As a provider of such solutions to retail for many years, I can tell you that real long-term value is extracted only after much trial and error. This is due in part to the hype and inflated claims by developers of analytics software who push the idea of elaborate systems paying for themselves in every conceivable retail environment.
Counting works, and — as others noted — is a critical metric for stores. Heat mapping works — to a degree — but the insights gained often justify only partial deployment across a chain, especially when you consider the cost of syncing the systems/data with numerous floor plans and changing planograms. Gender, age, and mood detection are bleeding edge — and borderline creepy. The accuracy is mediocre at best, so IMO, these are only worth considering in very specific cases.
Despite all of this, I'm optimistic about the future of such tech, since it gets better every year. We've helped pilot Wi-Fi analytics systems, BLE beacons, stereoscopic cameras, cloud video analytics, and a host of others — and they all have significant limitations. There is no perfect way to analyze brick and mortar shoppers — but every retailer has the challenge of trying to find what works in their environment.