Opens in a new tab

Using AI to Spot Shelf Problems and Guide Field Reps – with Stephanie Lilley of Reckitt

•

A 2024 study found that 29% of approximately 15,000 promotional displays tracked at one retail chain in 2017–2018 never reached the sales floor. The rest were present for only 62% of the campaign period. A missing display may remain unresolved until a field representative’s (rep’s) next scheduled visit.

A 2026 study in the Journal of Business Logistics reported that correcting inventory records increased store-wide sales by approximately 11% over the next two months at one European grocery retailer. The gains were concentrated in products whose recorded stock exceeded actual inventory.

Commercial leaders risk a promotion ending before teams resolve problems limiting sales. Field leaders must decide where a rep’s next visit can make a difference. Timely shelf checks and current store data can help teams choose whether to request replenishment, send a rep to correct a display, or preserve a scheduled visit.

Daniel Faggella, Emerj CEO and Head of Research, hosted Stephanie Lilley, VP Transformation Europe at Reckitt, on the Emerj AI in Business Podcast. Lilley discussed using image recognition to verify promotional displays and recent store data to adjust reps’ visits and tasks, rather than setting priorities four to six weeks ahead.

This article examines three practical ways commercial and field leaders can apply those insights.

  • Verify promotional execution with shelf images: Compare photos of simple displays with the agreed setup, and confirm the system can reliably flag discrepancies before extending checks to full aisles.
  • Prioritize field visits to protect sales: Send reps to stores where they can resolve urgent stock or display problems, while keeping regular visits that support retailer relationships.
  • Tackle the most urgent in-store tasks first: Use recent stock and sales data to decide which tasks reps should tackle first at their assigned store, based on what they can fix with store staff that day.

Listen to the full episode below:

Episode: Closing the Gap Between Commercial Plans and What Happens on the Shelf – with Stephanie Lilley of Reckitt

​

Guest: Stephanie Lilley, VP Transformation Europe at Reckitt

Expertise: Commercial Strategy, Trade Marketing, Retailer Partnerships, Enterprise Transformation

Brief Recognition: Stephanie Lilley leads enterprise transformation across Reckitt’s European markets. She previously led the company’s hygiene portfolio as Regional General Manager and Managing Director for the UK and Ireland and oversaw commercial strategy and major retailer partnerships as Sales Director for the same markets. Earlier, she established Reckitt’s global trade marketing function for Surface Care and managed the Tesco Healthcare and Asda Household accounts.

Verify Promotional Execution with Shelf Images

A centrally agreed promotion can leave commercial teams unsure whether individual stores installed the display or stocked the right products. Lilley describes using image recognition to compare photographs of the display with the agreed setup, helping teams identify where execution needs attention.

The difficulty of that check depends on what a person has to photograph and what the system has to recognize. Lilley gives the example of an end-of-aisle display with one or two products. A rep can take a photo, and the system has a relatively narrow question to answer about whether the display matches the plan.

Checking an entire aisle after a retailer changes its product assortment involves more products and more images. A rushed rep may capture blurry photos or miss part of the shelf, making it harder to assess whether the products are arranged as agreed.

One way for commercial leaders to apply Lilley’s distinction is to define what counts as a usable photo and a correct assessment before expanding coverage. A check for the presence of two products on a promotional display can have a clear pass condition. A full shelf-layout check also needs usable images of the relevant shelf space and enough detail to distinguish similar products.

Before widening coverage, teams can compare the system’s assessments with manual reviews and track how often people need to retake photos or resolve ambiguous results. A missing or unreadable image should trigger another check, since it leaves the status of the display unknown.

Lilley also describes how companies could extend these checks beyond their field teams’ regular visits by paying shoppers to photograph a section of shelving:

“You can get somebody when they are on their shop to take a picture of your bay, and then you can use image recognition software to understand ‘is this driving the compliance that I wanted it to drive?’”

– Stephanie Lilley, VP Transformation Europe at Reckitt

In her illustration of possible coverage, a brand that could afford to send reps to 100 stores might collect images from 1,000 through a crowdsourced model.

The commercial decision would depend on how many of those images are usable, what information the company would have to share with a third-party platform, and whether the team can act on what it finds.

Prioritize Field Visits to Protect Sales

Lilley describes how daily store visits could change when teams use recent data to set priorities:

  • Before — fixed visit cycles: Reps visit assigned stores at agreed intervals, so a stockout outside that day’s route may wait until the next scheduled visit.
  • After — flexible daily routing: Reps receive a daily store list based on company priorities and the expected return from the tasks they can perform, with stops potentially arranged into an efficient route.

At one location, a rep might verify whether stock shown in the retailer’s system is physically available and ask store staff to reorder if it is missing. At another, the rep might work with the retailer to put up a display or correct a shelf layout that has yet to be updated.

Field leaders can assess potential visits by asking whether the issue is still affecting sales, whether the rep can resolve it, and whether the store has the materials and staff support needed to act. A missing display during a live promotion may justify redirecting a visit if the rep can help install it that day. If the materials are unavailable, the immediate action is to arrange their delivery with the retailer.

A stock discrepancy may warrant a visit if the rep can verify availability and work with store staff to replenish the shelf or request a reorder. These checks help leaders prioritize visits with a plausible path to recovering sales opportunities. Teams can then track issue-resolution rates, time to resolution, and travel time to assess whether flexible routing improves field productivity.

Lilley points to European pharmacies as a setting where a fully flexible schedule could come at a cost. Individual pharmacies are important relationships in their own right, and regular visits give reps time to build those relationships.

She describes how a rep could divide the day between relationship visits and stops that respond to current store needs:

“Every day you are going to have four core stores that you need to visit because that’s where you are building relationships. But then you are going to have four flexible stores that are going to be based on where you will get the best return on investment.”

– Stephanie Lilley, VP Transformation Europe at Reckitt

The four-and-four split is an illustration leaders can use to consider which visits require continuity and which could change with store conditions.

Tackle the Most Urgent In-store Tasks First

In some markets Lilley discusses, including parts of Latin America, a company rep may work in one store every day and operate much like a member of the store team. Changing that person’s route offers little benefit.

If current sales data show a product performing well, the rep could ask about adding a display that day. If a campaign is about to go live, the rep could build the display earlier where the store can accommodate it.

Lilley describes AI as a way to turn store-level signals into instructions that reps can act on:

“It is about being more flexible, and it is about being able to react to what you are seeing in the data. And when I say you react to what you are seeing in the data, I mean the AI reacts to what it is picking up in the data and then provides instructions based on that.”

– Stephanie Lilley, VP Transformation Europe at Reckitt

Lilley contrasts retailers that provide large volumes of detailed data with those that offer very little. Commercial leaders can identify which locations provide reliable store-level signals and which still require observations from field visits.  

A delayed stockout report may send a rep to a shelf that has already been replenished, after several days of missed sales opportunities. Even an accurate report can arrive too late to guide the day’s tasks.

Where reps supply the evidence, the team needs a consistent way to capture the store, product, issue, and time of observation before using that information to assign work.

Leaders can trace a stock alert from detection through resolution in each staffing model:

  • Reps covering multiple stores: Check whether the alert changes the day’s route and whether the resulting visit resolves the stockout.
  • Reps based in one store: Check whether the alert changes the order of tasks and whether the rep resolves the stockout with store staff.

Share article

Recommended from Emerj