Every retailer now understands the strategic importance of AI. Yet ask whether their own data is accurate, complete and ready for AI, and the picture is often far less clear.

For Richard Widdowson, VP of global retail solutions at SAS Institute, this has become one of retail’s defining challenges. Speaking with Amie Larter for Inside Retail‘s Retail Untangled podcast series, Widdowson argues that while retailers are racing to adopt AI, many are overlooking the quality of the information powering it.

“The more powerful the technology becomes, the more brutally it exposes the quality of what you’re feeding it,” says Widdowson. “With AI starting to shape how consumers discover products, compare availability, and make purchase decisions before they’ve visited a single retailer’s website, inaccurate or fragmented data has become a commercial liability.”

Many retailers, he says, mistakenly assume AI will identify poor-quality information; he points to the familiar principle of “garbage in, equals garbage out”.

“If you are not using accurate data, your AI engine is going to produce something that’s not accurate. That’s the big thing that people are not comprehending when it comes to data governance.”

Widdowson is equally direct about the consequences. If a business has incomplete, inaccurate or poorly integrated data, “you are in trouble”.

“Imagine the scenario when you walk into a pet food store, and you have a cat, and you get a message saying there is 50 per cent off dog food. That’s a bad experience. That kind of fragmented data and inaccuracy across the end-to-end supply chain and the retailers is probably going to turn you off and make you go somewhere else – somewhere that understands who you are and what your needs are. That’s a big risk across the retail industry.”

How Ulta Beauty excels

He points to US retailer Ulta Beauty as an example of data-driven personalisation done well. Around 44 million customers belong to its loyalty program, representing about 95 per cent of its customer base. Rather than relying on broad customer segments, the retailer tailors communications to individual shoppers.

“When you can achieve that level of personalisation, it works; the opposite, obviously, is the customer goes somewhere else, somewhere that does happen.”

Accurate data also underpins customer trust throughout the shopping journey.

“If the customer is taking the time to research a product before they walk into a retail outlet, and they want to use online retailers to look at the reviews and at what the best product is, then imagine the opportunity. I have had this experience: I’ve done my research, I found the right product, I found out where I could buy it, then I click to buy – or walk into the store – and it is not there. That really is the business risk. If you have created the capability to provide those insights, and the availability, and the pricing, and the delivery information, but then you can’t fulfil on that promise, that’s a bad experience.

“If you are saying your product is available, make sure it is available. If you are saying it is going to cost $10, make sure it costs $10. If you say you can deliver it tomorrow, make sure you deliver it tomorrow.

“Make sure you are leveraging the data, using the insights, and are able to predict exactly what information you should show to the customer. The ideal scenario saying I can deliver it to you tomorrow, but it turns up tonight: That’s the right experience we want, not the opposite.”

The danger of missing links

Beyond data quality, retailers also need to address missing information.

“If you don’t have good data, that’s going to drive a bad result. But the other thing that we’ve seen is gaps in data, which could be internal or external. Maybe you have got missing data for a time period or a part of the country, a distribution centre or a region. 

“If you have a gap in the data, that is where you start to see some of those challenges; where it goes wrong. We have a capability where you can actually create synthetic data to fill in those gaps, which helps. But the other big gap that we are starting to see – and people are realising – is bringing in the external data where it matters.”

One area where AI can quickly influence the bottom line is pricing and markdown optimisation, where many retailers still rely on manual processes or simplistic tools.

“If you make a bad pricing decision, you know that that can have a significant impact on the profitability of the company.”

For example, if markdowns are too aggressive, stock will clear, but margins disappear. Real-time visibility allows retailers to respond faster to customer demand.

“You might think you came up with the best red t-shirt ever on the catwalk, but if customers don’t like it you have to make a decision quickly to say, well, let’s run a promotion, let’s start some pricing activity, because maybe you got it wrong, and if you don’t make those decisions quickly, you are going to end up losing a lot of money. People say the first loss is the best loss, because it’s the least loss. You need the technology that understands that, so you can start to be able to impact that.”

Why trust is vital 

Ultimately, Widdowson believes the retail industry’s success with AI will depend not on adopting the technology itself, but on ensuring retailers can trust the information behind it.

“AI is not going away. AI is embedded in all our solutions. We have been using AI for many years.”

“But you have to be able to trust the data and be able to turn into an action, at scale and at speed.

“There is so much data now, if you aren’t quick enough, your customer is going to the competition. So you have got to leverage that data, leverage the AI capabilities, and connect end to end be able to operate at scale.”

• Listen to the podcast to hear Widdowson warn retailers of the dangers of an in-house silo mentality, where every department believes it has the solution to a problem; and how using hyperautomation and integrating data can enable faster decision-making by creating real-time insights.