The question
Kalshi lists a contract on Target's monthly foot traffic for August 2026. It settles on September 7 against an index produced by a data company called Carbon Arc. The contract we studied asks whether the index finishes above 94.
What we thought 94 meant
The research brief described the index as one where 100 equals flat year-over-year store visits, which would make 94 roughly a 6 percent decline.
We could not independently verify that. Kalshi's contract terms for this series are not in its public regulatory documents. Carbon Arc publishes no methodology for a foot traffic index. CFTC self-certification filings turned up nothing for the series or for Carbon Arc as a source agency. No historical value of the index has ever been published anywhere we could find.
We built the forecast on the supplied interpretation anyway. Everything below carries that condition.
What Target's actual traffic evidence says
On whether people are walking into Target stores, the evidence is unusually good.
Target's filing for the quarter ending August 1 reports comparable traffic up 3.6 percent with ticket up 0.2 percent. Growth was driven primarily by traffic rather than ticket. Placer.ai, whose Target estimates have tracked reported revenue within about 1.2 points a quarter over five quarters, has visits up 4.7 percent in Q2 and up 7.3 percent in July, the best month of the year.
The comparison being lapped is soft. August 2025 fell 3.3 percent, the seventh straight monthly decline, during the boycott that followed Target's DEI rollback.
Back-to-school helps. The 2026 season set a record at $146.8 billion, and in early August shoppers had finished only 44 percent of their lists, with most of the rest planned for the fortnight before classes started. Online lost share to stores. Target priced 95 percent of its school supplies at or below last year.
One live negative. On August 24 the American Federation of Teachers called for a back-to-school boycott over Target's silence on immigration enforcement in Minneapolis. It landed with seven days left in the month, after most school shopping was done, making it a bigger problem for September.
The forecast we locked
We locked 93 percent YES, median index 104, before looking at any price.
Read it precisely. That is a 93 percent probability that Target's store visits decline by less than 6 percent year over year. It becomes a probability about the contract only if the index measures what the brief said.
What broke
The ladder does not behave the way that interpretation predicts.
Kalshi centred the Target strikes at 94 and the Costco strikes at 97 for the same month. Placer has Costco visits up 8.2 percent in Q2, up 6 percent per location. Under the supplied interpretation Costco should sit near 108. Two unrelated retailers, both measured as growing, both centred at or below 100, with a gap of about 11 points in each case.
We looked for a version of the index that fits everything. A shrinking measurement panel would explain Target and Costco, but then Sweetgreen would not be centred at 114 on a low-growth turnaround. A trailing-average base fails, because August is one of Target's biggest visit months. A 2019 base fails, because Target's visits already run about 8 percent above 2019.
None of the interpretations we tested reconciles those three numbers. That is not proof the brief's definition is wrong. It is enough to stop us relying on it.
Why the 44-point gap is not actionable
The market is extremely thin. Total volume across the Target ladder was $141, with a six-cent spread at our strike. Two contracts in the category showed zero volume, and almost every strike priced between 42 and 59 percent. That is consistent with thin market-making around the listed strikesINFERENCE, though we cannot establish how much genuine directional information sits inside the 49 percent.
The deeper problem is that our number and the market's may not be probabilities of the same quantity. Ours is about Target's store visits. The market price is about whether the Carbon Arc index settles above 94. Same statement only if the index maps to visits as we assumed. Our rules compare two probabilities and act on the difference, and that comparison means nothing until both sides are known to describe the same event.
The decision
PASS.
Not because the gap is small, and not because we lost confidence in the Target research. Because we could not show the research reaches the question the contract asks. The 93 percent is recorded permanently, labelled conditional. No position, on paper or otherwise.
What settlement will teach us
On September 7 the index prints a real number. That will be the first observed value of the series we have ever had, and the first chance to measure its distance from Placer's figure for the same retailer in the same month.
Five more foot traffic contracts settle that day, covering Costco, McDonald's, Chipotle, Shake Shack and Sweetgreen. Comparing them shows whether the distance is a constant, something specific to big-box formats, something retailer by retailer, or no pattern at all. We are not assuming it is uniform until we see it.
What PastBehavior learned
Before assigning a probability you would act on, establish the exact resolution condition, identify the resolution source, and confirm your historical evidence can be translated into the settlement variable. If that last step fails, record the probability for research and pass.
The third condition is the one we skipped, and it is the easiest to skip, because a supplied definition of a metric reads like a fact when it is an assumption. We had good evidence about Target. We never checked whether it was evidence about the contract.
This is Prediction Markets No. 007.