The following is an interview with Erik Chelstad, the CTO of Observa.
Could you provide a brief overview of Category Intelligence?
Observa adapts quickly to the retail environment by being constantly in it. We have hundreds of thousands of Observers in stores across the US every day, which gives our clients access to a consistent stream of photos and quantitative data about the state of the shelf.
We use this massive amount of information to train our AI, which enables it to accurately recognize all the products in a given category and reveal actionable insights every month for our customers.
As you know, brick-and-mortar retail is chaotic. In any given category, we constantly see new products coming in, old products going out, and changes in packaging across entire product lines. Needless to say, having up-to-date information is essential if you want an accurate assessment of the shelf conditions. That’s not just for your products. It’s for your competitors’ products, too!
How does Observa’s AI work?
We use image recognition based on neural networks to track products on the shelf. We receive millions of photos from Observers every week, which provides a continuous pipeline of training data.
This enables us to automatically identify the latest packaging and product changes that are so important to this industry. Retail has a massive churn rate, so if you don’t retrain your models each month, you’ll start missing the products in your category. That number builds up over time, so by the end of the year, you’ll miss a significant percentage of the items on the shelf.
Is the AI able to recognize subtle distinctions between products?
Definitely! It’s worth noting that packaging changes can be very subtle. Sometimes, it’s just a tiny thing like a competitor changing a font or adding a little star that says “gluten free.” If our AI notices that something is different, it will alert us as the platform’s maintainers; honestly, this is where it gets really cool.
Our ability to recognize these changes and retrain our models guarantees that our clients are always seeing the most accurate, transparent view of their products and their competitors’ products nationwide. That’s not something that you glean with POS data or by doing your own sampling in a few stores.
How accurate is Observa’s system?
It’s extremely accurate! Over 99% accurate, in fact.
As I mentioned earlier, we’re able to keep our system up-to-date, because Observers are consistently in the field collecting data. Furthermore, our AI’s findings are double-checked by a team of human validators. This approach gives our customers the confidence that our insights are sound, and it also allows us to continually train the AI, making it better and better at automatically recognizing all the products in the category.
Is this giving the same power to brick-and-mortar that e-commerce has had for years?
Definitely, and it’s been a long time coming!
One of the main reasons that e-commerce has been so successful in the last decade is because marketing and sales agencies have access to a vast amount of information on everything: their shoppers, retailers, products… You name it!
In brick-and-mortar, you have access to some information like POS data, but that’s only one piece of the puzzle. There are still massive gaps in visibility about the true nature of what’s happening in-store. At Observa, we believe that brands should have as much visibility as possible, and with this newfound visibility comes a whole world of new opportunities.
Imagine being able to conduct A/B tests or even multivariate tests in physical retail. If you were only to use POS data, you’d be out of luck, as execution rates in-store are quite low, but by combining both data types, you can effectively carry out these tests. Brands selling in physical retail stores can finally quantify and streamline their shoppers’ buying journey.
And when I’m talking about these tests, I’m not talking about just conducting them in smaller markets like the current traditional approach for physical retail. Observa’s ability to guarantee shelf conditions enables you to expand your tests across multiple chains, regions, and even states, which is so exciting!
What data points do customers receive?
The amount of data you receive depends on your subscription level. While each level provides value, the higher ones will naturally offer the deepest insights. I’d recommend consulting our sales team to determine which one makes the most sense for your organization.
At the upper subscription levels, you can drill all the way down to the SKU level, empowering you to see the big (and small!) picture. Observa’s Category Intelligence helps you measure your share of shelf, approved product lists, and compliance by retailer, brand, product, and so much more.
How is the data collected?
Each month, Observers collect data at 3,000 different grocery stores across the United States. Our platform directs them to the locations that we need for statistically significant sampling across banners and geographic areas. It also ensures that we don’t constantly go to the same stores.
After an Observer submits an observation, the data is double-checked by fraud prevention algorithms and a dedicated team of human validators. We pride ourselves on producing high-quality insights, so we discard any data that doesn’t meet our standards.
What can you do with the data?
With Observa, you can accurately assess your brands’ and your competitors’ share of shelf and on-shelf availability across the country. You can visualize your data via an exportable .csv file, Observa’s API, or a data visualization dashboard. Sorry, I should say “data exploration” instead of “data visualization”.
Why do you prefer data exploration over visualization?
In general, data visualization relies upon someone else’s theories and hypotheses. It’s prepared for you and doesn’t have a lot of flexibility. At Observa, we believe it’s important for you to have the ability to test the hypotheses that are specific to your category.
Do you need to be a data scientist or have an insights team?
Not at all! The data is presented in a way that anyone can use, and the more knowledge you have in your specific field, the better off you will be while exploring.
If you are asking questions such as “How am I doing across large format stores in the midwest compared to my competitors?”, then you will find valuable information without having to be a data scientist or know any coding or Excel.
No matter what role you play in your organization, or how experienced you are, we want to remove the stress of collecting and working with data. We’ll help you get the information you need, delivered in a way that works for your business.
At Observa, we pride ourselves on being customer-oriented. We develop our product set based on what matters most to you. By doing this, we remove the need for you to set up an entire division around data science. Instead, we know you can use the data coming in over time and add more intelligence to it yourself by actively using our platform.
Simply put: we make it easy for you to make smart business decisions fast.