Description of a business game used to demonstrate a hybrid business simulation approach. Uses by discrete event simulation to model the operations of a pub, and a Agent Based Modelling approach to model a population of customers.
Game can be played by up to 5 teams
1. Build your own pub!
Kevis Pachos & David Buxton
kevis.pachos@decisionLab.co.uk
david.buxton@decisionlab.co.uk
Young OR 18, Exeter University
10th April 2013
2. This Objective
• Win 2 bottles of cheap fizz!
• Get an overview of simulation
• Take a quick look at AnyLogic as a simulation tool
• Get some understanding regarding the different
approaches to simulation
– Discrete event for the business processes
– (System Dynamics)
– Agent-Based for the business strategy
• See how a business game can be used to bring learning
into an organisation
3. About Decision Lab
• Operational Research (OR) consultancy
• Business set-up
– 5 permanent OR modellers
– Plus 3 to 5 technical associates
– Geographically spread
– London office
4. What are we going to do?
• Build a process model of a pub in AL
– Test and learn to understand how our operations perform
• Bring the pub models (the ones you’ve developed) into a broader
game
– to understand how the drinking strategy of a town might play out
• Imagine that..
– The town planners want to know how many pubs the town can
support
– The council want to know if you can create a happy society
– The brewery want to know if it can make money
• (and if you’ll be a good strategic decision maker)
5. What you’ll be doing
• ‘Right size’ the operations of the pub for a given number
of customers
• Using AnyLogic to build a simulation of the pub process
– Using the fundamental concept of discrete event simulation
• To do this, you’ll need to install & activate the AnyLogic
software
– Then follow the instructions put together by Dan
– Given the time constraints, Dan has given you a starting point
6. What you’ll be doing…cont’d (supporting
strategic DM)
• Decisions
– The town planners want to know how many pubs the town can support
– The council want to know if you can create a happy society
– The brewery want to know if the system can make money
• To achieve this we will extend using models using some agent based
thinking
• Playing the game
– In teams, Pub’s will compete against each other
• You will set the strategy for
– Pricing
– Capacity
» # of bar tenders (affect how quickly people get served)
» # of seats & # of standing positions (adds to the cost of running the business, but increases the comfort
level of the customer)
» # of spaces at the bar (affect’s the customer’s perception of ‘busyness’
• Strategy can be updated weekly
7. Why would you ever want to do Agent Based
Modelling?
• Most problems may now be described by the term ‘complex adaptive systems’ which
interact and evolve
– Our traditional, top Down (Reductionist) approaches are not sufficient to deal with these issues
– ABMS (using a constructivist philosophy) allows us to embrace the complexity
• Really good for looking at business strategy problems
– Generally include more about user perception of localised decision making
• Creates a closer 1 to 1 relationship between the real problem and abstracted model
• To the customer, feels like to more natural way to represent the problem
– Better client engagement
• Allows the capturing of more complex behaviours which influence system performance
• Allows the adoption of a rule based approach at a micro level
– Reduces reliance on probabilistic methods
– i.e. BDI
8. • But first, let’s build the pub
– Gain experience of using AnyLogic
– Introduction to process simulation (for those that
haven’t used it previously)
10. The rules
• The customer agent will decide who to buy from
– Where shall I go decision?
• Is it too costly?
• Is it too busy?
– Customer’s will remembers previous performance & can
select pubs as a favourite
– Customer will decide to have another drink, if they are
satisfied tonight
• If they aren’t then they will go to a competitor, unless everyone is
too busy, then they go home
• There’s about 800 customers per night to fight for…
11. What we know about running a pub..
• Fixed costs are a function of
– Seating capacity
– Standing capacity
– Bar spaces
• Variable costs
– Staffing (£75 per day)
• Cost of sales
– All drinks cost you £1 (you are free to set the resell price)
• Each week there’s a management meeting with the
brewery
– This is when you can change the configuration of the pub
12. What we know about the town..
• The town has a population of 10k, and
between 7 and 10% drink regularly at the
weekends
• During mid-week, only about 5% go out
• People tended to drink more on the weekends
13. What we know from the customer focus
groups..
• (remember ABMS is a bottom up modelling approach)
– (An agent based models can only reflect the behaviours and interactions that are incorporated into
the design
• Its interesting to do because of the interactions of the population over time, and the ‘live’ decision making
• This results in the system complexity and the emergence often talked about)
• About 50% of customer’s said that they sometimes choose a pub because of its
environment and comfy seats even though the prices might be higher
– But if the price is too high, then they’d rather drink down the road
• Nobody liked having to queue at the bar, all respondents said that they’d probably
look to go elsewhere if the pub was too busy
• Customers drink faster when standing up, but prefer to sit down when choosing
where to go
• Customer’s have their favourite pubs which they tend to go to first
14. Thank you for playing
For more information –
• www.decisionlab.co.uk
• www.anylogic.com
• http://www.theorsociety.com/Pages/Training/Courses/3238.aspx
Hinweis der Redaktion
We build models (mainly optimisation and simulation models)