After working through dozens of Halo SAF implementations, the most common failure pattern isn't technical — it's scope. Organisations either model too much (every asset, every dependency, whether it matters or not) or too little (a few vague services with no ownership data attached). The golden rules exist to stop both.
Before you define a single service, you need to understand the ecosystem around it: who provides and supports it, who owns and governs it, and who actually requests it. Then you model only what's necessary — what has an owner, what supports an ITSM process, and what delivers a tangible business outcome.
The two questions that cut through almost every modelling debate are these: does this service need ITSM support, and would it survive a modelling review in six months? If the answer to either is no, leave it out.
A lean, well-governed SDM outperforms a bloated one every time.
There is also a practical rule of thumb worth applying at every workshop: if you can't name the owner of a service within ten seconds of being asked, it's not ready to be modelled. Ownership is the foundation. Without it, the service data model becomes infrastructure with no accountability attached — and a model with no accountability is just as useless as no model at all.