Editor usage stats
This page has been thoroughly redesigned in mid-2012. If you are looking for the older raw statistics, check the archive page.
Note that most plots on this page include at least one logarithmic scale. This may be unfamiliar to some.
Remark on updates. The information on this page has been compiled by Oli-Wan. It will occasionally be updated, but I cannot promise to maintain regular intervals. If you think another update is past due, please remind me.
Contents |
The "market share" of an OSM editor can be defined using different weights: the number of changesets created, the number of its users, or the number of edits actually made, and various others (including combinations of the former). Each definition has its advantages and disadvantages, but each also tells something about how a given editor is (or was) used.
In the tables below, the market shares of various editors according to the aforementioned three definitions are presented, broken up into years. For the corresponding plots, the time resolution is six months.
Methodology
If you are not interested in technical details, feel free to skip this section.
All the information presented on this page is extracted from changesets-yymmdd.osm.bz2 files. An editor is identified by the created_by tag of each changeset. (This implies that 2009 data cover only the time following the startup of API 0.6, when changesets - and changeset metadata - were introduced, and that statistics for earlier years cannot be generated in this simple way.)
In all the tables, 2013 means "2013 so far". For obvious reasons, full-year statistics cannot yet be generated.
The tables and plots show, respectively, the number of changesets created using an editor (a simple counting exercise in processing the changesets file), the number of total edits made with it (computed by summing up the num_changes of each changeset), and the number of distinct user IDs (uids) seen in association with each editor. The latter is assumed to correspond to the number of actual users/mappers (though some may use more than one account, which is not considered here). The data are resolved by years for the tables and by half years for the plots. Market shares (as percentages) are calculated with respect to the overall sums.
In the "distinct uids" table, mappers editing with more than one editor will be counted once for each editor. This double-counting implies that the sum of all is not equal to the total number of mappers editing in a given period of time. The "market shares" in the corresponding plot are, however, calculated in relation to that number. Therefore, the market shares do not add up to one; instead, their interpretation is as follows: if an editor has, say, a market share of 0.5 in a given period of time, 50 % of all mappers have used that particular editor at least once in that period - but some of them may have used other editors as well. For example, in the first half of 2012, Potlatch 2 and JOSM had respective market shares of roughly 85 % and 36 %, the sum of which already largely exceeds 100 %.
The double-counting of users could in principle be remedied by "splitting" users according to some sort of weight, for example, the number of changesets created with each editor, but that would alter the definition and distort the distribution, diluting the desired information about the user basis. Therefore, such is not done here.
Observations
It is obvious that according to both the number of changesets and the number of edits, JOSM has never been challenged as the number one editor (the only nominal exception in 2009 being a special case, when huge numbers of superfluous tags from the TIGER import were removed by woodpeck_fixbot using osmtools). JOSM's dominance is not so clear in terms of changesets created - this has to do with Potlatch 1 creating zillions of empty changesets in live editing mode, but also with the fact that Potlatch users tend to make smaller changesets on average.
Also, the common notion that Potlatch "wins" by the number of users, is confirmed - with Potlatch 1 being largely superseded by Potlatch 2. In fact, the number of actual edits made with Potlatch 1 has gradually decreased with increasing popularity of Potlatch 2, dropping to the permille level in 2012. Before being removed from the standard choices on the website's edit button, Potlatch 1 was still used by quite a number of users, possibly trying Potlatch 1 out of pure curiosity.
One may also be surprised by the small market share of Merkaartor - often considered one of the "big three" editors. Its declining popularity may relate to its stalled development. By now, Merkaartor has fewer users than Go Map!! and Vespucci.
Most other editors have never reached percent-level market shares (the exceptions being import scripts, used by only few users but editing large amounts of data). Recently, some smartphone applications have obtained a non-negligible audience (making very few actual edits per changesets, which of course reflects their goal of filling the niche for quickly editing, e.g. the opening hours of some POI).
On the other hand, JOSM's market share in terms of users has continually decreased over the past years (unlike its market shares by changesets and edits, which have remained roughly constant). The number of mappers has grown, but the number of JOSM users has not kept up with the pace. In other words: a growing percentage of users has been using exclusively Potlatch. It is not clear at this point whether this trend derives from an increase in the number of actual mappers, who use only Potlatch for their (presumably occasional) edits, or from an increase in the rate of one-time contributors (nearly all of whom use Potlatch).
Tables and figures
by number of changesets
| editor | 2009 | 2010 | 2011 | 2012 | 2013 |
|---|---|---|---|---|---|
| JOSM | 1 201 159 (46.8 %) | 1 759 942 (53.3 %) | 1 976 916 (57.7 %) | 2 398 347 (56.9 %) | 935 432 (58.9 %) |
| Potlatch 2 | - | 58 154 (1.8 %) | 940 064 (27.5 %) | 1 340 965 (31.8 %) | 508 161 (32.0 %) |
| Merkaartor | 136 926 (5.3 %) | 172 561 (5.2 %) | 149 106 (4.4 %) | 111 796 (2.7 %) | 29 358 (1.9 %) |
| Pushpin | - | - | - | 10 060 (0.2 %) | 28 853 (1.8 %) |
| iD | - | - | - | - | 28 321 (1.8 %) |
| Go Map!! | - | - | - | - | 13 404 (0.8 %) |
| Vespucci | 673 (0.0 %) | 4 141 (0.1 %) | 9 036 (0.3 %) | 17 380 (0.4 %) | 10 336 (0.7 %) |
| Potlatch 0.x/1.x | 1 073 205 (41.8 %) | 1 068 397 (32.3 %) | 215 062 (6.3 %) | 52 644 (1.2 %) | 4 885 (0.3 %) |
| MapStalt Mini[1] | - | - | 3 177 (0.1 %) | 4 789 (0.1 %) | 3 795 (0.2 %) |
| OsmAnd | - | 4 784 (0.1 %) | 11 445 (0.3 %) | 7 284 (0.2 %) | 2 589 (0.2 %) |
| Roy | - | - | - | 8 660 (0.2 %) | 2 525 (0.2 %) |
| iLOE | 1 325 (0.1 %) | 4 236 (0.1 %) | 14 393 (0.4 %) | 16 954 (0.4 %) | 1 298 (0.1 %) |
| OpenMaps | - | 8 987 (0.3 %) | 1 745 (0.1 %) | 3 463 (0.1 %) | 350 (0.0 %) |
| osmapis | - | - | - | 14 272 (0.3 %) | 261 (0.0 %) |
| ArcGIS Editor | - | 293 (0.0 %) | 3 809 (0.1 %) | 10 193 (0.2 %) | 226 (0.0 %) |
| upload.py | 35 808 (1.4 %) | 830 (0.0 %) | 281 (0.0 %) | 443 (0.0 %) | 179 (0.0 %) |
| Mapzen | 5 659 (0.2 %) | 30 745 (0.9 %) | 10 119 (0.3 %) | 1 417 (0.0 %) | 154 (0.0 %) |
| FreieTonne | 9 576 (0.4 %) | 9 717 (0.3 %) | 5 941 (0.2 %) | 5 158 (0.1 %) | 131 (0.0 %) |
| osmtools | 10 446 (0.4 %) | 5 121 (0.2 %) | 312 (0.0 %) | 505 (0.0 %) | 118 (0.0 %) |
| Redaction bot | - | - | - | 141 693 (3.4 %) | 48 (0.0 %) |
| PythonOsmApi | 946 (0.0 %) | 17 091 (0.5 %) | 7 809 (0.2 %) | 2 346 (0.1 %) | 45 (0.0 %) |
| FindvejBot | 293 (0.0 %) | 117 489 (3.6 %) | 48 021 (1.4 %) | 18 884 (0.4 %) | 21 (0.0 %) |
| bulk_upload.py | 62 064 (2.4 %) | 11 889 (0.4 %) | 1 317 (0.0 %) | 319 (0.0 %) | 18 (0.0 %) |
| OSMPhpLib | - | - | - | 9 201 (0.2 %) | - |
| KMLManager | 17 443 (0.7 %) | - | - | - | - |
| osmitter | - | - | 9 663 (0.3 %) | 1 (0.0 %) | - |
| Other | 12 204 (0.5 %) | 29 297 (0.9 %) | 15 690 (0.5 %) | 41 054 (1.0 %) | 16 380 (1.0 %) |
by number of users (distinct uids)
| editor | 2009 | 2010 | 2011 | 2012 | 2013 |
|---|---|---|---|---|---|
| Potlatch 2 | 0 | 3438 | 33743 | 59514 | 32349 |
| JOSM | 10054 | 14762 | 16720 | 20190 | 12218 |
| iD | 0 | 0 | 0 | 0 | 3485 |
| Go Map!! | 0 | 0 | 0 | 0 | 912 |
| MapStalt Mini[1] | 0 | 0 | 512 | 1167 | 856 |
| Vespucci | 58 | 235 | 457 | 952 | 697 |
| Merkaartor | 2156 | 2680 | 2139 | 1670 | 632 |
| rosemary | 0 | 0 | 0 | 758 | 606 |
| OsmAnd | 0 | 191 | 636 | 1119 | 511 |
| Pushpin | 0 | 0 | 0 | 389 | 484 |
| Potlatch 0.x/1.x | 18112 | 29480 | 16298 | 11652 | 387 |
| OpenMaps | 0 | 607 | 674 | 974 | 190 |
| OSMPOIEditor | 0 | 0 | 0 | 335 | 190 |
| RawEdit/Osmose Raw Editor | 93 | 75 | 100 | 181 | 88 |
| OSMapTuner | 0 | 0 | 0 | 241 | 86 |
| iLOE | 131 | 326 | 330 | 341 | 73 |
| QGIS | 34 | 156 | 120 | 159 | 63 |
| Mapzen | 774 | 3228 | 2130 | 232 | 58 |
| ArcGIS Editor | 0 | 45 | 107 | 100 | 27 |
| YAPIS | 0 | 0 | 0 | 229 | 24 |
| osmtools | 61 | 108 | 53 | 42 | 14 |
| OpenSeaMap-Editor | 18 | 114 | 100 | 124 | 12 |
| BigTinCan Upload Script | 164 | 186 | 101 | 86 | 11 |
| bulk_upload.py | 65 | 38 | 17 | 6 | 2 |
| wheelmap.org[2] | 0 | 262 | 1064 | 349 | 0 |
| Other | 308 | 541 | 399 | 839 | 204 |
See comments above to understand why the market shares in this plot do not add up to one. (It's not a bug, it's a feature.)
See also this plot for the absolute numbers - the normalized plot above may be somewhat misleading if interpreted improperly.
Number of users as a percentage of the total user-base
by number of edits
| editor | 2009 | 2010 | 2011 | 2012 | 2013 |
|---|---|---|---|---|---|
| JOSM | 125 615 526 (28.7 %) | 494 458 271 (78.6 %) | 539 079 358 (80.8 %) | 542 105 707 (74.2 %) | 192 926 988 (69.5 %) |
| Potlatch 2 | - | 7 584 901 (1.2 %) | 92 218 969 (13.8 %) | 146 250 937 (20.0 %) | 58 794 532 (21.2 %) |
| osmupload.py | - | - | - | - | 17 623 237 (6.4 %) |
| Roy | - | - | - | 1 216 065 (0.2 %) | 2 940 171 (1.1 %) |
| Merkaartor | 5 378 663 (1.2 %) | 10 887 187 (1.7 %) | 10 250 061 (1.5 %) | 9 702 870 (1.3 %) | 1 468 264 (0.5 %) |
| iD | - | - | - | - | 1 040 658 (0.4 %) |
| osmtools | 150 412 240 (34.4 %) | 8 994 328 (1.4 %) | 981 921 (0.1 %) | 1 229 689 (0.2 %) | 753 793 (0.3 %) |
| upload.py | 31 167 015 (7.1 %) | 10 166 002 (1.6 %) | 2 435 996 (0.4 %) | 421 238 (0.1 %) | 426 895 (0.2 %) |
| Potlatch 0.x/1.x | 44 030 413 (10.1 %) | 57 141 365 (9.1 %) | 8 756 458 (1.3 %) | 1 471 290 (0.2 %) | 218 695 (0.1 %) |
| osmapis | - | - | - | 4 169 283 (0.6 %) | 48 713 (0.0 %) |
| Redaction bot | - | - | - | 21 303 508 (2.9 %) | 7 783 (0.0 %) |
| PythonOsmApi | 366 913 (0.1 %) | 836 867 (0.1 %) | 9 502 989 (1.4 %) | 24 803 (0.0 %) | 1 328 (0.0 %) |
| bulk_upload.py | 64 669 041 (14.8 %) | 24 649 625 (3.9 %) | 1 442 184 (0.2 %) | 759 090 (0.1 %) | 367 (0.0 %) |
| FindvejBot | 262 (0.0 %) | 3 797 787 (0.6 %) | 358 760 (0.1 %) | 104 606 (0.0 %) | 175 (0.0 %) |
| bulk_upload_sax.py | 7 781 579 (1.8 %) | 722 332 (0.1 %) | - | 52 554 (0.0 %) | - |
| AND node cleaner | 3 314 543 (0.8 %) | - | - | - | - |
| Other | 4 359 343 (1.0 %) | 10 153 193 (1.6 %) | 2 155 541 (0.3 %) | 1 354 388 (0.2 %) | 1 165 563 (0.4 %) |
- ↑ 1.0 1.1 writes a "create_by" tag instead of created_by=* in changesets
- ↑ now identifies itself as rosemary (the library employed by the wheelmap editor) in the created_by tag
Editor profiles (2012)
In this section, a selection of editors are displayed at their position in a two-dimensional space spanned by various quantities which allow some characterisation of the editor. (The selection is more or less arbitrary. If your favourite editor is missing, please complain.) All figures are for edits in the year 2012 only.
The first plot displays each editor by the number of its users and the total edits made using it, which may be interpreted as the editor's total impact on the OSM database. The distribution of editors across the plot is roughly diagonal: There is a mainstream region in the upper right, where editors have a large user basis and are used to create or modify large amounts of data: JOSM, Potlatch 2 and - to a lesser extent - Merkaartor and Potlatch 1. All other editors are used by relatively few mappers (down to just one), and consequently, relatively few edits are made with them. Only a few specialized tools for large-scale edits appear somewhat off-diagonal.
The second plot again has the number of each editor's users on the horizontal axis, but now the average size of each changeset is plotted on the vertical axis. This tells us something about how each editor is used - for editing only few objects, or for larger changesets. The number of edits in a changeset using one of the mainstream editors is of the order of 100. Surprisingly, the average JOSM changeset is only roughly a factor of 2 larger than the average Potlatch 2 changeset. Various other editors (often mobile applications) usually make only a few edits in each changeset. As expected, specialized tools for large-scale edits (used by only few mappers) generate large changesets. Of course, these are all just average values, which tell nothing about the distribution of changeset sizes. We will address that issue below and try to explore the distribution of changeset sizes.
As mentioned, the average size of a changeset (i.e. the number of edits in it) provides only very little information about how an editor is used - some number appearing as an average value may result from very different distributions. To visualize the distribution of changeset sizes, quantiles are plotted for the mainstream editors in the plot to the right. Each box describes the changeset sizes which make up 50 %, 90 % etc. of all changesets created using a given editor (e.g. 50 % of all Potlatch 1 changesets contain 4 objects or less, 90 % contain no more than 58 objects). The 50 % quantile is the median; for both Potlatch versions, this is roughly identical to the average. For JOSM and Merkaartor, both measures differ significantly, hinting a larger tail of the changeset size distributions.
Note: I am aware that this plot badly needs some cosmetic improvements. Please be patient.
Update discipline of JOSM users
How regularly do JOSM users update their editor? This plot shows the JOSM versions in use in a given month, weighted by the number of changesets and clustered by the version number's leading digits. The majority of users clearly keeps their JOSM largely up-to-date, working with at least a recent "tested" release - commercial software manufacturers must envy the JOSM developers. The average version in use is roughly three months behind the current release; however, also very old versions are still being used by a small but nonvanishing and apparently die-hard number of mappers. Even the ancient release 1566 from May, 2009 is still seen occasionally.
There is also another plot which covers a longer term.
