using the lab’s results, category managers have achieved a step-change in the quality of discussion with suppliers.

A category manager interviewed suggested that the benefits are more significant when there is a competitive tender at stake. When representatives from the supplier know they are in a dominant position, it is more challenging to get their team to be open and discuss detailed design. But he also highlighted that benefits on services can be significant. Indeed, the Design2Cost team performs analysis within the field on services. The map processes analyse them and spot improvement opportunities.

To summarise the benefits, the lab and its Design to Cost process provides new insights that can be used in commercial negotiations. Second, this offers a full understanding of designs and of their impact on cost and customer preferences, creating options for the company to choose what is essential and what is not critical from a customer standpoint. This allows for the value of every component to be challenged. For instance, the team in the lab highlight on a recurring basis that black paint on a board that nobody will see is “vanity not value”. Finally, this allows them to have in-depth, fact-based conversations about the opportunities to jointly optimise the end-to-end value chain with Vodafone’s partners. This is key to opening the door to more collaborative and innovative relationships with suppliers.

How was the change implemented?

Implementing such an initiative requires a change in mind-set. On the supplier side, not all were open to discuss and share their costs. The CPO and CTO of Vodafone sent a letter to all key suppliers’ Customer Account Teams and CEOs, explaining to them that the Design2Cost Lab had been established and that it would lead to new ways of working.  Also, from the start of a project with a supplier, an evaluation of their openness is done. This is important to create the right relationships and forms part of tender evaluation criteria. The team has seen some suppliers coming to them with services or products where they lose money and offering to work together to see how this could be improved. New entrants in the market often want to work with VPC’s Design2Cost team and are open to discuss cost. Incumbent suppliers are more reluctant, but in a competitive context or when the process brings real tangible opportunities for both sides, progress has been made. Perseverance and consistency are essential here.

On the internal side, a lot of time has been invested in explaining the benefits and the way it works. This is perceived as an opportunity for procurement teams to be empowered with more information and knowledge, so it really helps them make progress with some vendors.

Business statistics concept.

Looking into the future

As the team progressed, it realised that additional benefits could be unearthed from the data collected and the assets they had created. For instance, some of the data can be used to map sources of components and react to any disruptions or bans. There is also the possibility to offer a Purpose-led Design2Cost capability. This means identifying impacts on the environment and society as part of the exercise so further feedback can be provided to suppliers. This can help identify if refurbishing a product is an option or not, can help eliminate plastic and serve as a good basis to understand CO2 emissions. It can be used to make the necessary total cost calculations for implementing circularity. Finally, one option for the ream is also to monetize its capabilities outside of Vodafone to other telecom operators who cannot invest in such a lab. So, the future is looking great for the Vodafone’s Design2Cost Lab.

All this demonstrates that technology often makes an organization more efficient thanks to automation. But when technology augments the work of the best professionals then the benefits are huge. This requires maturity, and for many organisations, further investigations of their investments.

About the Authors

Hervé Legenvre

Hervé Legenvre is Professor and Research Director at EIPM, an Education and Training Institute for Purchasing and Supply Management. He manages educational programmes for global clients, conducts researches and teaches on innovation and purchasing transformation. Hervé Is the author of the book “Fifth Generation Purchasing”.

Gavin Hodgson

Gavin Hodgson manages Vodafone’s Hardware teardown lab. He has two decades of global experience in procurement and supply chain management, and holds degrees from Cambridge University, and KEDGE. He is currently working on solutions to assess hardware CO2 footprints by using teardown insights.

Govind Khandelwal

Govind Khandelwal is Head of Core, Software & Transmission Technology Procurement and Design2Cost Lab at Vodafone. Govind is a Strategic and business focused senior international telecommunication industry executive with 20+ years of Supply Chain. He has International work experience in India and Europe.


Please enter your comment!
Please enter your name here