Patent Application
20090248488
The invention relates to integration of risk assessment and forecasting in a supply chain such as for services and/or goods, and in particular, but not exclusively to enabling a user to build sensible scenarios especially, but not exclusively, through a suitable user interface with a computer system adapted to determine appropriate risk forecasts.
Best-in-class companies require a responsive and resilient supply chain to withstand against major disruptions occurring in the business environment. Globalization has increased the supply chain complexity. Disruption in one geography can lead to disruption of the entire supply chain. It is imperative now to have comprehensive enterprise-wide risk management practice. While Risk Management practices have matured in finance markets, they still remain in a nascent stage in supply chains.
Supply Chain disruptions are difficult to forecast. Events like natural disaster or 9/11 cannot be predicted by most sophisticated forecasting engines. However, these events have a very low probability of occurrence. Most of the events impacting supply chain planning are planned (adaptive) or unplanned (reactive). Risk coming out of planned interruption (sales promotion, plant shutdown) can be managed by accepting it and monitoring it, while risk of unplanned events can be controlled or mitigated. In each case, it is extremely important to understand the extent to which it could cause the disruption.
No two events are of the same nature. The context changes with the factors of time, geography, product and person. Macro-economic parameters, geo-political conditions, different regulatory environments, culture and strong human involvement provides multiple combinations of the same events to be managed. Organizations are exposed in different business environments and risk assessment has to take into account the local context. More often than not, business managers are aware of the impact of the event through their past experience, however, they find it difficult to apply that knowledge consistently in the planning. Also, with the exit of experienced personnel their knowledge also goes out of the organization. In the absence of a formal risk assessment process, which enables users to assess the probability and impact of events, risk assessment remains gut feel speculation at best. This results in organization either being vulnerable to residual risk (unmitigated risk) or spending far more money on risk hedging and losing opportunities due to an over cautious approach.
Forecasting science provides strong theoretical approaches to capture the impact of demand and supply drivers on the forecast. However, business users find it difficult to understand and interpret these complex models, limiting their potential of providing best analysis through rigor of mathematics and statistics. Complexity of these models also makes it difficult for the users to completely trust and accept the output and at times, even pushes them to make their own decisions. Lack of ownership in formal Forecasting & Demand Planning process can create ripples in an entire supply chain and makes it susceptible. Scientific modelling captures the historical data and derives intelligence from the past patterns. But in the absence of a history of calendarized events, their performance would be compromised. It is impractical to design different models for different geographies and businesses. Moreover, analysis provided by these models is retrospective—looking at the past, while business requires prospective analysis of the events—looking at the future, with the capability of running “what-if” scenarios for different combinations.
Risk analysis carried out in isolation to the capacity view, will not yield any conclusive results. Hence, it is important to consider events, which not only impacts the demand but also the supply conditions. In modelling environment, this poses further challenges and increases the complexity.
US 2007/0208600A1 for example provides a computerised system for pre-emptive operational risk management and risk discovery and even to forecast the nature of these risks over time in future, but the methodology remains quite constrained and inapplicable for practical implementation in a supply chain. In particular this prior art document does not teach scenario building, nor does it consider demand or supply scenarios, nor the impact of inter-relationship over time of multiple events for example.
An aspect of the invention provides a supply chain forecasting system enabling integrated risk assessment and forecasting of the impact of events on a supply chain comprising a processor adapted to receive, process and output data for a forecast, and a user interface adapted to enable a user to input parameters to facilitate processing of a forecast, wherein the user interface provides a user the ability to select an event which is likely to have an impact on a supply chain and to view a visual representation of different types of impact over time of an event thereby enabling a user readily to understand the nature of the impact of an event on a supply chain and hence to select the most appropriate type of impact over time for an event.
Beneficially a user such as a local manager involved in operating a supply chain can interact intuitively with the forecasting system using knowledge of a type of event which is likely to arise and the nature of that event and hence its impact over time for the local supply chain.
In a global business environment, organizations are impacted by a number of events in the external environment. To cope with the risk due to such events and to mitigate the impact of it, it is imperative to have true assessment of risk organizations are posed with. An empirical approach is used here, which captures the experience and local environmental understanding of the planner, different probabilities of the event occurrence with different magnitude stretching for different time durations, and provides aggregated picture through a Risk Index.
After carefully assessing the challenges in designing and implementing statistically rigorous models, an effective and implementable approach has been designed and described here. The primary benefits of the system are—
1. Risk assessment process is easy to interpret for the users.
2. The local environment context and the experience of planners (bootstrapping) is captured and hence, provide consistency in decision making.
3. The framework takes into account
4. The system also takes into account, both demand and capacity views.
5. Classification of events is based on Impact and Probability to apply correct risk management practice.
A Risk Assessment framework is used as a compliment to the statistical forecasting engine, in stead of a substitute. Sophisticated time series models captures pattern in historical data and generates the base forecast. Changes in business environment and macro-economic factors happen over a long period of time. Impact of these changes can be captured by statistical models as they span over longer time. However, events which are unexpected or impacts the planning horizon in short term, cannot be captured by such models. The Risk Assessment Framework, hence, focuses on impacts of the events for which historical pattern is not available. We could enhance the base forecast accuracy by integrating this simulated risk assessment approach with the statistical forecast using ‘what if’ scenarios. Users can apply their own understanding and experience in generating simulation scenarios, and the output could be used in meaningfully modifying the base forecast. This also enables users to understand the individual and cumulative impact of various events on the planning. A further aspect of the invention provides a supply chain forecasting system enabling integrated risk assessment and forecasting of impact of events on a supply chain comprising a processor adapted to receive process and output data for a forecast and a user interface adapted to enable a user to input parameters to facilitate processing of a forecast, the system comprising a catalogue of pre-defined events which might be influential in the forecast, each event having associated settings for at least one of risk profile, impact, probability, and impact versus time characteristics.
A yet further aspect of the invention provides a supply chain forecasting system enabling integrated risk assessment and forecasting of impact of events on a supply chain comprising a processor adapted to receive process and output data for a forecast and a user interface adapted to enable a user to input parameters to facilitate processing of a forecast, the system being adapted to provide a user interface which enables a user separately to select events according to their likelihood of impact on one of the demand side and the supply side of a supply chain.
Another aspect of the invention provides a supply chain forecasting system enabling integrated risk assessment and forecasting of the impact of events ion a supply chain comprising a processor adapted to receive, process and output data for a forecast, and a user interface adapted to enable a user to input parameters to facilitate processing of a forecast, wherein the user interface provides a user the ability to select an events which is likely to have an impact on an aspect of a supply chain and to create a scenario comprising a number of such events wherein the processor is adapted to combine the effects of the events selected for a scenario in order to determine an overall impact of the events in a forecast, the combination of impacts being weighted according to a predetermined conversion of the sum of the impact values (risk indices) for the individual events.
Another aspect of the invention provides a method of providing an integrated risk assessment and forecast of the impact of events on a supply chain comprising the steps of cataloguing events having a potential impact on a supply chain, enabling a user to build a scenario of selected events, analysing the risk and impact of the events in the scenario, and integrating risk over time of the events to provide a forecast of the behaviour of the supply chain in the scenario.
Further aspects of the invention includes methods of creating supply chain forecasts comprising certain of the steps set out in the following description and as defined in the appended claims. Moreover, further aspects include computer program products comprising instructional data which when implemented by a computer system enable the methods according to the invention, and a yet further aspect includes user interfaces comprising a visual display to enable a user to create a supply chain forecast. These aspects, steps and features of the invention are apparent from the following description of an embodiment of the invention and the appended definitions of the invention set out in the claims. Moreover it should be realised that each feature or step of the invention is combinable in any combination with any other step, steps, feature or features of the invention thereby beneficially to provide novel combinations according to the invention.
An embodiment of the invention will now be described by way of example only with reference to the following drawings, in which:
Referring to
Computer 14 comprises a processor 22, or central processing unit which can be in the form of a microprocessor; a data storage device 24 or memory, which can include a range of devices including volatile and non-volatile storage units including RAM, registers, cache memory and/or mass storage devices such as hard drives.
Accordingly, computer system 12 in particularly comprising display 18, is able to provide a user with a graphical user interface thereby to enable a user to interact with a program running on processor 22.
Computer 14 further comprises input and output ports and devices 26 enabling communication outside the computer system 12 for example through use of a software product 28, such as a CD Rom or other such device e.g. a flash memory, comprising data and/or software code (instructional data) which is insertable in a suitable drive forming part of the input/output 26 of computer 14. Similarly, a suitable output connection is provided between input/output 26 and a network 32. Network 32 might comprise one or more servers, provide a local area network and/or might suitably, for example with appropriate fire walling, enable access of computer system 12 to the internet and/or specific extranet having one or more devices and/or computers. In
Here, the processor 22 is a device capable of executing computer programs and in particular of receiving, processing and outputting data for use in integrated risk analysis and forecasting on a supply chain. Such as known processors or central processing units for a computer include microprocessors available from many manufacturers such as an Intel (trademark) and Motorola (trademark). Accordingly processor 22 is suitably programmed or programmable to enable the risk analysis and forecasting process described herein to be performed.
The user interface comprises sufficient elements or devices to enable a user readily to interact with the computer system 12 thereby to interact with the processor 22 and hence to influence the risk assessment and forecasting process, including the input, processing and output of data in that process. Such electronic devices can comprise output devices including a visual output such as a display, and or a printer, and an audio output such as a speaker; input devices such as a keyboard and or a mouse; and/or an interactive display such as a inductive touch screen display. This list is not exhaustive.
Referring to
The process 60 further comprises the steps of cataloguing of defined events, their impact, magnitude of impact and other characteristics as shown at step 66. An example of such a catalogue is shown in
A user, such as a local manager is able to interface with process 60 as shown at step 68 to effect a local planning function. At step 70 in the process, the user builds a scenario by defining probability, start date and duration periods for events which are likely to impact on a supply chain. At step 72, the computer system 10 and in particular local computer system 12 is able to determine risk index and quadrant information (as described later) and to output this to the user as shown at step 74 thereby to enable a user to create mitigation strategies. At step 76, of process 60, the risk index can be integrated with a base forecast using an engine forecast 78 thereby to provide an output at step 82 for review by a user. The output can be reviewed at step 80 for demand planning and/or to iterate the process to better refine the risk assessment and forecast of impact on a supply chain as shown by the return at step 84 to the cataloguing step 66.
In order to achieve process 60 described in relation to
Referring to
Beneficially, each event has certain default characteristics as shown in
Impact Weights are beneficially expressed in ordinary language terms but have associated mathematical values such as a percentage or value within a predefined range such as: High—100, Medium—66, Low—33. These weights are positive for inflationary events and negative for deflationary events. If a user thinks an event will have no impact it can simply be deselected from the scenario. The quantitative weight value attached to “High”, “Medium”, “Low” can be changed as required. Ideally, these values should be defined by historical analysis to see the impact of some of the events on demand. Additionally it is possible to use linear (1,2,3 or 2,4,6 or 33,66,100 etc.) or geometrical (1,3,9 or 2,8,64 etc.) or any other relationship, which can best describe the difference between “High” and “medium” and “low” impact.
In one version, display panel 90 automatically displays events selectable from one of three categories one of demand impact, supply impact or shock. Accordingly, the user is able to define impact direction (see
Referring to
Similarly, the default impact characteristic 100 can be modified by a user via display panel 90.
Further user tool buttons 106 are provided in display panel 90 including the option to save, add (an event), create a scenario, and refresh.
Assuming that a user selects the “create scenario” button 106, then display panel 90 is slightly modified as shown in
The user is required to select only two of the start date 114, end date 118 and time period 118, and the system automatically then selects the third of the three time characteristics for each event.
A user is able through display panel 90′ to define a probability of recurrence of each event in the scenario, thereby enabling the user to take advantage of any local knowledge under the selected filter parameters and in particular local geographic factors. Referring again to
Beneficially, the probability of occurrence of an event is displayed to the user everyday or in easily understandable terms such as certainty, almost certainly, maybe, maybe not, and bleak possibility for example.
The phrases “Certainly”, “Almost certainly” etc. could be change as per business user requirement. Also, the quantitative probability value attached to it. Hence, this remains flexible to suit a typical business requirement.
Accordingly, the local computer system 12 is able to determine based on the input probability 120 of the likely recurrence of an event and hence use this in the risk assessment analysis. A drop down panel 122 is provided in order to help the user to select the most appropriate term under expression of probability based on local experience.
Additionally, a user is able to determine the impact on demand 124 of any event as shown in the column labeled “impact on demand” 124. Beneficially, the impact on demand is labeled with a simple alpha/numeric label and this can form part of a standard characteristic of an event as held in the default parameters (albeit not shown in
Referring to
Accordingly, with each of the
Beneficially, the user merely has to select the shape of the impact versus time whereafter the system 12 determines whether or not the primary function f(t) or its inverse f′(t) is selected depending on whether the event has an inflationary or deflationary risk profile (and/or is demand side impact for the scenario as discussed). And moreover, calculates the impact over the selected time for the specific event, that is between the start and end date selected in columns 114 and 116 shown in display panel 90′. As shown in
Further types of impact function over time are shown in
a F(t)=Ae−xt, F′(t)=1−Ae−xt
b F(t)=1, F′(t)=−1
c F(t)=t, F′(t)=1−t
d F(t)=At2, F′(t)=A(1−t2)
e F(t)=1/4(t−1/2)2, F′(t)=1−1/4(t−1/2)2
f F(t)=4(t−1/2)2,F′(f)=1−[4(t−1/2)2]
Based on these characteristics, the local computer system 12 is able to determine the probability of impact and weight of impact of an event over time as set out below.
Event Probability—Certainly (i.e. 1.0)
Event Type—Inflationary
Impact Weights—High (i.e. 100)
Impact Duration—15 days
Impact Profile—A (exponential decrease Yt=Yo. e−kt) where Yo=Probability×Impact Weight=100
Impact on Day 1=100.e−k x (1), where K is a decay constant, calculated for decaying of initial impact of 100 to end impact of 1. In this case, value of K is −0.3070 Hence,
Yt
1=100.e−(−0.3070)×(1), where
Yt
2=100.e−(−0.3070)×(2),
Yt
3=100.e−(−0.3070)×(3),
Yt
10=100.e−(−0.3070)×(10).
Here an impact value or risk index of 118 is calculated for day 1, 49 for day 2, 8 for day 3 and 1 for day ten using the above equations.
Having built a scenario comprising 8 different events each having defined start end date, probability, impact and risk profile, the system is able to calculate risk elements for each event over the selected time period as shown in
The overall risk index, being the sum of the total columns in the tables E of
In one embodiment the proposed forecast in the final column of the table in
The user is able to view the overall risk index shown in the first column of the table in
The manager is also able to make use of the risk quadrant which can be displayed as shown in display panel 166 in
Referring to
Finally, the user is able to view the impact of the scenario on a base forecast. Beneficially, a display panel 166 is presented to the user comprising the engine (or base) forecast, the proposed forecast and overall risk index. Hence the user is able to view the overall risk index as shown also in
A Risk Assessment Framework is described below in terms of cataloguing events, scenario building, analysis, and integrating risk index with forecasting as now reiterated.
An important factor in the success of effective implementation of Risk Assessment Framework is cataloguing of events. The applicant has designed three categories to classify the events based on their impact:
Demand Side—Events which impacts the demand directly are put into this category. (For example, Heavy Rain, General Demand Decrease/Declining in the market, New Contract Signed, Existing Contract Cancelled, Seasonal Demand Increase, Seasonal Demand Decrease, Promotion, Discount Scheme, Advertising Campaign, Competitor's Price Reduction, New Product Cannibalization, Product Phase Out, Price Reduction etc.)
Supply Side—Events which impacts the supply directly are put into this category. (For example, New Entrants in the market, New contracts increasing the capacity, Temporary Decrease in overall capacity, Raw Material Price increase or decrease, Raw material supply increase or decrease, Logistical Improvements, Logistical Problems etc.)
Shock Events—Events which are sudden and cannot be categorized into above two category, shall be catalogued here. (For example, Company related positive or negative events, Macroeconomic surprises—Positive or negative, R&D News, Regulatory changes, Disasters/Calamities)
Distinction has to be made to include only information which has direct impact on either Demand or Supply. For example, rain could dampen the demand for cement so it should be included in demand category. But legislative change to attract FDI in cement industry will not have direct impact on demand or supply in short term. Competitor's price reduction will have an impact on demand, but its taking over another firm may not have direct impact on demand in short term. Catastrophic events must be put under shock category; which means that whenever this event happens again in future, their impact can be judged. Also, by notifying these events with dates alerts the forecasting engine, not to include those data, or refine them, before using in time-series. It is important to observe here that classification between Demand and Supply side events is not crystal clear always. For example, increase in raw material price can drive the manufacturing cost higher and due to increased price of the finished product, demand is reduced. However, the relationship between raw material price increase and demand reduction is not direct. In this case, raw material price increase is more of a Supply side event then the Demand. Regulatory events are sometimes treated as surprise. However, if their impact can be clearly recognized on either supply or demand side, they should be put into one of them. Above mentioned categorization is illustrative. Idea is to be able to classify events to understand their impact on Demand.
Each of the event should be ascribed with Inflationary (increases the demand) or Deflationary (decreases the demand) impact, further categorized into High, Medium and Low. Quantitative weights to these events remains a business decision. It could be applied with linear weights, or geometrical weights, or any suitable way to numerically capture the difference between high, medium and low impacting events. We will use the scale of 1 to 100 for the same, 100 being the high for Inflationary impact and correspondingly, −100 being high for Deflationary impact. Here, we also define the recurrence of the event, which can feed into Forecasting engine. Please refer to the
The cataloguing of events helps users to create and maintain events specific to their particular scenario. The same type of event could, for instance, have different impacts in different areas. In order to capture this, a two step customisation process is proposed. The first step captures the type and the general nature of an event. A typical example is a storm which results in a sudden demand (repair) increase that dies down over time. The user defines the event type “storm” and the general shape of its impact. A library of standard event impact shapes/functions should be provided for simply selection. This library could include many different functions, examples of which are given below.
Without loss of generality, only inflationary functions over [0,1]×[0.1] are given here. The second step sees the creation of a specific event. This is achieved by selecting an event type from step 1 and then specifying the event's actual start and end time, its expected severity/maximum impact, and the likelihood of the event happening. If these parameters are given as ts, te, A and p, then a generic impact function ƒ:[0,1]→[0,1] chosen in step 1 can be translated into a specific impact function {tilde over (ƒ)}: [ts, te]→[0, A] for step 2:
This equation has the characteristics that it is defined between the start and end time of the specific event and that the resulting impact values are constrained by the maximum impact. This formula represents the impact of an event without considering the event likelihood. This probability can be considered by including it as a factor:
F(t) allows the calculation of the expected impact of a single event over the planning horizon. Once such daily impact values Fi(t) are obtained for all specified events, a cumulative impact for time t can be calculated by
This sum defines the Risk Index.
Another important graph can be prepared here is called Risk Quadrant. There are 4 risk zones and based on impact and probability, each of the event would fall in to one of these four zones.
The Risk Quadrant helps in applying the appropriate risk mitigation option to each of them. The options are
These options can be correlated with the Risk Quadrant in the manner shown in
Integrating Risk Index with Forecasting
The Risk Index and cataloguing of events based on their impact and probability prepares organization to apply the right risk mitigation strategy and channelize the hedging effort. If we could integrate Risk Index with the forecast, it could provide users a true picture of variation in base forecast for each simulated scenario. For the given value of Risk Index, how much impact shall be taken to modify the statistical forecast, can be decided by business experience or by solving a small optimization problem to increase MAPE. For example, if the value of Risk Index is between −1 to −5, the statistical forecast will be pulled down by 5% and if SI is between +1 to +5, the statistical forecast will be pulled dup by 5%. Variation in these weights would allow to modify, and not replace, the base forecast. However, for the large value of Risk Index, modification in forecast could be as large as 100%.
The process maintains the essence of carrying out statistical forecast through complex forecasting models, and bootstrapping the experience of business users in understanding future events and their impacts. The former is retrospective while the later is prospective. The right combination of the two helps forecasters in making consistent decisions in planning. Below is one of such simulated scenario. The graph shows actual volume against both, statistical forecast and proposed forecast using Risk Index. The graph shows the improvement in Forecast direction and MAPE by 7%. The mean value of adjusted forecast is also much closer to the actual mean.
The framework is flexible to adopt different requirements of different business and geographies. While its flexibility and ability to capture and apply human experience is a strength, that is also its weakness too. Defining weightages to the event impacts and integrating Risk Index with the statistical forecast, are two important areas and requires users to try few combinations before reaching the right one. Sensitivity analysis of events on the forecast could come very handy, if such history is available.
Although the present invention has been described in connection with various exemplary embodiments, those of ordinary skill in the art will understand that many modifications can be made thereto within the spirit of the invention and the scope of the claims that follow. Accordingly, it is not intended that the scope of the invention in any way be limited by the above description, but instead be determined by reference to the claims that follow.
