Honeypots are increasingly used in Industrial Control Systems (ICS) to divert attacks from critical assets and study malicious behavior. While prior work has examined specific aspects of ICS honeypot design, a comprehensive understanding of cost-effective deployment strategies is still lacking. This work investigates how interaction level, network type, and geographic location affect the attractiveness of ICS honeypots. We deploy both low- and high-interaction honeypots, alongside a physical device, across corporate and cloud networks in various geographic regions. We collect and analyze network interactions involving HTTP, S7Comm, and Modbus protocols from 16 honeypots with diverse configurations over a three-month period. Our results show that network type has the largest impact on ICS honeypot traffic, while interaction level and geographic location play a minor role. We also find that low-interaction honeypots capture traffic comparable to high-interaction setups, supporting their use for general threat intelligence.