As of October 28, 2026, Circana is arguing that quarterly reviews and static reports are too slow for the pace at which consumer demand, category performance and competition now move. In an article published on October 26, the market research firm said companies that still rely on periodic reporting respond late to shifts they could have seen earlier, and it pointed to its Liquid AI platform as a way to combine consumer, shopper, retail and market data with predictive analysis. For brands managing availability across store networks, the practical claim is narrower and more useful than the headline: the interval between reads is itself a business decision, and a quarterly interval hides stockout patterns that a weekly interval exposes.
Quarterly reviews leave availability gaps open for weeks
Availability problems have a shape, and the shape is invisible at quarterly resolution. A distribution gap opens when a store runs out of a SKU and does not receive a replacement within the normal replenishment cycle. If the gap is measured once a quarter, a two-week stockout registers as a single data point with no indication of when it started, how long it lasted or whether it recurred.
At weekly resolution the same gap becomes a pattern. Brands can see whether the gap began after a promotion ended, whether it clustered in specific stores or regions, and whether it repeated in the same SKU. None of those questions can be answered from a quarterly snapshot, which is why Circana's argument is less about artificial intelligence than about read frequency.
The 1,500 hours an Australian retailer recovered
Circana cites an Australian retailer that saved about 1,500 hours a year on recurring brand performance analysis by using Liquid AI, with the freed time redirected to strategic work. The number is worth reading carefully: 1,500 hours is roughly three-quarters of a full-time employee's annual hours, which suggests the saving came from eliminating repeated manual assembly rather than from a single analytical breakthrough.
That distinction matters for anyone planning a similar programme. The hours are usually spent pulling data from multiple sources, reconciling definitions and rebuilding the same report with a new date. Automation removes that work, but it does not remove the need to decide what the report should measure. A team that automates an unclear metric gets an unclear metric faster.
Assistant access changes who reads the numbers
Circana says it can make its data available through assistants including Copilot, ChatGPT, Claude and Gemini using Agent2Agent and Model Context Protocol, technologies designed to connect AI systems and data sources. This is the part of the announcement with the widest operational consequences, because it moves analysis from a specialist function to a conversational one.
When a category manager can ask a question in plain language and receive an answer drawn from retailer-authorized POS data, the bottleneck shifts from report production to question quality. The risk shifts too: an assistant that answers without exposing its filters can produce a confident number built on the wrong store set, and the person receiving it may not have the context to notice.
For availability monitoring specifically, assistant access is most useful for exception review rather than routine reporting. Asking which stores moved from on-shelf to out of stock in the last seven days is a bounded question with a verifiable answer. Asking how the category is performing is not, because the answer depends on definitions the assistant may not surface.
Three signals that move before a stockout
If reads move from quarterly to weekly, the metrics should change too. Availability rate alone is a lagging indicator: by the time it falls, the sale has already been lost. Three upstream signals tend to move first, and all three are observable from data brands already receive.
| Signal | What it shows | Lead time |
|---|---|---|
| Order fill rate | Share of ordered units delivered | One to two cycles |
| Shelf facings | Space held versus planogram | Two to four weeks |
| Replenishment cadence | Days between deliveries per store | One cycle |
Order fill rate is the earliest of the three. When a retailer's orders are only partly filled, the shelf has not yet emptied, so availability still looks healthy while the pipeline is already short. Watching fill rate rather than availability converts a lagging measure into a leading one without adding new data sources.
Shelf facings move more slowly but carry a different warning. A brand losing facings is losing distribution space before it loses sales, and the loss is usually negotiated rather than accidental. Tracking facings against the planogram catches the drift while there is still a commercial conversation to be had.
Why POS data alone is not enough
Retailer-authorized POS data tells you what sold, not what was available to sell. A store that stocks three units and sells three units shows the same sales as a store that stocks thirty and sells three, and the first store is a stockout risk while the second is a demand problem. Without an availability measure alongside sales, the two are indistinguishable.
This is where the combination Circana describes becomes relevant. POS data shows the outcome, category shopping patterns provide the expected baseline, and availability measures show the constraint. Any two of the three produce a plausible but incomplete story, which is why single-source dashboards tend to generate confident conclusions that do not survive a store visit.
Building a weekly availability read
A weekly read does not require a new platform. It requires fixing three things: a stable store sample, a fixed metric definition and a comparison baseline. The store sample should be large enough to represent each region and stable enough that week-over-week changes reflect reality rather than sample churn.
The metric definition matters more than most teams expect. Availability can mean in-stock at the shelf, in-stock in the store backroom, orderable in the retailer's system or shippable from a distribution centre. Each definition produces a different number, and mixing them across weeks produces a trend line that means nothing.
The baseline should be seasonal rather than absolute. A store that runs at 94% availability in a normal week and 91% during a promotion is not necessarily failing; a store that drops from 97% to 91% while its region holds at 96% is. Comparing each store against its own history and against its region separates structural problems from temporary ones.
FAQ
Does a weekly read require new data sources?
No. Order fill rate, availability and facings can all be derived from data a brand already receives. What changes is the interval and the discipline of using one definition consistently.
Why is order fill rate a better early signal than availability?
Because the shelf has not emptied yet when fill rate falls. Availability still looks healthy while the pipeline is short, so fill rate turns a lagging measure into a leading one.
What is the risk of querying data through an AI assistant?
Answers may not expose their filters. A confident number built on the wrong store set is hard to detect, so assistant output is best used for bounded exception questions and verified against source data.
How should the store sample be built?
Large enough to represent each region, stable enough that week-over-week change reflects reality rather than sample churn, and documented so the same rules apply every period.
