I suspect other data points would also provide a better picture with regards to processing times. I do not recommend this information is shared due to privacy concerns but all applicants should logically and subjectively consider:
Language test scores
Length of time working for the company
Salary
Health concerns
Nominated occupation in relation to the industry the company operates in
Age
Residential address (statistically some postal codes may have a lower compliance rates than others)
CV
Reference letters
LinkedIn profile and other online platforms
Country
Other citizenships
I have no doubt an individual with a high level of proficiency in English earning a decent salary with a solid history with an established company that was willing to commit to a DE application (and pay the levy) soon after the 457 with no health concerns would make the decision for a CO easier. Extra points if your LinkedIn profile with 500+ connections confirms the above.
The CO is ultimately a resident just like us with KPI's and their own biases that would ultimately play a role. Some of our documentation expired as an example but the CO used their discretion to push it through because ultimately they knew the result would more than likely be the same and there is no need to clog the system for the sake of bureaucracy.For the record, the documents only recently expired and were still within the discretionary time frame.
These systems become more automated over time and all these individual data points start forming models that will be used to evaluate applications. I suspect the reduced processing time post the SAFe levy has more to do with automation due to the reduction in complexity allowing the CO to perform a quick evaluation before moving on.
That's my 2 cents, your mileage may vary. I am a BA, my brain is wired to deconstruct business processes.
Apologies if the above offends anyone, I am simply providing my insight as an individual that has worked on many similar systems. Immigration for Australia is all about boosting the economy and reducing risk. Each application will be weighted based on their economic contribution (less potential cost) with fraud detection mechanisms built in.