It is not possible to get these numbers. Each group is autonomous. Of the many groups I am familiar with none has ever tracked this type of information (over the last 30 some years). If you had 20 or so very dedicated researchers who wanted to attend the meetings of 20 different groups once a week for 5 years,it might be possible to get some good numbers.

I imagine they don't have any data to publish. I can't see how they would collect it.

This article has some good insights into why that is and why previous studies have been biased.

https://www.theatlantic.com/magazine/archive/2015/04/the-irr...

tldr: Most people who respond to surveys asking about the efficacy of the program are sober and skew the results, the anonymity of people in programs makes it difficult to identify who are potential candidates for study/surveys for external researchers, and no central authority monitoring programs so little standardisation between them meaning some groups may be more effective than others (this is somewhat reflected in a comment above about trying other groups which may mean that the real effectiveness comes from the community it brings, not the program itself, but that's pure speculation on my part)

Here's another

https://www.salon.com/2014/03/23/the_pseudo_science_of_alcoh...

Don't they give out tokens for days sober? That volume could be informative.

Well, yeah, but you don't have to collect them. I haven't gotten a token since I passed the one year mark, and it's been six.

They used to. It's 5%.

Effectiveness for quitting cold turkey...is 5%.

No. That number was drawn from an internal report on triennial membership surveys[1] made for a meeting briefing for the board of trustees, and is based entirely on the inability to read a chart (specifically, figure C-1 on page 12). I'll admit, it's a bad chart, and of a poor type for the data it represents. The misreading sparked a "back to basics" movement in AA because it didn't jive with Bill Wilson's off-the-cuff estimate that about half of the people who gave it a really good go (and who weren't psychotic) would recover on their first go, and half of the initial failures would make it on a subsequent attempt. That, of course, raises significant questions around just what, exactly, would constitute giving it a really good go, and how, exactly, was the "psychotic" diagnosis of failures arrived at.

The actual number was closer to 21%... but that only refers to the number of people in their first year of membership who were still members at the completion of 12 months. The problem is that the chart lists membership percentages per month (as a percentage of all members with less than 12 months) using a sort of fever graph with a distinct line for each of the years being compared. That led some people to believe that it was a continuous line following a cohort through their first year. That's impossible, since by that interpretation there are a large number of people in 1986 (for instance) who failed to remain sober for six months but somehow made it to seven. The lines don't start at 100% and work down; there is no indication at all of how many people dropped in and never came back, etc. And do keep in mind that nobody was actually compelled to complete the survey. All it can tell you is that about five percent of the people who have 12 months or less of sobriety in AA during the survey period(s) and who were included in the survey have twelve months.

[1]https://hams.cc/triennial/triennial.pdf

I can't imagine they ever published the data since I can't imagine how they could even collect it.

Anyone can walk in and no one needs to provide any identifying information so how would you be able to track anyone over any period of time?

They used to publish the effectiveness but quit in the early 80's.

It's easy to track anonymous users when the same anonymous users show up for the same meetings for years. AA has had a LOT of meetings over the years, so coming up with a baseline of effectiveness isn't a vexing problem.

Why they quit publishing those numbers isn't exactly a mystery, either.

There's a perfect analogy to religion.

One guy sees X% of the population is a member of faith group A, and X% is a member of faith group B, where A and B are incompatible, therefore the guy is depressed that somehow proves no religion has the one true answer and all are false insert peak nihilism here.

Another guy sees the same data and is elated that the solution set of "happy successful peacefully coexisting religious people" is such a large number provably at least two times X percent.

Which outlook resonates better, depends mostly on your personal model of how you want the world to work, rather than what actually works.

Is it possible the populations in each 5% are disjoint and one should not be too quick to dismiss the program’s results?