Product data is the foundation of a strong-performing ad. It is the fuel that allows Connexity to boost brands to maximise their reach and achieve results. Yet too often, poor-quality information hinders the success of an otherwise star product. We sat down with Nick Shelbourne, Senior Technical Account Manager, to discuss all things bad product data in the first of a two part series . From its key causes to preventative measures, we dived into everything you need to know.
Nick’s Role in Connexity
Nick is a Senior Technical Account Manager at Connexity. His role blends two key areas: technical operations and customer enablement. He acts as the technical lead across the customer lifecycle, ensuring integrations, data, and systems drive measurable business outcomes for customers
Nick explains, “Product data gets sent from our advertisers into our systems, where we enrich it and optimise it with our campaign management tools and human analyst team to deliver their performance goals. I’m a human face for some of our more technical processes and questions. Especially with our bigger accounts, when we think about compliance and quality, if there are any issues, then I have full ownership of that and am able to design bespoke solutions.”
The Key Platforms He Uses
Nick works across the entire technical life cycle, and in terms of product data, he uses Connexity’s own in-house feed management tools. This system ingests data through file drops, API, or scheduled pulls. He works closely with the DataOps team as they extract, process, and map data to the company’s internal database. After ingestion, custom pipelines are used to validate data integrity, check for missing attributes or sudden drops in offers, and enrich the data by processing titles and categorising offers.
“It’s essentially a pipeline where you receive the data from the customer and then you process it before you pass it on. The final platforms that we work with when it comes to publishing and syndication are our Google Merchant Centre, Bing Merchant Centre, and then affiliate and publisher networks” Nick adds.
Defining Bad Data
A person’s definition of ‘bad data’ can very much be open to interpretation. A common description given to this type of information is simply data that is missing or incorrect. In the context of performance marketing and working with clients, the ultimate goal is to meet key performance indicators (KPIs) for merchants. Any data that hinders or limits the ability to achieve these KPIs can therefore be considered “bad product data”. This can be broken down even further to highlight whether certain attributes are missing. These issues quite often result in the data being unusable; however, this isn’t the only type that causes problems.
“In our feed specification, there are some fields that we say are necessary and then there are some that are nice to have. But in reality, those nice-to-have attributes can really change the outcomes for the campaign. So you can have bad data that’s usable, but it’s going to really affect the performance downstream,” says Nick. “These could be elements such as not having the right colours (on an item) or not having the right Global Trade Item Numbers (GTINs). Although some might think they don’t need that field or that it’s not required, it actually will affect the outcomes.”
Time is a factor too: “Data can be bad as it is missing attributes, but it can also be stale or out of date. We need the data to be delivered frequently and it sometimes depends on how quickly they’re changing the prices. Some of our partners have a very dynamic pricing strategy. For example, large international marketplaces, like AliExpress or Temu, are changing the price throughout the day. So by the time they’ve sent us the data, it’s almost already out of date. In those cases, it might be that all of the fields are there, but the price is no longer accurate because by the time it’s been ingested and processed and published, the price on the website is different.”
Unpacking the Complexity of Its Creation
Merchants work with lots of different channels and partners. Prior to onboarding with Connexity, they often have all of the information that Connexity needs but not necessarily in the right format. This then requires some technical work on their side to create a new pipeline for Connexity. This is where many problems can occur: not passing full and accurate product data. Without the data, the quality of the product ad is then put in jeopardy.
“That’s probably the main driver of the issues: the merchants have the relevant data, but there’s some technical overhead required to create the pipeline that passes it to us. Often our direct contacts are driving the onboarding and it’s their job to bring more traffic there in the marketing team. However, they regularly don’t have the ability to actually set the feeds up. Sometimes they’re using a tool where they have the ability to do that within the marketing team, but often they have to book in time with their operations or their engineering team. People can be sceptical about investing in technical work without proven results of how successful a partnership can be. This can lead them to then give us a very bad quality feed with only the required attributes. As I mentioned before, this directly ties to outcomes.”
Marketplaces present their own challenges, “I would also say that with large marketplaces we work with, there is an element of human error because their product data is often uploaded by external sellers. They add their products to the marketplace and they have to add the attributes and they may make mistakes, like entering a product’s colour wrong. This is very difficult to deal with at scale so those issues can fly under the radar, but they can really impact performance and business outcomes. The main issues are more technical, but obviously human error can come into it.”
