Predictive Tools

Assess, Predict and Improve Performance

It’s all about the future.

Spectrum Analysis Predictive Tools Services

Sales prediction modelling, neural network,
retail gravity, sales cannibalisation,
upgrade prioritisation reporting
and store scoring check chart.

Sales Prediction Modelling

Are you responsible for calculating sales predictions for multiple sites across Australia? 

Do you know how to fairly assess a site’s performance based on specific criteria for your industry or target market or against the performance of other current or future sites?

Our tailored Sales Prediction Models forecast site sales using a mathematical formula that we develop specifically for your enterprise and industry.

We create your Sales Prediction Model by analysing your facility and site location characteristics against the performance of your existing network. We identify the underlying factors (or drivers) responsible for sales performance and then build your tailored prediction model.

Your Sales Prediction Model accepts facility and site location characteristics as inputs and returns a sales performance prediction for one or more sites. Benefits include:

  • prioritising and assessing new site potential
  • the ability to assess the economics of site relocations
  • identifying possible sales cannibalisation impacts of a new site over existing sites
  • comparing actual versus potential sales performance for any existing site

We can use our six step data analysis process to help you:

  • explain how various factors can affect sales performance (for better site management)
  • estimate the sales performance of proposed sites for any location
  • select the most appropriate facility for a given site location
  • calculate the effects competitors sites have on sales performance
  • calculate the sales cannibalisation effect of opening a new site close to an existing site
  • provide scenario planning
  • avoid ‘gut feel’ decision-making
  • compare actual versus predicted performance and actual versus potential performance of existing sites
  • provide quadrant analysis to categorise sites as poor performers, under performers, good performers and excellent performers and plot the sites on a graph against predicted performance versus actual performance (a great visual representation)
  • identify and prioritise ‘new build’ sites with facts and data for evidence-based decision-making

Artificial Neural Network Modelling

Would you like to calculate sales predictions for multiple sites across Australia using the power of advanced analytical techniques?

Traditional sales predictions have typically used the ‘least square regression’ technique. This plots individual sales predictions as close as possible along a line.

Technology has helped develop Artificial Neural Networks that allow the algorithm to learn from historical data and train itself iteratively over time. 

Artificial Neural Networks can detect complex non-linear relationships between drivers and the dependent variable so it doesn’t just measure the performance of a site, it shows the interactions between the drivers of that specific site.

We can also incorporate the most relevant data for each existing site by visiting the site, taking photos and grading the site on key criteria including:

  • site-specific factors including size, position, access, visibility, traffic flows, appearance and local business generators
  • market factors including social and business demographics
  • competitive factors including distance to and size of local competition

We add this information into a database and then use the Artificial Neural Network Modelling to:

  • perform a correlation analysis to identify the key drivers that could be affecting sales performance and list these as variables
  • choose an Artificial Neural Network Design and develop Artificial Neural Network Models
  • select the Artificial Neural Network Model with the maximum level of accuracy for a reasonable set of variables to make future predictions
  • test the Artificial Neural Network Model to determine its predictive ability with either a dataset portion (if there is a large amount of data) or k-fold cross validation where you divide the data into k parts of approximately the same size and each time, leave out one of the parts for testing and the rest for model making so you can utilise all of the available data (if there is less total data)
  • use the Artificial Neural Network Model with the Spectrum Analysis scientifically proven Sensitivity Analysis System so that the values of the inputs can be varied around selected critical points to observe changes in the Artificial Neural Network Model output
  • calculate the contribution of the individual drivers towards actual sales

We can use our six step data analysis process to help you:

  • benefit from our neural network software with state-of-the-art neural model designs and various scientifically proven algorithms developed in-house including our Sensitivity Analysis System
  • secure more accurate performance forecasts for proposed site locations
  • examine the contribution of various factors on sales performance
  • understand the effect competitors have on sales performance
  • provide scenario planning
  • avoid ‘gut feel’ decision-making
  • compare actual versus predicted performance and actual versus potential performance of existing sites
  • identify and prioritise ‘new build’ sites based on the most aligned and non-linear drivers 

Retail Gravity Modelling

Would you like to calculate sales predictions for multiple sites across Australia based on the probability of individuals or households at any location visiting any one of several stores in their area?

Retail Gravity Modelling (also known as Huff’s Gravity Model) is a modified version of Sir Isaac Newton’s Law of Gravitation.

Retail Gravity Modelling looks at a specific node, normally a Census Collection District, travel distances to each store in that area and incorporates a variety of attractiveness measures (including size, population density, internal or external characteristics, choice of products and services, appeal to a certain lifestyle and brand) to define a probability model that will determine the likelihoold of a customer patronising (visiting and/or purchasing) at a given site.

Retail Gravity Modelling is particularly well suited to understanding trade areas and sales potential for shopping centres or shopping strips.

More specifically, Retail Gravity Modelling requires an analysis of the estimated spend potential for each node in the model. Estimating competition sales volumes can significantly improve the predictive ability of the Retail Gravity Model.

We can use our six step data analysis process to help you:

  • accurately define the market boundaries of trade areas
  • estimate the sales performance of proposed sites at any location including shopping centres, shopping strips, homemaker centres and homemaker strips
  • complete accurate ‘what if’ and market share type analysis
  • understand the cannibalisation impact of introducing a new site both on your existing sites as well as your competitors within a trade area
  • visualise the findings using the latest mapping tools as a Probability of Patronisation Map, Likely Retail Choice Map and/or a Trade Area Map 

Sales Cannibalisation Modelling

Do you have an established or mature franchise, corporate or retail network and want to understand the impact of a new site on an existing site or sites?

A new site has the potential to generate acceptable sales, but if a high proportion of these sales have been stolen (cannibalised) from one of your existing sites, the new store may not be economically justified.

In a franchise network, cannibalisation issues are particularly important as existing franchisees need reassurance that franchise network expansion will not impact upon their livelihoods.

Typically, Sales Cannibalisation Modelling is encapsulated in Sales Prediction Modelling. However, Spectrum Analysis can develop independent models using a history of site openings and the impact, pre and post new site opening that the new site has on surrounding sites.

We can use our six step data analysis process to help you:

  • estimate the impact a new site will have on the sales at other sites
  • run a Net Gain Analysis before committing to a new site
  • identify potential sites that do not cannibalise existing sites

Site Potential Check Chart

Would you like a Site Potential Check Chart assessor’s check list designed specifically for your enterprise that can be filled out by anyone and doesn’t require access to demographic or competition data?

Our Site Potential Check Chart can be used to develop a qualified score to help you make an informed quantitative decision on the level of interest to give to a site. It is a first level site evaluation tool that can be used before any major analysis is completed and can quickly filter out poor sites.

We help you define an accurate Site Potential Check Chart that includes identifying the drivers responsible for sales location and facility characteristics as inputs and returns a sales performance prediction for the site.

This Site Potential Check Chart can assess new sites and assist with the evaluation of the current franchise, corporate or retail network and your understanding of the dynamics of your enterprise.

We can use our six step data analysis process to help you:

  • avoid wasting time evaluating poor potential sites and focus on good quality sites
  • estimate the sales performance of proposed sites at any location
  • select the most appropriate facility for a given location
  • provide scenario planning
  • avoid gut-feel decision-making
  • compare actual versus predicted performance and actual versus potential performance of existing sites
  • identify and prioritise ‘new build’ site locations