I am the mail admin at a manufacturing corporation with about 2000 email
users and 3500 messages a weekday.
Here is a two week sample of how razor v2 is doing at our corporation
compared to other spam blocking mechanisms. "Jammed" is a word recipe
blocker and "SA" stands for SpamAssassin, threshold 5, no network checks.
"Blacklisted" checks the envelope sender and the "From:" address
and supports regular expressions. My Bayesian filter corpus is
getting dated. With an update, I think I could boost it 10% on
average.
I review messages not caught by at least two filters. Blacklisting
counts as two filters. About 50 to 100 messages a day are caught by just
one filter. I have about seven false positives a day from that 50 - 100
that are worth forwarding on to the recipient. There are probably
ten a day more false positives that aren't worth the e-paper they are
printed on, so I pitch em and/or whitelist the sender so I don't get bothered
further.
Razor is still a valuable filter at 40% (all parts must be spam). I
haven't had time to make these live and publically available as has been
requested often, but I wanted to share.
|
| |||||||
| Date | Spam Count | Jammed | Razored | Blacklisted | Trollbox | Bayesian | SA |
| 2003-03-03 | 936 | 166 (17.74%) | 327 (34.94%) | 321 (34.29%) | 259 (27.67%) | 756 (80.77%) | 706 (75.43%) |
| 2003-03-04 | 1390 | 232 (16.69%) | 533 (38.35%) | 517 (37.19%) | 373 (26.83%) | 1062 (76.40%) | 1033 (74.32%) |
| 2003-03-05 | 1204 | 204 (16.94%) | 488 (40.53%) | 367 (30.48%) | 369 (30.65%) | 979 (81.31%) | 913 (75.83%) |
| 2003-03-06 | 1111 | 174 (15.66%) | 460 (41.40%) | 370 (33.30%) | 341 (30.69%) | 864 (77.77%) | 829 (74.62%) |
| 2003-03-07 | 991 | 143 (14.43%) | 432 (43.59%) | 426 (42.99%) | 275 (27.75%) | 799 (80.63%) | 709 (71.54%) |
| 2003-03-08 | 762 | 130 (17.06%) | 360 (47.24%) | 248 (32.55%) | 233 (30.58%) | 678 (88.98%) | 568 (74.54%) |
| 2003-03-09 | 686 | 130 (18.95%) | 320 (46.65%) | 210 (30.61%) | 244 (35.57%) | 583 (84.99%) | 523 (76.24%) |
| 2003-03-10 | 1006 | 145 (14.41%) | 380 (37.77%) | 316 (31.41%) | 262 (26.04%) | 756 (75.15%) | 737 (73.26%) |
| 2003-03-11 | 1042 | 174 (16.70%) | 395 (37.91%) | 400 (38.39%) | 273 (26.20%) | 806 (77.35%) | 767 (73.61%) |
| 2003-03-12 | 1246 | 169 (13.56%) | 465 (37.32%) | 428 (34.35%) | 388 (31.14%) | 998 (80.10%) | 892 (71.59%) |
| 2003-03-13 | 1183 | 213 (18.01%) | 453 (38.29%) | 380 (32.12%) | 373 (31.53%) | 952 (80.47%) | 849 (71.77%) |
| 2003-03-14 | 973 | 176 (18.09%) | 368 (37.82%) | 368 (37.82%) | 305 (31.35%) | 824 (84.69%) | 688 (70.71%) |
| 2003-03-15 | 810 | 162 (20.00%) | 350 (43.21%) | 332 (40.99%) | 269 (33.21%) | 721 (89.01%) | 547 (67.53%) |
| 2003-03-16 | 691 | 115 (16.64%) | 302 (43.70%) | 243 (35.17%) | 257 (37.19%) | 605 (87.55%) | 487 (70.48%) |
| 2003-03-17 | 1004 | 228 (22.71%) | 327 (32.57%) | 385 (38.35%) | 305 (30.38%) | 796 (79.28%) | 676 (67.33%) |
