Market Basket Analysis

What is Market Basket Analysis?

Market basket analysis studies the concept of affinity. Affinity is the natural liking or understanding of something. It can also mean the degree to which something tends to combine with another.

We measure and leverage this property of substances to naturally combine with others. All we need is to be able to identify an individual and have a detailed history of the products that they buy.

This type of analysis is now well known, especially in the on-line world. In e-commerce we are now all used to seeing the “Customers also bought…” suggestions. IP television providers show programming choice that changes by what we, and other customers, watch.

The Psychology

The reasons we buy are complex. A great many purchases are made on impulse. If we see an item or are made aware of it then a good percentage of people will buy. In order to make more money a business just needs to trigger that impulse. Businesses make suggestions that resonate with the customer.

The Big Issue

The theory and mathematics behind Market basket analysis are simple. The issue is with the sheer volume of data that we need to analyze to tease out interesting affinities. Firstly, we usually have a large volume of transactions to study. However, the issue is the number of combinations of products. Here are some numerical examples:

If a company only has 10 products then the number of “baskets” of 2 items is 45.
If we increase our product range to 100 or 1,000 then the number of baskets jumps to 4,950 and 499,500.
Notice that a tenfold increase in products results in around 100 times more combinations.

Analyzing baskets of two items will tell us about affinities. However, with just two items, we usually already know the answer. If the output tells us that people who buy bread also buy butter then this is not useful. To really gain the competitive edge we need to find more interesting behaviors and that means considering affinities between three or more items. If we repeat our numerical example with a basket of three we see that the number of combinations explodes:

A company with 10 products will have 120 baskets of 3 items.
Increasing the number of products to 100 or 1,000 pushes the number of combinations to 162 thousand and 166 million.
A company with 10,000 products will have too many 3 item baskets to compute.

A Better Approach

Whilst it is possible to run large calculations, we can also employ differential analysis. This is benchmarking of sales in different stores. We could also benchmark different seasons, demographic groups or other attribute to see if we can see a significant improvement in a particular group. Having identified the improvement the business can then investigate the underlying reasons. Any factors that can be understood are then rolled-out through the business.

Another approach for market basket analysis is to look for predefined baskets to offer a targeted promotion. Searching for customers buying charcoal, firelighters, matches, meats will tell you they are planning a barbeque. This then tells you the customers who, in all probability, have and enjoy a garden and who could have a propensity to buy other outdoor leisure goods.

The Secret Recipe For Success

The scope of an affinity analysis must be very carefully defined with upper management. It is very easy to get “lost in the weeds” doing this kind of work.
Don’t scope something that is computationally hard. Is three week’s compute time reasonable in the context of the rest of the project?
Bring the objective down to something practical and manageable.

Our Other Data Solutions Services

Business Intelligence (BI)

From simple reports to complete BI strategy that fully meets your business needs, we can help. Enjoy the flexibility that mobile and self-service BI brings to an organization without re-inventing the wheel.

Market Segmentation

Market segmentation involves dividing customers into groups of people with similar characteristics or interests in order to make your marketing more effective.

Data Modeling

We have built many market models for our clients over the years allowing them to better understand the dynamics of the market and also to measure their performance. Market models explain what happened in the recent past and also allow us to forecast what is likely to happen.

Advanced Analytics

We can apply complex algorithms and Machine Learning and techniques of predictive analytics to a great variety of problems. Advanced analytics is a subset of data science that uses high-level techniques to predict future trends and behaviours.

Text Matching and Analytics

Being able to match two data sources is vital to good analysis and we have a proven track record and methodology. We can also apply statistical and machine learning techniques to textual data to extract information and sentiment from the data.

Data Processing

Good Data Science starts with a deep understanding of how data should be manipulated and stored. Traditional data and big data need different approaches but the underlying discipline is the same. The way we store data differs greatly and we can advise on how to adapt your platforms for better insight generation.

Integrated Security

We take integrated security very seriously at JTA. We have the most stringent security practices and hold ourselves to a higher standard than most. JTA can build dashboards that intelligently adapt their content depending on who is viewing it to help users comply with corporate accountability legislation.

Predictive Maintenance

Predictive maintenance is a benefit that we can bring to an organization by predicting failures and quality issues. This can avoid downtime and reduce maintenance costs.

Customer Churn

Using data prediction to identify when a certain customer is at high risk of churn allow clients to prevent attrition and keep existing clients. This helps immeasurably and keeps you with one in the hand rather than two in the bush.

Internet of Things

The Internet of Things is the network of small devices that have an embedded processor to measure, control and connect to the internet. Many such devices benefit from sophisticated algorithms and data science techniques. We can develop solutions ready to roll out for deployment on the microcontroller.

Outsourced Analysis

Companies which prefer to outsource data analytics instead of deploying a solution in-house can use JTA to form a virtual team to manage outsourced data analysis. Our team would typically access your data and your infrastructure and do the analysis remotely.