Trying to figure out this number is fraught with a lot of difficulty, because it runs into issues of documentation, responsibility, and intentionality.

Historians have access to much, but not all, of Stalin-era archives. For example, in Oleg Khlevniuk's work on the Gulag, he was able to access state prosecutors' documents, and documents in the former Communist Party archives, but not directly many of the documents in the former NKVD archives, although often he was able to work around or piece together this based off of NKVD documents that had been literally carbon copied and sent to these other institutions. And archival documents that are available (many were destroyed) are often scattered across many different sites and post 1991 different countries, so it is a logistical challenge to review the available materials.

Even when you have documents available, they often need a bit of interpretation when you want to assess numbers For example, a certain number of death sentences doesn't equal the number of executions, as death sentences could be stayed or replaced with prison terms. On the other end of things, someone could be given a prison term, have their health terminally ruined, and be released in order to not have their death be on the books. So when an historian is trying to determine how many were killed in the Soviet penal system, you have baseline figures, but these need a bit of interpretation to work out and estimate a number of total victims.

The bigger questions arise when you look outside of official punishment at events like famines - it's much harder to determine famine deaths (often these get calculated as "excess deaths"). Part of the reason for this is because during a famine, not all (maybe not even most) of the victims die from hunger, but die from other causes and complications caused by malnutrition, such as disease epidemics. These also occurred during a time of massive economic dislocation (whether state-directed, such as collectivization and dekulakization, or more voluntary, such as massive immigration to cities), so it can be hard to tell just how accurate records are for people.

Something interesting from Khlevniuk around dekulakization is that some 1.8 million people were deported (either to "special settlements" or camps), and only something like 1.3 million were recorded a year or so later - basically, hundreds of thousands of the healthier and able-bodied just took off - escaped. The Soviet system was often really, really ramshackle. Also: how do we know these people escaped, instead of dying? Khlevniuk cites inquiries from regional governments and party organs specifically asking what to do about escapees, and the NKVD and procuracy sending circular letters weighing the pros and cons about going after such people.

A final point is that even in cases of resettlement or famine, if you can tease out numbers of deaths in any reasonable sense, there is the question of intentionality and responsibility. Most historians accept that the deaths caused by famines are the responsibility of the Soviet government and its collectivization program, but there is debate (notably between Stephen Wheatcroft and R.W. Davies on one side and Michael Ellman on the other) as to the intentionality of the mass deaths. Nowadays no historians, even Ellman, would argue that Stalin deliberately set out to kill millions of people through a famine, and all would agree that the Soviet regime showed a remarkable degree of callousness towards the death and suffering, but the debate is mostly around how much Stalin knew about the deaths, and how OK he was with allowing them to continue.

I would say that overall, academic historians are in the 6 to 9 million deaths range, or close to it. Most of the differences in estimates depend on how famine deaths are estimated, and how much of that is directly the responsibility of Soviet policies. I'd actually be curious to know who is seriously arguing 11 to 20 million based off of detailed research as opposed to just relying on older estimates.

I'm champion Khevniuk a lot here, so I will close with one final argument he makes. He takes the broader view that focusing on "deaths" is kind of misleading in that it implies some sort of "hard", objective number. He takes a broader view and argues for looking a larger category of "victims of repression" at the hands of the regime. This could range from the mild (like losing a job) to the severe (like decades of forced labor or execution), and puts that at something like 1 in 7 Soviet inhabitants. It helps to give a better sense of the scale of what Soviet people were facing in the period.