Analytics

A Better, Cloud-Native Data Architecture for E-Commerce Analytics

By Nate Nunta - September 1, 2020

For e-commerce analytics teams, creating a source of truth for your disparate data across multiple e-commerce channels like Shopify, Amazon, and your own site can be a daunting endeavor. However, if you can build a single view of your transactions across channels, you’ll accelerate your ability to understand why key metrics like AOV, LTV, and acquisition change.

When you can monitor these metrics in real time, your analytics team can finally provide the business with actionable insights, informing everything from promotions to product packaging to future product decisions.

At Sisu, our Implementation and Customer Success teams have supported dozens of analytics teams in creating cloud-native data architecture tuned for the realities of e-commerce analytics. Here are a few of the best practices and tips we’ve picked up along the way. While some of these recommendations are common knowledge for data engineering and analytics teams, others result from our practical experience delivering faster, comprehensive, and more accessible analytics for customers.

 

Step 1: Establish a Cloud-Native Data Warehouse for E-Commerce Analytics

Before you can begin analyzing your e-commerce data, you’ll need all your disparate data stored in a central location — ideally a cloud-native data warehouse.

There’s a myriad of benefits to using cloud data warehouses that we go into in our full-length guide. For e-commerce analytics, what’s important is moving your data sources to a centralized online location to create a nearly infinite, low-cast, and scalable solution. Data warehouses also move both the security and management of your data to the cloud, providing additional security that is hard to replicate if you’re storing data on your own.

 

Step 2: Create a Data Map to Organize Disparate Data Sources

With your cloud-native data warehouse set up, outline the key metrics your business will want to track and identify, particularly if there are any unique metrics you need to pay attention to for your online sales.

Then you’re ready to create a data map. While it may seem tedious, this data mapping step is crucial to the success of the project. Different sales channels report transaction data in different ways. When you’re moving to consolidate the data, you need to ensure it’s moving from your disparate data sources to your data warehouse consistently.

To make this more concrete, imagine you’re an analyst at the Awesome Energy Bar company. You sell five flavors of energy bars (Chocolate, Peanut Butter, Coconut, Cherry Vanilla, and Chia Crunch) through an online marketplace, an e-commerce platform, and your website. Each sales channel provides slightly different ways of reporting transactions.

In your data map, you should include everything your business will need to know in the current environment and everything you envision they’ll want to know in the foreseeable future.

For example, Awesome Energy Bar currently only cares about how flavors and product categories impact AOV, but they’ve earmarked a budget to add marketing attribution data in the next six months. That means that you’ll need to start moving to the cloud warehouse without that attribution data in place. But with a data map, you can plan for this by researching the metrics the team will want to monitor, identifying the unique data factors the new marketing application provides, and leave room for those factors on the data map.

Example Data Map for E-Commerce Analytics

We’ve found that adding the intel you gather to the data map as early in the process as you can will make your e-commerce analytics easier in the long run. Plus, it has an added benefit of allowing you to answer unexpected questions from the business unit.

 

Step 3: Take the Wide View of Your Transaction Data

Particularly when you’re dealing disparate data sources like those in e-commerce analytics, having a single table that combines them all into a standard schema will allow you to see correlations within and across the funnel. It will help you create a single definition for a transaction, even if each sales channel provides different factors.

There’s an added benefit of allowing you to capture insights that may not have been obvious correlates: for example, at Awesome Energy Bar company, if the business was intent on capturing geolocation metrics and had aggregated their data according to that metric, they may have missed something less intuitive: that flavors with labels listing organic ingredients are actually a more significant factor in sales than location.

Think through the structure of your data table, including:

  • Accessibility: Can a stakeholder identify where a metric of interest is located in the data?
  • Cardinality: Is the data rich enough to contain the factors of interest?
  • Granularity: Is the accessible data fine-grained enough (e.g., transaction-level) to perform the desired analysis?
  • Actionability: Are the factors present in the data actionable to the stakeholder?

 

Step 4: Accelerate your Fact-finding with Augmented Analytics

In the past, analyzing your data across each online channel would have been slow, using CSVs, or focused only on one data source at a time. Now that you have all your disparate data in a central location, you can point an augmented analytics tool like Sisu at it to diagnose changes in your e-commerce data in real time.

For our fictitious Awesome Energy Bar company, they’ve connected Sisu directly to their data warehouse and can quickly see actionable facts from across their retail channels like:

  • New Yorkers are increasing their Average Order Value by 6.4%, which increases the overall change in AOV by $1.10.
    AOV results for e-commerce analytics using augmented analytics
  • Chia Crunch performance improves among customers who purchased in the morning, versus when it is purchased in the evening.
  • Los Angeles customers love the Protein Bars, increasing sales by 25.6% month-over-month.
  • The AOV for mobile orders dropped 8.6% month-over-month, which hurts the overall change in AOV by $0.92.

With a cloud-native data architecture for e-commerce analytics, you’ll finally be able to dig through the mountains of data you’ve been collecting to provide valuable input for the business. This is how you go beyond producing data for data’s sake and become an integrated and critical service within your business.

To learn more, see how augmented analytics is helping companies like Kitu Super Coffee understand their online sales in our recent case study, or get in touch with our team.


Read more

Designing Datasets: Four Principles to Advance Data Diagnosis

With more transactional data in cloud-native warehouses than ever before, analysts should stop aggregating their data for business intelligence tools. To help, here are four principles on designing datasets for cloud-native diagnosis.

Read more

Four Reasons Why Now is the Time for Augmented Analytics

Insights from Sisu customers on augmented analytics and how adopting this proactive approach lets analytics teams fully unlock the value of their data.

Read more