flow

I may be misunderstanding the situation, but are you applying the test across rounds, or only to assess reliability in the final round?

It makes sense to me that a reliability statistic could seem low if it's based on data from multiple rounds, since the point of a Delphi study is that the experts see what other experts think and modify their answers accordingly.

If you're just looking at the final round, then it might be an indication that the experts just disagree about some of the topics you asked about. Maybe you can identify those topics and collect some qualitative data to characterize those disagreements. Generally, finding persistent disagreement can be a valid and interesting result in a Delphi study, like the one below for instance.

Ager, A., Stark, L., Akesson, B., & Boothby, N. (2010). Defining best practice in care and protection of children in crisis‐affected settings: A Delphi study. Child development, 81(4), 1271-1286.

https://srcd.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1467-8624.2010.01467.x?casa_token=xq2mfmx5MD8AAAAA:ii1soEoi-kp6wNiwtMH-TpxaNBPQIeF6iyKHX0I5f1SHc8WB4BmQOu18AyQ1psz6HLAF8PIXOlur0hFq

Hello,

Check the work below to have an example of an applied criteria to select the Delphi ranges:

Article A roadmap for a leanness company to emerge as a true lean organization