User retention rates on MathOverflow

A two-input table :
First input : number of users with more than $500, 1000, 2500, 5000$ or $10000$ reputation points.
Second input : number of users not seen for more than $1/12, 1/4, 1/2, 1, 2$ or $3$ years.

On September 10th 2013:

$\begin{array}{c|c|c|c|c|c|c} &\text{seen in last month}&1/12 & 1/4 & 1/2 & 1 & 2 & 3 \newline \hline 500 &891&345& 202 & 136 & 82 & 37 & 10 \newline \hline 1000 &592&152& 88 & 62 & 35 & 18 & 4 \newline \hline 2500 &310&43& 28 & 22 & 14 & 7 & 2 \newline \hline 5000 &171&19& 15 & 11 & 8 & 4 & 2 \newline \hline 10000 &87&7& 7 & 5 & 5 & 2 & 1 \end{array}$

How explain the results of this table ?

Remark : This table has been completed by using this query.

On March 6th 2015:

$\begin{array}{c|c|c|c|c|c|c} &\text{seen in last month}&1/12 & 1/4 & 1/2 & 1 & 2 & 3 \newline \hline 500 &1096&494& 383 & 305 & 241 & 114 & 50 \newline \hline 1000 &730&231& 168 & 139 & 100 & 53 & 25 \newline \hline 2500 &391&73& 55 & 46 & 30 & 15 & 8 \newline \hline 5000 &208&36& 27 & 22 & 14 & 8 & 4 \newline \hline 10000 &109&17& 13 & 10 & 5 & 4 & 3 \end{array}$

On September 7th 2016:

$\begin{array}{c|c|c|c|c|c|c} &\text{seen in last month}&1/12 & 1/4 & 1/2 & 1 & 2 & 3 & 4 \newline \hline 500 & 1189 & 734 & 559 & 469 & 367 & 263 & 191 & 74 \newline \hline 1000 & 797 & 330 & 251 & 212 & 162 & 116 & 81 & 35 \newline \hline 2500 & 453 & 122 & 89 & 77 & 59 & 43 & 24 & 11 \newline \hline 5000 & 247 & 46 & 34 & 30 & 22 & 17 & 12 & 7 \newline \hline 10000 & 130 & 19 & 16 & 13 & 10 & 7 & 5 & 4 \newline \hline 25000 & 35 & 3 & 2 & 1 & 1 & 1 & 0 & 0 \end{array}$

On September 16th 2017:

$\begin{array}{c|c|c|c|c|c|c} &\text{seen in last month}&1/12 & 1/4 & 1/2 & 1 & 2 & 3 & 4 & 5 \newline \hline 105 & 3232 &5540 &4428 &3682 & 3003 & 2210 & 1674 & 1287& 688 \newline \hline 250 & 1896 &1813 &1415 &1196 & 998 & 776 & 590 & 466& 202 \newline \hline 500 & 1310 & 830 & 642 & 540 & 451 & 349 & 266 & 202 & 84 \newline \hline 1000 & 888 & 365 & 274 & 230 & 198 & 154 & 117 & 85 & 40 \newline \hline 2500 & 507 & 124 & 93 & 80 & 72 & 55 & 40 & 25 & 13 \newline \hline 5000 & 288 & 49 & 36 & 30 & 26 & 19 & 15 & 11 & 8 \newline \hline 10000 & 146 & 21 & 16 & 14 & 13 & 12 & 8 & 6 & 5 \newline \hline 25000 & 50 & 4 & 2 & 1 & 1 & 1 & 1 & 0 & 0 \newline \hline 50000 & 12 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \end{array}$

• Next challenge: find a (nice) function fitting the data, and use it to predict the numbers of users at each point level who will absent themselves in the future. Then check back in a year to see how good those predictions were. Sep 8 '13 at 23:41
• I added the first column, showing users seen during the last 30 days. It seems impossible to interpret the rest of the data without this column! Sep 9 '13 at 0:03
• One thing that table won't show is the slow-down in visits. I still visit MO, but much more sporadically than before, and interact less when I'm here. There's a complicated reason involving frustrating behaviour of my browser around script-heavy sites in combination with my university's proxy which I don't understand. This was a problem with the SE network as it is now, and wasn't in MO under SE1.0, so I stopped fighting and slowed my visits down. (Edit: just posting this comment triggered said behaviour...) Sep 9 '13 at 6:53
• @David: I hope that the said behavior is not a system admin coming into the room and smacking you on the head with a teaspoon!
– Asaf Karagila Mod
Sep 9 '13 at 8:56
• I don't understand at all why this question was closed as off-topic. I've created an answer on the 'Requests for reopen votes' threads meta.mathoverflow.net/a/798/3 to discuss this. Sep 11 '13 at 0:24
• To me it looks like (up to a rescaling) this data says that it's "similarly difficult" to get high reputation as it is to quit MO. Sep 11 '13 at 19:48

It is not quite clear to me what type of explanation there could bein this form beyond the "obvious" one:

Some people use MO for a while and then stop doing so for some reason.

Just this might be considered as a "stupid" answer so some more thoughts.

1. I do not see anything unusual about the data, it seems just distributed as one would expect. The more points the less quits; the shorter the period for 'temporal(?) quit' the more.

2. I cannot find anything that unusual about people stopping to use MO for a while or at all. There are not few things in my life (mathematical and otherwise) that I did for some time an then stopped doing. Sometimes there was a good reason, sometimes it just happened. To give an example, for some time I really like to read the Notices of the AMS and read essentially all articles. Now, I read them rarely. I could not say why exactly. Or, for a while I browsed MathSciNet a lot, for fun mainly; I don't anymore. Also, my MO usage is not uniform. Sometimes I turned it down for a specific reason, sometimes I just have less time, sometimes I just do other things or do not happen to be in front of a computer that much.

3. For specific users one could guess or sometimes also know for a fact why they stopped using MO. For some others it is not clear to me and for very few I would be curious. But this is not really a data-driven discussion but would be asking individuals about their motivation. I guess the reasons one might hear would be contradictory (from one user to the other).

In brief, it could be interesting to know some people's motivations, but I doubt we can infer them from data. Also, I doubt it is a good idea to speculate regarding some specific users motivations.

• Another remark, tho not one we like to think about, at least two 10k users have passed away during MO's history... Sep 8 '13 at 21:12
• Yes this is mainly what I had in mind with the "now for a fact." I did not make it explicity as I was worried to formulate it not well and uninenionally make it not tactfull enough.
– user9072
Sep 8 '13 at 21:24
• My MO use has slowed down a lot, but not really for any important reason. Sep 8 '13 at 23:04
• I totally agree with the final two sentences of quid's answer. This thread seems a bit pointless to me unless someone collects reasons directly from the "deserters" why they "deserted". (I've been busy! Leave me alone! (-: )
– Todd Trimble Mod
Sep 9 '13 at 0:08
• I agree that the data does not look at all like it needs any special explanation. In any online community, people tend to just stop participating for one reason or another (or none), and the numbers here do not seem to me to be large at all. Sep 9 '13 at 6:40
• Of the 5 living 10k users on the list of missing people, 2 essentially stopped participating after racing to 10k in the first six months of MO, one is (I think) busy editing Wikipedia, and I don't know the other 2 personally.
– S. Carnahan Mod
Sep 9 '13 at 23:52

This seems to be asking for rather a lot! Why don't you first produce this table, using either the Stack Exchange API or the Data Explorer. Both will provide the answers to your questions. After that, we can get to analysis and explanations!

• Thank you Scott for this information about Stack Exchange API and Data Explorer. Sep 8 '13 at 12:05
• I have edited an empty table. Could you explain a little to me how does these sites work for having these informations ? Sep 8 '13 at 14:18
• The basic help page is here: data.stackexchange.com/help The language used is T-SQL, which is a Microsoft variant of IBM's SQL language. It might be easiest to start by modifying other people's scripts, and reading on-line sql tutorials. Sep 8 '13 at 14:34
• That's it, the table is completed ! Sep 8 '13 at 20:20
• The table doesn't count the deleted accounts. Sep 11 '13 at 20:13