Reshuffling Your Data

Data is the future. But how do we use data to shape the future of our partners and allow them to weaponise their data? 

Data on its own is fairly useless; a large mix of digits and letters with no discernible order, ranking or informational value. This data could be the CRM database, the webshop interactions, the weather forecast vs realised or the past credit worthiness. All of these items are possible useful datapoints but on their own and without context they will most likely lead to imperfect decision-making, or worse colossal failure.

The underlying reason for this is the patterns that seem apparent in large amounts of data are often not there. Those patterns either do not exist or are caused by other factors. An easy example of this is found in the discussion around socio-economic outcomes with individuals; is this related to sex, ethnic background, education or even religious factors? An interesting discussion which however is pointless if the starting wealth (parents, inheritance etc.) is not taken into account.

Ada Beta uses an advanced algorithm to properly dissect the data, test all data for relevance (statistically and for causality) and regroup the data in a multi-dimensional group structure that allows for targeted decision-making. Decision-making based on actionable data!

In step 1 Ada Beta receives the raw data from our partners. This will be all the internal data that they have combined with any data enrichment, where possible. We will test this this data for easy grouping; outliers are identified and tested, basic patterns are used for hypothesis formulation and we focus also on the non-data. Non-data being the clients, products, services that were lost opportunities or that do not appear in the data but are well-known for appearing with relevant competitors.

How the data looks before we have analysed it

This raw data can be seen as a sea of dots, some with already easily identifiable quantities (size, sex, age, etc.) but also with loads of missing relevant data points.

We then move to a standard sub-selection of the data where all the usual groups are formed. This is not done for any specific decision-making purposes but it helps our partners to get a baseline for their dashboard. This process is done in a few steps to show how adding additional data or enriching the current raw data can lead to better splits creating a more balanced and informative dataset.

What the dataset looks like after a simple K-NN model

This is best illustrated when thinking about a supermarket modelling its data; who buys wine, who buys bread, who only buys milk on Fridays etc. Who are the basic customer categories and how do we best tailor our products, shop layout and pricing strategies to create loyalty and maximise the bottomline.

Ada Beta then applies its algorithm; rather than following a one-dimensional approach we release all known data (raw, enriched and otherwise) to create the right grouping for the overall True Customer Value. This allows us to incorporate referrals, weather patterns, economic circumstances, shopping patterns, evolving family structures, actual business value (cashflow and margin %s) and many more data points.

With this we create a multi-dimensional grouping that correctly identifies those clients that are truly worth chasing after and the clients that you’d happily see at your competitors.

Now you finally have transformed your raw data into actionable information; no longer an ocean of data points but clear instructions on which clients should be the core target audience for your business!

What the dataset looks like after the algorithm of Ada Beta

Would you like to experience how powerful our modelling can be for the future of your company? Contact us at info@adabeta.nl