Lessons from the Science Museum
It might seem a little surprising to begin an article on Marketing Analytics with a trip to London’s Science Museum. Or perhaps not, given the rising importance of Data Science in analysing marketing spend and performance.
Amongst the jet engines, satellites and space suits is an odd machine that houses a number of pipes and tanks, and that exhibits a number of brass wheels on the outside. This analog computer is a recreation of similar machines that were built by Bill Phillips in the 1940s and 1950s, and other replicas are housed all over the world.
The machine aims to use the flow of water through valves and tanks to simulate a country’s economy. The water represents the flow of money through the economic system, and the tanks represent the accumulation of savings or investments. The brass handles operate valves that can be used to stimulate or restrict the flow of water, representing the levers of Government and the tightening or loosening of monetary policy.
The machines were used as a training device to teach Economics at a number of business schools and Universities. The importance of this approach was to show how relatively small changes to the economy could quickly result in large imbalances, and that the delayed effect of any corrective actions made problems difficult to rectify. The economy is a dynamic system, and it is the rate of interactions, as well as the interactions themselves, that need to be accounted for. And that can be particularly difficult to model using spreadsheets or simple calculators.
A key advantage of teaching Economics in this fashion was that it precluded any necessary advanced mathematical knowledge. Anyone could now easily understand some of the relationships that exist in an Economy, even some of the more complex and non-linear ones.
Modern Marketing Analytics
it’s easy to see how these principles could be used to model other dynamic systems, and in particular when thinking about marketing approaches. Like the movement of money, the changing habits and movements of consumers can also be thought of as ebbs and flows. The consumers of your services are a resource, but they are not an unlimited resource, and they may use your services unevenly.
Fortunately, building this type of model no longer needs any feats of engineering or plumbing skills, but merely access to a web browser. And as with Phillips' model, building these models does not rely on advanced mathematical skills, but simply a deep understanding of the relationships in your business. The remainder of this article walks through how to build this type of model using Silico’s unique solution
Step 1 - Defining Consumer Types
The first step is to map out the relationship between types of customers. For the purpose of this model we will define our customers as being in one of four groups:
1. Loyal Brand Customers
2. Current Brand Customers
3. Competitive Brand Customers
4. Non-customers of any brand
These four customer types can be thought of as being connected together in four buckets from which customers can move freely from one to another. In Silico our easy to use interface makes creating and naming these groups straightforward.
Step 2 - Define the relationships
The next step is to define the relationships between these types of customer or potential customer.
The graphical interface makes it as simple as possible to connect my groups using drag and drop functionality. For this model we are assuming that customers can move freely between groups, but that Loyal Customers only interact and interchange with my Current Customers.
These structures can be as simple or as straightforward as my real-world customer relationships.
Step 3 - Setting my population sizes
Once we have defined our groups and determined their relationships, we can then set the size of each population.
By simply clicking on each of my Groups a dialogue box opens in which I can enter a value, define a formula, or even import a time series of historical data.
For this demonstration model we have added an addressable market of 100M consumers, and also defined the number of consumers in each of our other consumer buckets.
Step 4 - Adding Marketing Campaigns
Next I want to represent the marketing activites that I have at my disposable and the influence that they might have on the consumer types that I’ve modelled.
In this example we are going to add some marketing campaigns. Each campaign is specifically aimed at a different customer type. These then feed into the model structure and influence the flow of customer types as they move between groups.
Again, I can parameterise the campaign’s influence on my customers in order to explore sensitivities and relationships.
Step 5 - Set a scenario
A key benefit of this type of approach is that we can now overlay scenarios on top of the framework that has been created.
For example, I might decide that my ‘Choose Our Brand’ campaign is going to behave in a certain way. I can use a formula to define this, or indeed import data to define the scenario.
I can then use my model to see how this scenario plays out and interacts with my other campaigns.
Step 6 - Dynamic Dashboards
In 5 easy to follow steps I’ve created my marketing model. As I run my model through time the Silico dashboard both visualises the results, but also allows me to play with the inputs.
The dynamic dashboard allows me to change the values for anything that I’ve parameterised, and to instantly see how that plays out in my results.
I can also export the data that I’ve generated into other data science workbooks or visualisation tools for further analaysis.
Exploring the Results
Even in this relatively simple and easy-to-build model I am already able to produce some pretty interesting results. In my dashboard I have created a slider to change the influence of the Brand Loyalty Campaign and examine the results this has. This is represented by the Campaign: Brand Loyalty slider at the top of the Dashboard.
As I move my slider from left to right, changing the influence that my campaign has, you can see that the results are not linear. The marketing campaign does not have a straightforward effect on my customers, because of the relationships between those customers and other customer types. This type of insight that takes into account second and third order effects is incredibly powerful, and is difficult to achieve so easily with other non-graphical tools.
The obvious next step now that I have generated these insights is to try to optimise my campaigns to produce the best possible outcome. Silico also provides optimisation tools to ensure that this can be achieved. In running an optimisation analysis I can define the outcomes that I want to achieve, and allow the Silico solution to search for the right mix of marketing campaigns to produce that result.
Marketing Analytics is just one area that Silico can help with. This same approach can be used to explore other business challenges, from Workforce Planning, to Supply Chain Modelling. And, even more powerfully, models from across your organisation can be linked together and, for the first time, give you a truly holistic modelling framework on which to base important decisions.