Freddie Turner is Managing Director, Chalice AI EMEA and a regular columnist for NDA.
Every peak-trading season brings a new race to adopt the latest marketing technology. This
year, that conversation is dominated by artificial intelligence. Retailers are being promised faster
campaign creation, smarter media buying, better personalisation and automated optimisation
across nearly every channel. But simply adding more AI into the mix does not guarantee better
performance.
The retailers getting the most from AI are starting somewhere else: with a very clear
understanding of the commercial outcome they are trying to achieve, and then building the
technology and data strategy around it.
That matters even more as AI begins to change how people shop. Festive product discovery is
no longer limited to a familiar sequence of search, social, marketplaces and retailer websites.
Consumers can increasingly ask AI-powered search tools for recommendations, comparisons
and answers tailored to a specific need, compressing parts of the discovery and consideration
journey into a single interaction.
For marketers, that means another layer of fragmentation and another powerful intermediary influencing which products make it into a shopper’s consideration set.
As those behaviours become more common, simply having AI in the marketing stack will not be
a differentiator. The bigger question is whose AI is getting smarter from your customers. If all of
the intelligence stays inside someone else’s platform, retailers risk spending their festive
budgets training systems that may ultimately work just as hard for their competitors.
That makes moving intelligence closer to the retailer increasingly important. The models working
on a retailer’s behalf need to learn from its own commercial priorities, customers and data, then
use that intelligence across channels rather than leaving it locked inside individual platforms.
They need to interpret a wider set of signals, react to changing demand and allocate investment
towards the places that can actually influence the commercial result.
Every retailer also enters peak trading with a different definition of success. One may be
focused on driving shoppers into stores during key promotional periods, while another needs to
shift seasonal inventory before year-end. Some are prioritising customer acquisition, while
others are protecting margin or increasing loyalty among existing customers.
Those objectivesrequire very different decisions, even when retailers are working with similar AI capabilities. The advantage comes when the model learns what matters to your particular business rather than applying intelligence shaped by the priorities of thousands of other advertisers.
Consider two apparel retailers investing across many of the same media channels. One wants
to drive younger shoppers into physical stores during a major promotional period. The other
wants to maximise profitability across its ecommerce business because inventory is already in a
healthy position. Applying the same optimisation model to both makes little sense. The desired
outcome should determine what the technology optimises towards and, importantly, which data
is used to inform those decisions.
For the first retailer, that could mean allocating media towards in-store sales rather than relying
primarily on online conversions. Doing that effectively requires stitching together the best
available signals: contextually aligned inventory, in-store sales data, promotional and discount
calendars, and the retailer’s own customer profile data.
AI can then continuously assess thosesignals and allocate investment towards the channels, audiences and moments most likely to deliver the retailer’s actual business goal. Crucially, those learnings become intelligence the retailer can continue to build on, rather than knowledge that exists only within an individual media platform.
We have seen this principle work outside retail too. In one UK campaign for a leading FMCG
brand, weekly sales data was connected to media delivery at postcode-sector level, allowing an
AI-powered allocation model to direct investment towards areas with the greatest room for
incremental growth. The campaign delivered 2.4x the beer and lager ROAS benchmark and
3.75x its incremental sales goal. The important point is not simply that AI was involved. It was
that the model was given a specific commercial outcome and the right business data to optimise
against it.
Traditional performance metrics still have a role. Clicks, conversions and return on ad spend
provide important signals, but they do not always capture the full commercial picture. A
campaign that looks efficient based on online conversions alone may be undervaluing media
that ultimately drives customers into a shop. Likewise, a channel generating a high volume of
sales may be less valuable if those purchases are concentrated around heavily discounted
products with limited margin.
The opportunity is to give AI enough context to understand those distinctions and ensure that
the resulting intelligence is working for the retailer. It can process enormous amounts of
information, identify patterns that would be difficult to find manually and continuously adjust
investment as conditions change. But deciding which signals matter, what outcome they should
ultimately serve and where those learnings live remains a strategic decision for the retailer.
That becomes particularly important during peak trading, when conditions can change quickly.
Consumer demand shifts, competitors launch unexpected promotions, inventory positions move
and individual stores or product categories can perform very differently from one week to the
next. A model learning directly from the retailer’s own customers and commercial data can
respond to those changes far more intelligently than one optimising towards a generic
performance benchmark.
So before adding another AI tool or model to the festive plan, marketers should do two things.
First, define the commercial outcome that matters most in concrete terms. Is the priority in-store
revenue, profitable acquisition, inventory movement, margin or something else? Second, audit
the data the business already has that can help the model understand and optimise towards
that outcome, from sales and inventory to promotions, customer profiles and contextual signals.
The stronger the connection between business data and business objective, the more useful AI
becomes, and the more of that intelligence the retailer can build for itself.
As AI-powered search reshapes discovery and AI becomes part of every marketer’s toolkit,
competitive advantage will increasingly come down to who owns the intelligence. Retailers
should be building models that learn from their own customers, commercial priorities and
business data, rather than allowing that intelligence to remain trapped inside individual
platforms.
This festive season, the question should not simply be how much AI is being used. It should be
whether the AI getting smarter from your business is ultimately working for your business.
