Merchandising & inventory
Computations for sell-through, stock and SKU performance must be consistent across systems and regions. (See example belor of same-store sales.)
Industry · Retail and consumer
In retail, comparable store sales needs to be right, every time.
Blue Mercury trusts AtScale as their universal semantic layer. AtScale for AI defines each retail metric once, so anyone using OpenAI or Gemini or any BI tools gets merchandising and finance numbers right, every time.
Use cases
Where do retail teams need accurate semantics?
AtScale computes each metric consistently for every analyst using AI and every AI agent you develop. Without a unified semantic layer, teams can get different answers to the same question, eroding confidence in AI.
Customer 360
Marketing and merchandising get one definition of customer value across CRM, POS, e-commerce and loyalty systems.
Omnichannel
In-store, online, marketplace, and partners. AtScale gives you one view of sales based on one model for every channel.
Demand forecasting
Forecasts and planning models get one number so they match.
Finance
Finance closes the month based on the same sales, margin and inventory figures merchants report.
Promotions and markdowns
A semantic layer ensures promotion lift and margin after markdowns is measured the same way by merchants and finance.
How does AtScale keep same-store sales the same for every AI agent?
Comps are up for merchandising and flat for finance? Guidance goes out Thursday, and you need to pick the right one.
Merchandising, finance and store operations each calculate same-store sales differently, and each method is reasonable. Without a semantic layer, an AI agent might picks the wrong one. AtScale computes it one, validated layer with full knowledge of based tenure thresholds, remodel rules, closure rules and the 4-5-4 calendar.
What a buyer prompts and what they meant can be two different things
Humans using AI ask vague prompts. Before AtScale computes anything, we use our knowledge of your semantics to disambiguate prompts so AI gives the right answer, every time.
| What a person types | What they meant |
|---|---|
| “How did sales grow?” | Same-store sales, stores open a full year, remodels kept in, on the 4-5-4 calendar, against the same weeks last year. |
| “Which SKUs are losing us money in the Southeast?” | SKUs whose margin after markdowns fell this quarter, in the Southeast as finance draws it. |
| “Did the promotion work?” | Did basket size rise during the promotion, net of demand pulled forward from the following month? |
AI agents in retail run on same-store sales metrics
When an agent asks for “same-store sales,” it asks the AtScale Ava API first and selects the right metric. Then the agent uses business computations your team has confirmed as the right ones. Some of these decisions can be fully automated, and some can be require a person to approve before actions are taken.
| What the agent is told | What ACE calculates | Who acts |
|---|---|---|
| Every Monday: “Post weekly comps to the store managers’ channel.” | Same-store sales on the 4-5-4 calendar, stores open a full year, remodels kept in, against the same weeks last year. | Fully automated. The agent posts on its own. A wrong post is easy to correct. |
| Every night: “Mark down any Southeast SKU whose margin after markdowns fell this quarter.” | Margin after markdowns by SKU, in the Southeast as finance draws it. | Person approves. The agent drafts the markdown list. A merchant approves it before prices change. |
| End of each promotion: “Rebook the promotions that paid for themselves.” | Basket-size lift during each promotion, net of demand pulled forward from the next month. | Person approves. The agent proposes next month’s promo calendar for a planner to sign off. |
Proof points from this sector
