In 1978, Atari released “Avalanche,” an arcade game in which the goal is to catch falling digital ‘rocks’ before they touch the ground. The player must slide the paddles left and right as the rocks fall at an ever-increasing speed. Inevitably, one slips by and it’s game over.
Surprisingly, many Consumer Packaged Goods companies run their eCommerce operations in a similar manner. With only a small team responsible for getting products out and onto the digital shelf, it’s a veritable avalanche of product information that must be processed and validated. When bad data makes its way to consumers...you’re going to need another quarter.
Many companies still rely on antiquated Product Information Management (PIM) software and business processes that feel like they were developed for an Atari 2600. Even throughout the most recent decade, the interfaces have changed but the fundamental pain-points around aggregating, cleaning, enriching, and distributing data remain largely unaddressed. These systems give you a place to store your data, but the onus is still on you to enter all of the information. You’re stuck playing the same game, just with updated graphics. The work for an eCommerce team is still tedious, error-prone and, with the deluge of data, it’s only getting worse.
It’s about time that we rethink how we handle product information in the new age of Artificial Intelligence. It’s time...for a game-changer.
Digging through the Rubble
Companies have traditionally used a PIM to manage all of the types of product data needed for the purposes of sales and marketing, including:
- Product features & benefits
- Marketing descriptions
- Cross-sell/up-sell information
Other information, such as weight, dimensions, color, materials or ingredients are typically stored in a separate Master Data Management (MDM) system. Any multimedia or unstructured data such as images, videos, and documents are stored in yet another system, either a Digital Asset Management (DAM) system or - because many users detest the poor usability of DAMs - a Content Collaboration Platform (CCP) such as Dropbox, Google Drive, Box, or OneDrive. Many of the tasks regarding how and when data needs to be created or updated (e.g. schedule a photoshoot) may originate from a workflow management system, sometimes referred to as a Project and Portfolio Management system (PPM).
The trouble is that in order to drive a great customer experience, all of this product information must come together in the eCommerce store. Most companies solve this today by trying to glue their MDM, DAM, PPM, and PIM together with brittle, custom integrations. It’s not difficult to imagine the chaos this can bring, but let’s make a top-ten list:
- Integration failures that are nearly impossible to debug
- Never-ending deployment projects of upgrades and patches
- Inaccurate data that trickles downstream from MDM to PIM
- Corrected or enriched data that is now only fixed in the PIM, but not upstream
- No universal user permissions or access controls
- Poor visibility over how different fields are being used and why
- Having to search in multiple places to find all the information
- Sharing spreadsheets to keep track of which images align with which SKUs
- No cross-system standards, measurement, or monitoring of data quality
- Difficulty tracking down who owns what data so that it can be improved
The ‘avalanche’ is the burden of constantly finding and fixing mistakes inherent in operating disparate systems. Having multiple brands and multiple channel partners only serves to multiply the problems. Fortunately, it doesn’t have to be this hard.
The Impira Data Intelligence Platform
So, why do companies operate separate systems in the first place? Up until recently, vendors have designed their systems around a certain data type. PIMs and MDMs store mostly structured data (meaning that the data can fit neatly into rows and columns like a spreadsheet). DAMs store unstructured information (i.e. files such as images, videos, and documents). The Impira Data Intelligence Platform brings structured and unstructured information together by treating both types of data equally (every piece of data is an ‘entity’) and focusing on storing the relationships between the entities.
For example, imagine a lifestyle photo that depicts a woman wearing a jean jacket, a pair of jeans, and boots - all of which are products available for sale. The photo was taken for the “Spring Marketing Campaign.” The photo, the products, and the campaign are all ‘entities’ within the Impira Data Intelligence Platform. When performing a search for ‘jean jacket,’ any one of those entities could appear in the search results, making it quick and easy to find what you’re looking for.
Impira leverages AI techniques to strengthen the relationships between entities as it collects feedback from the interactions of the platform users. It is a continuously learning system that builds an understanding of how products relate to other products, or to any other type of entity.
Having your data together in one place sounds great, but if you’re thinking “This would never work for us, we could never get everyone to agree to use a single tool,” well...you’re right.
Solving the Data Centralization Problem
If history has taught us anything, it’s that efforts to centralize data into one platform never seem to work. Everyone has different requirements, and while it looks good on paper to have all your data in one repository, the reality is that different users and groups will splinter off over time.
Just look at the DAM industry. DAMs are sold and marketed as a single place to store all of your images and video but, for a number of reasons, most companies that have a DAM still have their images all over the place. Perhaps the product photos are in the DAM, but the Marketing team has all of their campaign creatives in Google Drive, for example.
The smarter solution is to leave the data in-place and allow everyone to use the tools that work for them, but find a way to universally search and locate data across those systems, and monitor the flow of data between those systems to facilitate better collaboration and an overall higher quality of data. This is the way of the Impira Data Intelligence Platform. We’re not here to replace your systems or to enforce top-down constraints that frustrate users. Instead, Impira is more like the glue between your systems and the only place where you can see all of your product information behind a single pane of glass.
By modeling the relationships between your data entities, we solve the data centralization problem by avoiding it altogether. With the Impira platform, we no longer have the headache of transferring all your data into a single repository or satisfying each individual user’s preferences for a specific tool.
Applying AI to Product Information
Why is it important to capture all your data, regardless of where it’s stored? Aligning and indexing all your data is a fundamental step towards applying AI to your business. Businesses that adopt AI will outpace competitors with lower costs, improved efficiencies, and ultimately higher revenues.
In order to be effective, AI needs your data as input. Once you take stock of all the data you have across your enterprise, you’ll realize that you have a lot of valuable information that can now be used to train custom models. For example, all of those image tags that your team has been manually entering in your DAM can be used to train a computer vision model, which could then generate tags based on your own product taxonomy and vocabulary.
The Impira Data Intelligence Platform is a framework that allows you to bring all sorts of AI techniques to bear, not just tagging. It’s time you took advantage of OCR/text-recognition, speech-to-text, image similarity, facial recognition, and document understanding. Isn't it time you used intelligent systems to make your data work for you?