The headline number is real, and so is the caveat
The National Retail Federation's newest retail security survey, published in late July, found shoplifting and merchandise theft incidents declined 12.4 percent year over year, the first meaningful drop after several years of retailers reporting rising in-store theft. David Johnston, NRF's vice president for asset protection and retail operations, attributed the shift to sustained investment in technology and employee training rather than a single cause. The survey covered 66 retail companies representing 143 brands and roughly $1.7 trillion in combined 2025 sales, a large enough sample to take seriously.
The caveat is that outside data tells a messier story. Reported shoplifting across 36 major U.S. cities rose 4 percent in the first half of 2026 compared to the same period a year earlier, and 5 percent compared to 2019 levels, according to figures cited alongside the NRF release. NRF itself has a credibility scar here: it retracted an organized retail crime loss estimate in 2023 after reporters found the underlying math was wrong. Read the stabilization claim as directionally credible, not as a precise number.
Where the crime actually moved
The more useful finding in this survey is where the risk migrated once physical theft itself started to stabilize. Sixty nine percent of surveyed retailers reported an increase in phone scams targeting store associates, typically social engineering attempts to authorize refunds or divert shipments to a fraudster's address. Fifty one percent saw loyalty program fraud increase, and 42 percent reported growth in gift card fraud and theft. Forty percent of retailers said they are dealing with organized retail crime, and 37 percent are seeing more walkout or pushout theft specifically, so the older physical categories have not disappeared, they have simply stopped growing as fast as the newer ones.
Read Hayes of the Loss Prevention Research Council summarized the shift plainly: as physical theft risk stabilizes, the industry has to confront fraud channels that grew up around digital and remote interactions instead. Tony D'Onofrio of Sensormatic Solutions framed the response the same way most of this survey's respondents did, as a data and connected-technology problem that spans both the store floor and the call center, rather than a staffing problem that more guards or cameras alone can solve.
The AI adoption gap is the real story for tech leaders
Buried in the numbers is a statistic that should worry anyone running fraud or risk tooling at a retailer: only 37 percent of retailers currently use generative AI for fraud prevention, even as the fraud categories growing fastest, phone scams and account-level abuse, are exactly the categories generative AI and adaptive machine learning are best suited to catch. Adaptive systems already in production elsewhere reduce false positives by as much as 85 percent while roughly doubling detection of compromised payment credentials, according to figures cited alongside the survey. That gap between where the risk is moving and where the tooling investment has actually landed is the clearest signal in the whole report.
Eighty five percent of merchants surveyed cited preventing fraud without degrading the customer experience as their central challenge, which is the correct framing for anyone building this stack in 2026. Rules-based fraud systems built for a shoplifting-dominant threat model tend to over-flag legitimate returns and loyalty activity, frustrating good customers while missing the phone-scam and account-fraud patterns that do not look like theft at all, since they involve a person on the phone convincing a legitimate employee to authorize an action rather than an anomalous transaction a rules engine would ever catch.
Why the old loss prevention budget does not fit anymore
Most retail loss prevention spending was built around physical deterrence: cameras, EAS tags, guards, and store hardening. That spending appears to be working, per NRF's own numbers, which makes it tempting to declare victory and hold budgets flat. The survey's own data argues against that. The fraud categories rising fastest, phone scams, loyalty abuse, gift card fraud, live in call centers, mobile apps, and account systems that physical security tools do not touch at all.
This is a governance and org-design problem as much as a technology one. Loss prevention teams historically report through store operations or security, while fraud and payments risk often sit in a completely separate function under finance or digital. The retailers that closed this gap fastest are the ones treating loss prevention and digital fraud as one connected data problem with a shared detection stack, not two departments comparing separate dashboards.
What to do with this before next quarter
If your fraud detection is still primarily rules-based, this survey is a credible external benchmark to bring into a budget conversation. The 85 percent false-positive reduction figure from adaptive machine learning systems is a concrete number to test against your own vendor's claims, and the fact that fewer than four in ten retailers have deployed generative AI for fraud at all means being an early mover here is still a real differentiator, not a catch-up move.
The harder work is organizational. Pull loyalty fraud, gift card fraud, and phone-scam incident data into the same review as shoplifting and organized retail crime metrics before the next planning cycle. If those numbers currently live in different systems owned by different teams, that gap is exactly the blind spot this survey just quantified, and it will keep growing faster than the physical theft numbers you are already tracking closely.
