How to do a Customer Lifetime Analysis with Turbular's AI Assistant

Investigating our Data

In this tutorial we will use an Excel File containing sample order data from an online shop. This is how our file looks like For the scope of this tutorial we are really only interested in the order tab, as this the data from which we can extract the Customer Lifetime Value for each cohort by month.

Asking Kepler Medium To Generate A Customer Lifetime Value Analysis

We first start by naively asking "Can you give me a customer lifetime analysis based on order value ?". As you can see based on the result Kepler Medium did give use the Customer Lifetime Value, however it gave us the value for every customer. Normally when performing Customer Lifetime Value Analysis we are interested in grouping our customers into monthly cohorts. Since Kepler is just a basic AI model he is not aware of that context and just gave us naively the Customer Lifetime Value for each customer.

Refining our Question

In order to remedy our mistake in posing the previous question we are now asking Kepler specifically to group the customers by the month of their first order and to calculate the average of these groups. It seems like Kepler gave us a response, and we could now inspect the Data or download it via the "Message Details" button.

Visualizing our Results

Now in order to confirm the data Kepler has extracted and in order to present a chart to our boss we ask Kepler to visualize the result. As you can see Kepler returned as a Linechart with the customer cohorts by month on the X-Axis and the Average Customer Lifetime Value for each cohort on the Y-Axis. Based on the graph we can intuitively infer that the extracted data by Kepler is correct as there seems to be a downward trend in the Average Customer Lifetime Value over time, which makes sense given that older cohorts typically performed more purchases.

Change to Barchart

Given that a line chart is not too well suited for this data, I now want to change the graph. I could go click the "Message Details" button and change the graph manually. However, I am lazy therefore I just ask Kepler again to visualize the data for me as a bar chart. Now we have the result we want without ever touching a database or a single Excel function. We can now download the data and the graph via the interface which opens when I click the "Message Details" button and present the results.

Things To Note

As you can see from this conversation flow our Kepler model do not get things always directly right given that they lack context, understanding or your question is just too hard. However, by interacting with Kepler's response we can steer our AI Data Analyst into giving us the response we want.