CTRN AI Site Selection Now Screens Every New Store, at a Self-Reported ~90% Accuracy

Citi Trends (CTRN) has turned AI site selection into a standing step in how it approves new stores. On the 2025-12-02 earnings call, management made new-store site selection the pivot of its expansion story and attached a roughly 90% sales-forecast accuracy to it. By the 2026-08-25 earnings call, the tool had become a routine evaluation that every single new location has to pass — and the company did not repeat that number.

What the tool is and who uses it

CTRN is Citi Trends, a US discount apparel chain. Its revenue comes from merchandise sold in its stores, so where the stores sit and how much each one sells directly set the size of the business. The AI site-selection tool is used by the company's real estate team and by the executives who approve capital for new stores, to decide which markets to enter and which specific locations to open. It combines three years of actual transaction data from every existing store with geolocation studies to describe which customer and market characteristics make a store sell well, then predicts how much sales each candidate location can do and matches candidates against the profiles of the chain's best-performing stores. After the score, a location still has to clear the local real estate team's judgment and the company's new-store financial criteria before it reaches store approval. It sits inside core operations, and customers never see it.

How the application evolved

Across the disclosures, the users, inputs and outputs of this tool have not changed. What changed is where it sits in the decision chain. The 2025-12-02 call was the first disclosure in this window: management described new-store expansion as a combination of three things — AI-driven analytics, local market expertise and strict financial criteria — and said this data-driven approach had demonstrated about 90% accuracy in predicting sales [3]. That number belongs to the whole method. The independent contribution of the AI was never broken out, and the company did not say how accuracy is defined, how many candidate locations it covers, or how long the observation period is.

The 2026-06-02 call fixed the same statement in place. The roughly 90% accuracy was given again word for word, with the subject of the sentence moving from "this data-driven approach" to "this AI data-driven approach," and management used it to explain why it was confident enough to accelerate to approximately 40 new stores in 2027 [2]. The figure still arrived without a sample, an observation period or a baseline, and the same passage used "Beyond the analytics" to set the financial criteria apart — an independent gate still follows the model's score.

Current status

In the most recent quarter the tool has been written into process, yet there is less quotable result attached to it. The 2026-08-25 call said each new location is evaluated using the company's AI tools and held to strict financial return and investment criteria; the technology section listed real estate site selection alongside product allocation as AI tools already in use, and the close summarized the approach as data-driven site selection [1]. The only concrete store-level number this quarter is that nine stores have opened since the fourth quarter of 2025 and all of them are exceeding the company's internal expectations [1] — but the source does not attribute those stores' performance to this tool. The accuracy figure was not repeated this quarter, and that can only be recorded as not disclosed this quarter: the underlying transcript keeps only the speaking turns that were screened in and does not cover the whole call, so there is no basis for inferring that the tool was scaled back or discontinued.

The financial line this would touch first

The metric this tool would most likely move first is sales per store. The chain runs like this: the AI score decides which candidate locations get new-store capital; the selected locations become stores that actually open; if the hit rate really is higher, it should show up first in sales per store once new stores mature; and from there it works through store-level four-wall contribution into revenue growth and operating profit.

What the evidence does and does not support

The evidence is directionally relevant and not yet proven. As of 2026-08-25, management has not attributed any financial line item to this tool. The mature-store sales-per-store target and the four-wall contribution target given this quarter are approval hurdles applied after the AI score — filter conditions, not model output. The gross margin, SG&A and EBITDA improvements this quarter were attributed by management to merchandise margin, lower shrink, and store-level technology and fixed-cost leverage, none of which points to site selection. The three gates — the AI score, local real estate experience and the financial criteria — have never been separated, so no individual store's performance can be booked to the model, and nine new stores is too small a sample open for too short a time.

What to watch next

CTRN's new-store AI site selection has settled into a capital-approval gate that every candidate location must pass, but its contribution to sales per store still rests only on management's qualitative endorsement, with no independently checkable financial evidence. The evidence that would most change that judgment is the company disclosing the basis for the roughly 90% sales-forecast accuracy: how accuracy is defined, how many candidate locations it covers, how long the observation period is, and how much higher it is than the site-selection hit rate before AI was introduced. With that baseline, sales per store as new stores mature would have something to be measured against, and the three gates would become separable.

Application profile

AI-driven site selection analytics for new-store growth. WHO: the real estate team and the executives approving new-store capital. SITUATION: deciding which markets and which specific locations to open as the chain expands toward roughly 650 stores. WHAT THE AI DOES: predicts a prospective site's sales from three years of per-store transaction data plus geolocation studies and matches candidates against the profiles of the chain's most successful stores. Kept separate from the merchandise application because all three identity questions differ — a different audience, a different decision, and a different kind of service.

Business position: core operations · Application stage: limited production · Scope: single business unit · Value type: capital efficiency

Sources

[1] Drillr · CTRN (CTRN) · 2026-08-25 · earnings call

Each new location is evaluated using our AI tools and held to strict financial return and investment criteria.

[2] Drillr · CTRN (CTRN) · 2026-06-02 · earnings call

This AI data-driven approach has demonstrated approximately a 90% accuracy in sales prediction, helping us to identify and replicate our most successful store profiles while minimizing risk as we expand our footprint.

[3] Drillr · CTRN (CTRN) · 2025-12-02 · earnings call

This data-driven approach has demonstrated about 90% accuracy in predicting sales, helping us identify and replicate our most successful store profiles while minimizing risk.