Machine Learning: Key to JTI’s approach to Sales Forecasting

John Ham, End-to-End Supply Chain Vice President at JTI
As companies everywhere use AI to speed up processes and improve accuracy, John Ham, End-to-End Supply Chain VP at JTI, puts forward his recipe for success

When using artificial intelligence (AI) or machine learning (ML), there can be a tendency to focus on the solution instead of the end result. What’s most important is to be clear on the destination, regardless of whether it’s generated through AI or ML.

In JTI’s case, ML was the chosen solution for the better result.

“We took the decision to design a wholly JTI-owned solution for our sales forecasting,” explains John Ham, End-to-End Supply Chain Vice President at JTI.  

“To build our ML engine, we brought together three groups of experts: internal top talent from within demand planning; our own internal data science team, which analyses data across the business; and external data science and forecasting experts from Accenture’s Supply Chain Innovation Center.

"Together, they built and tested the ML engine and we made it our own intellectual property.”

Transparency and trust

John continues: “Given how important trust and transparency were to the success and adoption of this new way of working, building the engine together and understanding how it works was crucial.

“It wasn’t some mystery machine doing it for you. We were in full control every step of the way.”

Another way JTI invited collaboration and gained trust in the process was its conscious mindset of putting people first – before technology. 

“We communicated, communicated, communicated – and I’m not exaggerating,” says John.

“We explained what we were doing and why, and reassured the general managers and CFOs that they still owned their numbers. We gave them forecast numbers generated from our ML engine, but empowered them with the full autonomy to review, enrich and sign off. It was huge in building trust.” 

John expands: "With AI or ML, it’s easy to quickly focus on technology and algorithms, but it is vital for the success of the project to make sure people with the right capabilities and mindset are engaged. This is when you gain momentum. The human component cannot be overlooked.”

It took JTI and its external team approximately nine months to build the engine, before testing took place in Switzerland – one of the company’s strongest and most rigorous markets – which believed its own forecasting accuracy couldn’t be beaten. 

“Guess what? Our ML did beat them," John exclaims. "It wasn’t a competition, but it was a bit of fun. The exercise gave us the confidence to say that, if we can do it successfully in Switzerland, with already a very strong process, then in markets where there’s less capability and poorer results we’re going to get even stronger benefits."

Rollout across 19 country clusters, representing approximately 75% of JTI’s revenue, took one year. Lower volume and more complex markets – accounting for the remaining 25% – will take another year. 

Success factors

John credits the aforementioned success to JTI’s ruthless focus on working towards two objectives: more measurable forecasting and forecasting accuracy with less effort.

He adds that owning the IP of the ML engine, making the entire process transparent from start to finish, and listening and involving employees and partners have added to a win-win situation. 

Besides the successful rollout, the role and importance of a demand planner has been elevated at JTI. Less time will be spent on Excel spreadsheets and number crunching, with more time devoted to driving business and sales.  

“We place a great deal of importance on our workplace culture and opportunities for our employees to grow and develop,” John goes on. "The evolving role of demand planners – increasingly more valuable – is just one example.” 

In early 2024, JTI was recognised for the 10th year in a row as a Global Top Employer in 47 countries covering five regions. 

Concluding with one final piece of practical advice, he adds: "Don’t rush the process; use ML methodically and carefully. You get one chance to convince people that this is the future.

"If there is push back to go faster, instead focus on your success criteria as these are essential parts of the journey. So many AI initiatives become vanity projects, instead of real business solutions.”

******

Check out the latest edition of Supply Chain Magazine and sign up to our global conference series – Procurement and SupplyChain LIVE 2024

******

Supply Chain Digital is a BizClik brand.


Share

Featured Articles

Mars Snacking's Amanda Davies on Sustainability at P&SC LIVE

Amanda Davies, Chief R&D, Procurement and Sustainability Officer at Mars Snacking, discusses sustainable futures

DP World: Delivering the ICC Men’s T20 Cricket World Cup

Logistics giant DP World is among the companies playing a key role in the delivery of the ICC Men’s T20 Cricket World Cup hosted by the USA and West Indies

FuturMaster: Unlocking Untapped Potential in Supply Networks

FuturMaster, a pioneer in supply chain planning solutions, has launched the Network Insight Graph in a bid to unlock untapped potential in supply networks

Why you Should Automate your Supply Chain Analytics

Digital Supply Chain

P&SC LIVE New York welcomes Amanda Davies, Mars Snacking

Sustainability

P&SC LIVE New York welcomes Kirsten Loegering, ServiceNow

Operations