Inventory Optimization & Balancing – Industry Use Case: Airline’s Catering Rotables
Level 1: Easy; 15 minutes read
Audience: Subject Matter Experts
Author: Ramandeep Goyal, SAP S/4HANA RIG EMEA
This blog is to share the details of a PoC built for an airline during an innovation project. The complete solution is built using S/4HANA, UI5, SCP and Leonardo Machine Learning. The key modules used from S/4HANA are PM (Work order as a key object for integrated planning and cost collection), MM (Material Master, Inventory management, Purchasing, Service entry sheets) and FICO. User interface of the application is built on UI5 using design thinking methodology.
Definition of the problem
Aircraft cabin operations require items, known as rotables, on top of consumables, food and beverages. Such items are cutlery, trays, glasses, duvets, pillows.
The aircraft is loaded with required stock at the hub for various airlines (Example – Frankfurt, Atlanta, Dubai). Some of the items are used in the flight, some of the items are broken (while handling or while using), some of the items get lost and some are unused.
At destination (outstation), all the items (including the unused items) are unloaded and then processed (washed and repackaged) to be reused on a later return flight. The aircraft is loaded with fresh stock of such items for the return flight as part of the turn-around operations. They are handled by the local caterer.
As the flights are unevenly occupied, there is a potential rotable stock unbalance situation. If the flights from Hub (Example – Frankfurt, Atlanta, Dubai) are less occupied than the return flights, the tendency of the stock is to diminish. If the flights to Hub (Example – Frankfurt, Atlanta, Dubai) are less occupied, the tendency of the stock is to accumulate. To balance the stock, the current approach is that, for each flight, the amount of rotables on board on each segment is equal to the amount needed to accommodate the maximum number of passengers, on the direct-return flight pair. The approach keeps the stocks balanced, but costs and pollution is not optimized (mostly fuel consumption and environment impact due to washing unused items).
Objective of the solution –
The objective of the solution is to optimize the stock level and movements of the rotable items in order to reduce costs and environmental impact. The model will consider the constraints (like minimal safety stock, have at least the minimum stocks on board) and multiple flights in the optimized flight period (day, week etc.)
Customer ABC develops a capability to balance inventory of rotables across the network in an automated way end-2-end, i.e. from demand planning (requirements prediction) all the way to stock transport orders and recording into the Flight Fulfilment Order (FFO). FFO is the key object used for integrated planning of various objects required for a flight. Examples – Rotables, Meals, Crew etc. In the PoC, FFO is built using work order in S/4HANA. This contain the information about various resources required for a flight and the associated cost. The header of the FFO which contains information about the flight is built in SCP.
The use case in Customer ABC’s PoC was for catering rotables but the same principle is applicable to any type of rotables that are transported across the Customer ABC’s network, e.g. ULDs, meal carts etc. The main objective of this work is:
a) automate an important and very costly part of the catering supply chain using SCP developments (SAP Cloud Platform)
b) potentially and eventually introduce uplift and inventory holding, and handling costs benefits across the network
Prerequisites and Assumptions:
• The standard S/4HANA process in Materials management & Plant maintenance will be used with some customization development for Customer ABC.
• Meals forecasts are assumed to be available from Machine Leaning solution through a developed API at least at D-24 and D-6 for each flight.
• Passenger forecasts are also assumed to be available real time.
• The current process of rotables uplift to the 100% of cabin capacity irrespective of passenger loads is not efficient and not sustainable.
• An inventory planner will always be needed to handle exceptions and the user interface (UI) is built from the perspective of that role.
Description of Model Features:
The business process flow envisioned for this this PoC is shown in the following process map:
The idea is to use various data sources to solicit all data required for determining the right amount of rotables to be uplifted on every flight each day, so that the following two business objectives are met:
a) uplift enough rotables to service the forecasted passenger load
b) uplift the appropriate additional rotables for ensuring a balanced stock of rotables between Hub and all outstation caterer stock holding locations, where balance” is determined against a pre-determined and agreed with the caterer minimum stock level to be held at that outstation
The inventory balancing algorithm (SAP cloud platform) determines the right uplift amount for each flight by considering multiple parameters such as:
a) Passenger load forecast
b) Meal requirements forecast (for now, as a total across all menu items and in the future, if needed, by menu item category)
c) Aircraft configuration
d) Forecasted stock deficit if no balancing occurs
e) User-defined or auto-generated weightage factors for each SKU
f) Business-defined conversion factors depending on the length of the particular trip etc.
Once the appropriate rotables uplift level for each flight is determined, the PoC solution built provides the capability to communicate this to:
i) the Flight Fulfilment Order for the appropriate purchase orders and financial accounting thereafter and
ii) the caterer operations teams as a stock transport order for transporting the requested SKUs from stock area (reserve stock) to workflow (floating stock onboard aircraft). Data from SCP is pushed into S/4HANA by using standard webservices already available.
Quantities of the rotables to be processed are pushed into FFO which becomes the basis to create service entry sheet to pay for processing charges as per the actual quantities.
Based upon the same quantities, STO (stock transfer order) is also created to transfer the stock from the hub to aircraft & then outstation. This gives you the most updated status on the inventory levels at the Hub and outstation as well.
Key Lessons Learned and conclusions:
The initial solution built in PoC reflects data only for 3 selected outstations:
a) LHR which has 6 flights per day,
b) BCN which has 2 flights per day and
c) ORD which has 1 daily flight and is a ULR destination on the Customer ABC’s network.
Solution built –
The UI built for a future inventory planner is shown in the following screenshots. A home page lists all the outstations that the inventory planner has purview of and also highlights critical inventory issues as well as an overall outstation stock status for a cursory review.
The inventory planner can then select a date range which always start “today” and ends at a desired period length (e.g. one week later). Then, he / she can select on outstation at a time to view the detailed rotables stock of and take decisions on, such an example for one week of LHR data is seen in the screenshot below:
The planner can select a rotables category (e.g. plates) and see a high-level overview of all SKUs under that category at the top part of the UI.
The bottom part of the UI shows for each selected date, the detailed uplift solution of the inventory balancing algorithm for a pair of DXB-outbound and DXB-inbound flight to and from that outstation (e.g. DXB-LHR and LHR-DXB) operated by the same aircraft tail. The detailed view of the uplift for each flight includes all SKUs and for each of those, it depicts:
a) “Full Load” option: The uplift as per current process, i.e. loading at capacity regardless of passenger volumes
b) “Minimum” option: the uplift required ONLY for passenger servicing
c) “Balanced” option: the uplift required for passenger servicing AND for stock level balancing at a pre-determined stock level
In the screenshot above, the “minimum” option is highlighted in red colour as it is to be avoided and the “balanced” option is highlighted in green as it is the one that the planner should follow in order to achieve stock level balancing. The difference in stock level impact between following “minimum” vs. “balanced” option can clearly be seen for any specific SKU (e.g. casserole F/J) in the graphing capability of the UI in next screenshot.
In this graph, the planner can follow the forecasted evolution of stock levels of that particular SKU in the days to come starting today (may require left-right scrolling). For each date in the X-axis, the uplift for all inbound and outbound flights between DXB and that outstation is shown in bar format and above that there are two-line graphs. If “minimum” option is followed by the planner, the stock level evolution is represented by the black line and in this particular case it shows that over 4-5 days int eh future, there will be a surplus of stock building up at the outstation if we follow this option, i.e. if we do NO balancing at all. On the other hand, if the planner follows the “balanced” option, which is represented by the red
line in the graph, stock levels of that SKU at the time of each arrival and each departure in and out of the outstation will be more balanced and therefore closer to the dotted line of agreed minimum stock level to be maintained at that stock location.
The inventory planner has to review and approve / publish the inventory balancing algorithmic solution, but this approval process can also be automated in the future based on an agreed approval workflow. Using this UI, before approving and publishing, the inventory planner has the ability to amend the uplift level dictated by the inventory balancing algorithm solution, should he / she deem appropriate to do so based on inventory information that he / she has that the algorithm did not have at the time of solving.
The results shown for LHR in the above screenshot are actual balanced results produced by the developed algorithm which shows that the algorithm actually assists in maintaining the desired balance.
Business impact –
By implementing this solution, following optimization flexibility is expected to be achieved:
– Reduce return flight loading (less occupied return flights than direct flights) and stock on outstation for future more crowded return flights
– Reduce total amount of rotables in the system (reduces stock on outstations and base). Overall inventory footprint in the network is reduced.
– Less number of rotables on a flight will result in less wear & tear and reduced handling cost.
For verifying the improvements, the following KPIs will be calculated for both historical and predicted/optimized outputs:
– Rotables induced fuel consumption per passenger x class / km, on flight, outstation roundtrips and fleet level, aggregated on day, week and month
– Estimated volumes of unused rotables processed and flown per passenger x class / km, on flight, outstation and fleet level, aggregated on day, week and month
– The above KPIs quantified as costs
I hope that you enjoyed the blog and got some insights on a common problem in the airline industry. PoC was highly appreciated by the customer and internal LoB. This is a fit to use solution to optimize the quantities of the rotables to be used on a flight. A demo has been prepared and placed on the demo store to share the experience with other customers.
Additional Information –
SAP S/4HANA Discovery
• SAP Cloud Platform – https://cloudplatform.sap.com/index.html
• SAP Leonardo Machine Learning – https://www.sap.com/mena/products/leonardo/machine-learning.html
• SAP design thinking – https://design.sap.com/designthinking.html
• SAP user experience – https://www.sap.com/mena/products/fiori.html
• SAP Demo store – https://sapdemostore.com/sap/bc/ui5_ui5/sap/yunifiedstore/index.html
• SAP help.sap.com – S/4HANA 1809 , S/4HANA Cloud
• SAP Fiori Library – https://fioriappslibrary.hana.ondemand.com/sap/fix/externalViewer/