Spatial Interaction Modelling Flashcards
What is Sij in the SIM?
The flows between i and j
i = origin/demand zones
j = destination/supply zones
What is Ai in the SIM?
A balancing factor that takes into account competition and ensures all demand is allocated to stores in the model
What is Oi in the SIM?
Demand - amount of expenditure in zone i
What is Wj in the SIM?
Store ‘attractiveness’ - normally size = attractiveness
What is exp-betaCij in the SIM?
Cij = cost of travelling between i and j - thought to be that greater distance means less attractive
Beta used for calibration
Why are SIMs more advantageous than buffers?
- Not confined to shopping within buffers, therefore take into account distance decay
- Can calculate market share by adding the sum of all flows
What are the 3 outputs of the model?
Revenue
Market Share
Flows from origins and destinations
Why is knowing potential revenue important?
Gives a benchmark - how does predicted performance compare with actual performance? i.e. overperforming/underperforming
Can also look at revenue per sq. ft. which is a good comparison for comparing the success of stores of different sizes
Why is knowing potential market share important?
Stores compete for market share, SIMS show market share at a smaller scale than regional level
BUT important to think of location e.g. 80% market share in a rural area is not the same as 80% market share in an urban area
Why is knowing the flows from origins to destinations important?
Good to know where customers are coming from
How do you calculate total revenue and revenue per sq. ft.?
Sum of flows coming in from every zone
Revenue per sq. ft.: divide by floorspace
What is the calculation for market share?
(revenue of retailer in zone)/(total revenue in zone) x 100
What can be included in ‘j’?
Shopping centres
Individual stores
Stores within shopping centres
80% studies use store size as the attractiveness indicator, but what else can be used?
Retail Fascia (Fj) - brand loyalty. Likely to combine Wj and Fj Price (Pj) - impacts on socioeconomic classes Site Characteristics (SCj)
What things can be included in site characteristics as an attractiveness indicator?
Parking (but likely to be associated with store size)
Key Multiples - stores nearby that are also needed
Ease of Access e.g. bus/car
Visibility - how visible stores are impacts on passing trade
Pedestrian Flows - in shopping centres etc.
Access to other centres
Why is measuring the success of stores within shopping centres more challenging?
Have to take into account shopping centre success as well as store success
What did Eyre (1998) create in order to take into account centre attractiveness?
Field Survey Attractiveness
Stores given a boost or deboost based on shopping centre success e.g. Meadowhall vs. Bradford SC
What is Fotheringham’s Competing Destination Model (CDM)?
Included an extra term Aj to measure ‘distance accessibility’ for zone j
Measures how close each retail outlet is to other retailers
Takes into account store competition
Higher the score, the more attractive the store
Why did Fotheringham make a CDM?
Argues that stores near other stores are more attractive than stores with nothing near them
The Aj term is increased if a store has more around it
Used in the SCj characteristic: access to other centres
What is distance decay?
When the interaction between origin and destination decreases as the distance between them increases
Why are SIMs better than GIS for measuring distance?
GIS uses x, y coordinates thus straight line distance but this is inaccurate
SIMS use a distance/time measure to create a non-linear but exponential relationship
Why is an exponential function better for measuring distance?
Controls the rate at which distance decay occurs
Takes into account variations between urban and rural areas
How are urban areas affected by distance?
Big relationship between distance and interaction (at the start of the exponential function)
If there are 2 Tesco’s a km apart, no one would travel the extra km for the 2nd Tesco, therefore everyone would go to the first
How are rural areas affected by distance?
Small relationship between distance and interaction (at the end of the exponential function)
Less provided for in rural areas, therefore if there was a Tesco 9km and one 10km, the extra km becomes negligible, therefore it is assumed nos. at each store will be similar
What is the function of beta?
Controls for the importance of distance
Why does distance vary between consumers?
Car ownership Amount of spare time Brand loyalty Rural vs. urba Geodemographics - socioeconomic class Product type i.e. elastic vs. inelastic products
How is distance usually measured?
Travel time
What did Birkin et al. (2010) say was wrong with using travel time as the distance indicator?
Assumes everyone has a car
Doesn’t take into account topography e.g. hilly areas would increase time
Flows between any demand zone and store will be proportional to:
- Store attractiveness in relation to attractiveness of competing stores
- Accessibility vs, accessibility of competing stores
Ultimately: people want to go to the most attractive and closest store
Why do we need to calibrate models?
To turn theoretical models into models that reflect reality; need to compare/match model predictions with real observed data
Can access observed data
Want model predictions to be within +/-10% of observed data - most stores want +/-5%
Where can observed data be accessed from?
Loyalty cards
Consumer surveys
What did Wood and Tasker (2008) say about model accuracy?
10% variation in sales forecast for a medium store can equate to £5mil. therefore argue for a variation of 5%
What do high beta values say about distance?
Distance is important - people willing to travel less
What do low beta values say about distance?
Distance less important - people willing to travel further
How does beta get altered in order to give the best representation?
If market share is too ‘flat’ then beta is RAISED
If market share is too ‘peaked’ then beta is LOWERED
How can geodemographics be used to also alter beta?
If an area has a pop. that is not affluent with low car ownership, then distance is important so beta is RAISED
If an area has a pop. that is wealthy with high car ownership, then distance is less important so beta is LOWERED
How do we check calibration is accurate?
Compare calibrated model predictions with actual observed data
Issue is that unlikely to have a matrix of all flows between each origin and destination, therefore becomes a generalisation - calibrate based on known flows for some consumers and try and ensure it is representative of all consumer
What is Average Trip Distance (ATD)?
Measures the average distance travelled by consumers
Use of ATD helps get hold of observed data and therefore be able to calibrate beta
If model can replicate observed ATD then:
- Consumer flows likely to be accurate
- Store catchments and inflow likely to represent reality
- Store revenue predictions should be accurate
Why do we disaggregate beta?
In reality people have their own beta values based on independent decision making, but can’t do this
Therefore, people are grouped based on shared characteristics based on where they live
What is the Standardised Root Mean Standard Error (R2)?
SRMSE used to assess model fit
0 = perfect model fit
Want to get R2 as close to 0 as possible
What are stores doing to try and increase attractiveness?
Tesco: max. floorspace by putting in concession points and other services e.g. cafes
Sainsbury’s: want to capitalise on their unused floorspace by taking over Argos. BUT Argos and Homebase come as a pair - don’t want Homebase
What is agglomeration?
Grouping non-food stores in retail centres is important as allows for economies of scale
Retail adjacencies can be crucial for attractiveness e.g. importance of anchor stores such as M and S in attracting customers to surrounding stores
What did Fotheringham (1986) say about hierarchical choices?
SIM assumes straight-forward decision making re. consumer choice in where to shop
He argues that the CDM is better than the SIM as it allows for hierarchical decision making: consumers pick a cluster of retail activity and then decide which stores to visit within the chosen cluster
What are economies of agglomeration?
Benefits firms gain by locating near each other. Associated with the idea of economies of scale and network effects
Why is agglomeration important?
- Stores located near each other may be more attractive
- Enables comparisons between stores
- Competing retail centres close to each other may be seen as one destination by consumers, therefore increasing custom
- Adjacencies to key anchor stores may also be important
What does alpha do?
Takes into account store attractiveness
How does alpha alter store size to alter attractiveness?
Makes a unit of floorspace more or less attractive based on a range of factors e.g. brand
Alpha >1 = more attractive as floorspace increases
Alpha
How can alpha be influenced by consumer characteristics?
Clarke et al. (2012) argue that consumer characteristics impact on brand e.g. affluent more likely to shop at Waitrose than Tesco
Therefore input ‘m’ (consumer characteristics) after the alpha
How does the model adjust for store location and size?
Acknowledges that not all affluent customers will go to Waitrose purely because of its brand image if it is small and far away
Therefore, the model recognises that some will pick the nearer and larger Tesco
How can we check for model effectiveness?
Minimise sum of squares
Want to increase the sum of squares
What is Birkin et al.’s (2010) ‘Goodness of Forecast’?
Predictive experiments using un-modelled stores
If confident actual trading will be within 1-% of predictions then directors likely to invest
Why might the model just not work?
- Demand misestimations
- Not counted all forms of demand e.g. workplace
- Not accounted for all existing supply
- Product choice important but not in model
- Multipurpose trips not accounted for
- Calibration data outdated and/or unrepresentative
- Mismodelled level of aggregation
- Misjudged impact of e-commerce etc.
What are ‘objective’ factors in store attractiveness? (BIRKIN ET AL., 2016)
Quantifiable factors impacting on store attractiveness e.g. parking facilities, retail adjacencies, relative prices etc.
What are ‘subjective’ factors in store attractiveness? (BIRKIN ET AL., 2016)
Consumers’ perception of brand image and customer service
