In the part 1 of Purchase events post, we looked at the key events and base data models based on those events.

Purchase events can be modeled in many different ways to get insights. Some of these models should be scheduled as jobs because while others can be built as needed.

In this post, we will explore common analysis on purchase events and the data models required to run these analysis.

The models covered in part 1 are required for building the models detailed in this post. I refer to the models from part 1 as base models in the post below.

Use Cases

Checkout flow by session and user.

Two metrics that define the fate of most stores: Conversion rate and average order value.

Marketing teams calculate conversion rate very simply as – conversions/sessions. But this is just one way to look at conversions. And this is a useful metric to optimise marketing budgets, or run A/B test on your checkouts.

But it’s not a good metric for building your product.

Users don’t just come to your store for the first time and convert. They go through an entire journey from awareness -> purchase over several sessions, sometimes spread over days or weeks (for higher ticket purchases.).

To get actionable insights, you need to zoom in and zoom out from this metric. Let’s zoom in first.

Checkout flow

The base checkout_flow model is the zoomed in version of conversion rate. it has one row for every checkout initiation by the user.

A checkout initiation happens every time users click on ‘Checkout’ or ‘Buy Now’ buttons on your website. Assuming you don’t have a login, users will land on the checkout page in a couple of seconds. Most users will not initiate payment, they will either bounce or go back to browsing.

If you have a login before payment, many of your checkouts will not have a ‘begin_checkout’ row. Your checkout flow needs to take this ‘login’ into account as well.

You can dig into the product side by asking:

• how long does it take users to complete a payment?
• what is the dropoff at each stage of the payment funnel?

Checkouts_in_session

Then you can zoom out a little to the session level. This is similar to the conversion rate metric marketers use, but we are not interested in CR.
Instead we can combine all initiations in a session into one flow, and ask questions.

Questions to consider:

• How often do users initiate checkout before bouncing?
• How often do users initate checkout before buying?
• For returning users, how many sessions and checkout initiations before an actual purchase?
• For returning users, how many of them have initiated checkouts and not completed purchase?
  • what % of these return and complete a purchase in 7 days?
• For returning Buyers, how many sessions and checkouts before 2nd, 3rd purchase?

Data model

Enter data model for checkouts_in_session.

Checkouts_user

Let’s zoom out further. Here we are mainly interested in returning users and buyers. We want to focus on two things

• sessionwise journey of a user before making a first purchase
• journey of a existing buyer between 1st and 2nd purchase

Data model

Enter data model for checkouts_user.

Checkout with emailers/whatsapp

Stores get a significant share of purchases by converting abandoned carts. Often, these cannot be tied back to a users checkout funnel because of data issues. But if you can do this (with some id available in some cases), you have an excellent follow-up to the above data.

• what % of checkouts complete via emailers, etc.?
• what % of checkouts complete via emailers + discount?

Data model

Enter data model for checkouts_emailers.

Items

The other area of focus in the items being checked out.

For stores with a large selection of SKUs, divided into multiple categories: this analysis can lead to very interesting insights.

First time buyer preference

• What do users purchase when they are making first purchase from the website? For stores with a large selection, which categories fill the cart for first time purchasers?
• Which items/categories bring users for 2nd purchase, which ones don’t?

checkout flow by categories

How does the checkout funnel vary depending on the items in the cart, the price of the items. How can this lead to more optimised marketing efforts?

removed items

No user journey is complete till some items are removed from cart! In my experience, users bring down the cart value by 30-50% before finally checking out. The items that get kicked out are the ones added on impulse.

• Which items or categories are removed from cart most often?
• what is the % conversion rate of a category from added to cart to actual purchase?
• which items get reduced or increased in cart?

Why ask such questions? For creating bigger or bundled options of products, for rethinking the price of some items, for keeping this in consideration in your marketing efforts, etc.

Items added on checkout

Similar to items getting removed, users also add items. This is true when for small value items which can be added to cart without much thought – like a chocolate protein bar while purchasing protein supplement from a store, or an accessory while buying a formal clothing.

These can also be analysed with the same data.

The complete item funnel.

GA4 provides a particularly powerful view in ecommerce.
Items viewed -> purchase.

You can build the same and then make it a lot more useful with paid vs organic comparisons, additional data on prices, data on category of item, separating first time users, returning users, and returning buyers, and so on.

This funnel is among the most important views available in data, and one that can unlock many, many insights if you explore it thoroughly.

Combining purchase data

Frequently purchased together

Identify bundling and upsell opportunities with this short analysis of items that get purchased together.
For larger selections, look at categories that get purchased together.

??