Child enticement of minors occurs when an adult approaches a child online for sexual purposes, often through a manipulative trust process known as grooming. Due to increased caseloads, machine learning (ML) and natural language processing (NLP) researchers are developing triage tools to help law enforcement officers (LEOs) identify potential enticement conversations. However, the datasets most readily available consist of Internet sting data where an officer or vigilante, known as a decoy, imitates a minor while talking to a potential solicitor. Psychology researchers note motivations of decoys and LEOs differ from those of at-risk minors and thus the conversations are likely incongruent. Previous research in the area is limited. Existing studies either do not include all three dataset types (victim, decoy, and LEOs) or examine usage of grooming stages and grooming strategies by each group, but not how these constructs intermingle. We extend the literature by examining the usage and sequencing of grooming stages in a chat, and the use of grooming strategies in grooming stages for all three groups. We find the greatest differences in the flow of grooming conversations with decoys, victims, and LEOs exist within the risk assessment, meeting, and sexual stages. Implications for LEOs, ML and NLP researchers are discussed.