This paper investigates bid-evaluation practices and derives practical fraud indicators through a combination of expert interviews and text mining of interview transcripts. A total of 18 Malaysian procurement officers were interviewed, and the transcripts were analysed using Voyant to identify terms and patterns associated with collusion and other irregularities. The extracted insights were subsequently refined into potential variables for incorporation into a future machine learning–based detection tool. Business intelligence methods were applied to transform unstructured interview data into structured indicators suitable for automated screening. As an initial validation, association-rule mining was conducted on a 2015 ePerolehan dataset subset. The identified co-bidding links were found to align with the signals derived from interviews, indicating the practical value of rule-based screening approaches. This study constitutes a requirements elicitation effort that captures practitioner needs while proposing potential indicators for a future bid-rigging detection model. The modelling demonstrated is preliminary in nature, serving to illustrate how practitioner-derived signals can be operationalised into measurable features. Comprehensive validation using labelled investigation data is planned for future research.