Maximizing Data Insights: How to Overcome Data Silos and Improve Attribution
Get The Print Version
Tired of scrolling? Download a PDF version for easier offline reading and sharing with coworkers.
A link to download the PDF will arrive in your inbox shortly.
A successful ecommerce strategy does not come from following a “gut feeling” or taking a shot in the dark — it comes from looking at the facts, learning from previous marketing efforts and drawing data insights.
However, for merchants like HealthPost, data can often become fragmented across various platforms, making it difficult to prepare reports and gather meaningful insights.
To get more insights, we decided to chat with Overdose. ourselves. Sitting down with Kaustubh Joshi, Chief Data Officer at Overdose., we learned a few best practices for how brands can make the most out of their data.
Kaustubh Joshi: “It starts with accepting the fact that different data sets are going to be different — and not everything is going to be talking to each other. For example, your marketing data from your Facebook Ads and Google Ads are completely different from your ecommerce data. Because of user privacy, there’s a big problem of attribution, and you can’t really have a one-to-one linkage.
“Instead, how can we solve this problem in a different way by using modeling and correlations? Let’s say I have a marketing mix of 50%, 40% and 10% across three marketing channels, and I have a certain cost of acquiring customers by spending this amount of money every month. Based on that, I know my consolidated cost of acquiring a customer after all of that marketing effort.
“When I track this data over several months, I can start seeing insights. When I change my marketing mixes across different months, how does it impact my CAC (customer acquisition cost)? When I shift from 40% Facebook Ads to 60%, did I actually gain more customers? Did I actually reduce my CAC? All of these kinds of correlations are the parts where you actually get the insights that you were missing.
“That is basically the approach that we’ve taken in a project like HealthPost, which is that some data will talk to each other and some data will not talk to each other. But we can use correlations and modeling to actually understand and answer those questions.”
KJ: “The benefit of BigCommerce is that it’s one of the first platforms to actually say ‘We offer an integration natively, so you don’t even need a third-party connector. You just configure your Google BigQuery credentials, put in your key and the data is replicated.’
“I would say that is a really great thing for merchants to have, because most of the time, it’s the little costs that creep up — the fact that I need to pay $500 a month for another connector platform to simply replicate and copy the data. I think more platforms will eventually get there, but it’s great that BigCommerce has already done that.”
KJ: “First, continue with a deterministic approach (in simple terms, determining which channels a customer came from) as far as possible. For example, Google Analytics by default offers data-driven attribution and has now started using their own first-party data, called Google Signals, to mitigate some of those attribution losses.
“Second, use new methods like marketing mix modeling and correlations, where you do not try to determine attribution. The premise of this thinking is that it’s not one channel that brings you the purchase — it’s the mix of your marketing. Multiple channels play together, so we should analyze it as a mix.
“The third is the simplest approach for any business today: simply ask your customers. When you make a purchase, you have an order success page, and if the business asks you, ‘How was your experience?’ you don’t mind giving them an answer. If they ask you, ‘Where did you get influenced for this purchase? Was it a Facebook ad? Was it an Instagram post? Was it a YouTube video that you saw?’ You will get answers and the insight that you get from customers.
“Google actually calls this the Measurement Trifecta — which is attribution, marketing mix and zero-party data. So that is the way I would say that brands should answer the attribution question. None of them will give you an exact answer, but three of them combined will give you enough context for you to feel confident about your marketing decisions.”
KJ: “Initially, you should not need an entire data science team, because you can very well leverage an agency partner — like Overdose., for example — who can get you on that journey. We say to most of our clients, ‘If you want to build this yourself, you will need a data visualization person, and you’ll need a person who understands data warehouse, who can code SQL and who can do Python.’
“So either you need a unicorn of a person, or you will need to hire three or four people, which is a big cost. And obviously, no one is going to hire four people without having any kind of idea of what the outputs are going to be. So that’s where you work with an agency like us and we can get you on that path, because we have those teams and you don’t need four entire people for that.”
It’s never been easier for online retailers to access and analyze customer data. With an agency partner like Overdose. and tools you’ll find in BigCommerce’s Big Open Data Solutions, merchants can easily overcome data silos, improve attribution and bring data insights to life.
Ready to build a data-driven ecommerce enterprise? Check out BigCommerce’s Big Open Data Solutions page to learn more about how to ignite the power of your data.
Haylee is a Content Marketing Writer at BigCommerce, where she partners with the SEO team to craft narratives and blog content. She earned a B.A. in English Literature from the University of Texas at Austin and afterward spent a year abroad to pursue a Master's in International Management from Trinity College Dublin. When she’s not writing, you can usually find Haylee with her nose in a book, enjoying live music or scoping out the best local coffee shops.