Epic Battle

Replicating Android vs iOS

Market Simulation can replicate competitive battles in real-world markets, like the battle between Android and iOS.

Just as real-world markets have products, features, brands, stores, locations, and competitive rivals, so does a Market Simulation. But what makes a Market Simulation truly realistic are the customers. Simulations can generate tens-of-thousands of virtual customers designed to mimic the purchase decisions of real-world shoppers. Customers evaluate the differentiation offered by each product.

Market Simulations provide a way to understand the economic complexities of a market. Simulations are used by academics, students, consultants, and business managers to predict how customers will react to change. The change might include a change in price, a change in product assortment, or the emergence of a new competitor. These predictions lead to improved business strategies that increase market share, revenue, and profitability.

This Case Study provides a high-level overview of the workflow without detailed explanation. It assumes you are already somewhat familiar with KNIME and Market Simulation. If not, start by reviewing the Building Blocks and Community Nodes.


Market Simulation is built upon the same principles as Conjoint Analysis and Mainstream Economics. The simulation uses an Agent-Based Model (ABM) to replicate the decision-making process of individual customers.

Customers purchase those products that give them the greatest Consumer Surplus – that is, the greatest difference between their Willingness To Pay (WTP) for a product and its price. A customer’s WTP for a product is the sum of the “part-worth” values of its independent features. For example, a customer’s WTP for a phone might be their part-worth value for software plus their part-worth value for hardware.

The free Market Simulation (Community Edition) for KNIME comprises of more than two-dozen nodes dedicated to creating, tuning, and simulating markets. Almost any market can be simulated, although business-to-consumer (B2C) markets are easier to tune.

Scientific Stratey’s Market Simulation nodes for KNIME can be used to simulate any market – like this one for Android vs iOS phones.

KNIME Workflow

Five of the free Market Simulation (Community Edition) nodes have been highlighted within the Android vs iOS workflow.

The competitive battle between Android and iOS phones illustrate how the Market Simulation nodes work on KNIME.

Apple is the sole supplier of iOS iPhones, while Android phones are manufactured and sold by many suppliers. Apple is also the most profitable supplier of phones, with $1000+ iPhones that are more expensive than comparable Android phones.

While Android suppliers are less profitable and generate less revenue, Android phones have a larger overall market share by quantity sold. Many varieties are available, with the price of Android phones ranging from very cheap to very expensive.

A KNIME workflow that simultaneously models all these market characteristics can be easily assembled with Scientific Strategy’s Market Simulation (Community Edition) nodes.

Product Features

The Android vs iOS workflow creates an Agent-Based Model (ABM) of 10,000 Virtual Customers. The products in the market include:

  • Apple’s iPhone
  • Samsung’s Galaxy
  • LG’s phone
  • Google’s Pixel
  • Oppo’s OnePlus

The two features that customers most value are:

  1. Software (iOS or Android)
  2. Branded Hardware

The prices of the phones range from $1,000 (for the iPhone) down to $500 (for the OnePlus).

The manufacturing cost of the high-end Apple phone has been estimated at $800, while the cost of comparable high-end Android phones has been estimated to be only $700. While the component costs of the hardware going into each of these phones is about the same, the cost of developing and maintaining the iOS software is higher for Apple. This is because Android software is available for “free”.

Note that this workflow is only meant to illustrate how the KNIME nodes can be used to simulate a market. No specific product SKUs or market channel has been selected for simulation, nor has the model been tuned to make accurate predictions. While the Community Edition nodes do include these capabilities, they have not been illustrated here.

Value of Software

While individual customers place different levels of value on the Android and iOS software running in a new phone, the average value of software across all customers is the same for both.

The part-worth value that individual customers have for iOS software is different to their value for Android software. Android software provides users with more options, while iOS software has a better reputation for reliability.

However, the average value for Android and iOS is about the same and has been set in the simulation to $500. Because there is no difference in this average value, the software is said to offer no Vertical Differentiation.

The new Customer Distributions node then converts these average values into two distributions. Each distribution contains the individual part-worth values for the 10,000 customers in the market.

Software Value

The average part-worth “Value” of iOS and Android software is the same.

Vertical Differentiation

Individual customer Willingness To Pay (WTP) is normally distributed.

WTP for Software

Each of the 10,000 customers (C00001 to C10000) have their own individual part-worth value for both types of software.

Branded Hardware

Customers have more difficulty distinguishing “Android” branded hardware than “Apple” branded phones as Android phones “all look the same” to many customers.

The part-worth value customers have for branded hardware approximates the value of software at around $500. But in this case, the higher value for Apple’s hardware ($700) reflects the superior brand power of Apple. On the other hand, Oppo’s low value ($200) reflects the inferiority of its phone’s camera, processor, and memory.

The new Matrix Distributions node works like the Customer Distributions node by generating customer distributions of part-worth values for each type of branded hardware.

But here the Matrix Distributions node also considers the degree of similarity between features. In this case, many customers believe that “all Android phones are the same”. This belief is reflected in the high levels of “Conformity” between the Android phones but not the Apple phone (conformity is set between 0.0 and 1.0). This conformity means that Apple has Horizontal Differentiation with respect to Android phones.

Hardware Value

Customers place a higher average value on the Apple brand and a lower average value on Oppo hardware.

Horizontal Differentiation

“Conformity” is an indication of how similar customers perceive all Android hardware.

WTP for Hardware

The 10,000 customers also have their own individual part-worth value for each type of branded hardware.

Features to Products

The Product Generator converts features into products.

The part-worth values of software and hardware need to be combined into overall Willingness To Pay (WTP) values for the products. This is achieved with two new nodes:

  1. the Feature Table To List node, and
  2. the Product Generator node.

The upstream Table Creator node describes all the features that make up the products, along with the manufacturing cost and price of each.

The Product Generator node creates a final “Product Array” and “WTP Matrix”. This is all the data needed to run a predictive simulation of the market.

Combined Value

Each phone is made up of a branded hardware Feature and a software Platform.

Feature List

The ‘Feature Table To List’ node creates a list of Product Features.

Product Array

The Output Product Array lists all of the Products in the Market along with their Prices and Costs.

WTP Matrix

The Output Willingness To Pay (WTP) Matrix matches each Product to each Customer.

Market Simulation

The Simulate Market node predicts which product each of the 10,000 customers will buy based upon their individual Consumer Surplus.

The Simulate Market node is the last new node in the workflow. It takes both the Product Array and WTP Matrix and calculates the Consumer Surplus for each customer (recall that Consumer Surplus equals WTP minus price). The Simulate Market node then predicts which product each customer will purchase (or “No Sale” if the customer finds none of the products attractive).

The results from the Market Simulation reflect the results from the real-world market:

  1. Android has a bigger overall market share (by quantity sold) than the iPhone
  2. Apple generates the most revenue
  3. Apple is also much more profitable than all Android phones combined

The Simulate Market node replicates the decision making process of each Customer. Customers who buy the iPhone pay only the $1000 price but might have paid up to their WTP. The difference is their Consumer Surplus which reflects their overall satisfaction with their purchase.

Final Results

Market Simulation reflects the same results seen in the real-world, with 54% of customers buying an Android phone and only 42% buying an iOS phone. There are also 3.9% of customers who don’t buy anything.

Market Share

Predicted market share (by quantity) of each phone in the market.


Quantified Differentiation

Market Simulation works because the differentiation offered by each product has been quantified.

On average, customers perceive no difference between Android and iOS software. Hence the software provides no Vertical Differentiation. However, the fact that individual customers disagree whether Android or iOS is better means that the software does provide Horizontal Differentiation.

The differentiation provided by the branded hardware is also a factor. While the Apple brand does offer some Vertical Differentiation, what’s more important is the fact that the Android hardware is mostly undifferentiated (customers believe “all Android phones are the same”). This leads to greater rivalry among the Android manufacturers which drives down the price.

When all types of differentiation are quantified the dynamics of the entire market can be modeled.

Data Sources

The input parameters for this Market Simulation are loosely based upon the range of phones available in the USA. While Chinese made phones have a large global market share, customers only make purchase decisions from the range of products available in their own geography.

The input data for this analysis was inspired by the CNET article “Why your iPhone and Android phone will cost more in 2019” by Jessica Dolcourt (2-Jan-2019):


Cost estimates were inspired by the analysis by HiSilicon found in the article: Economic Research Working Paper No. 41: Intangible assets and value capture in global value chains: the smartphone industry by Jason Dedrick and Kenneth L. Kraemer (Nov-2017):