Lens Repair Data 4.0

Published September 20, 2010

What Is This?

We started doing this several years ago. We have a unique opportunity: we own a very large number of lenses (5,600 have passed through our system) subjected to rather harsh conditions: they get packed in boxes, tossed around by UPS, and sometimes the user isn’t as careful with them as you would be with your own lenses.

Our numbers reflect how often lenses fail under fairly harsh conditions: your personal lenses shouldn’t fail nearly as often; they aren’t subjected to rental conditions. But since all of our lenses are subject to roughly the same number of rentals per year, and the same rental conditions, it does provide some comparison about how fragile various lenses are compared to other lenses. It’ not scientific, but it’s a bit more useful than posts on a forum going back and forth between “never had a problem with it” and “mine sucked.” For those of you interested, I’d also suggest looking at the LensPlay Database. Again, it’s not scientific, but it has responses from thousands of users, so it is a large series.

This list is not a comment about how good a lens is, it’s about how often it breaks under harsh conditions. Some of my favorite lenses are on this list.

What Is Different This Time?

First, this will be our last Lens Repair Data report, because our new website will soon start putting the repair data for each lens on the item page in real time. I may write an editorial or two in the future about repair rates, but the up-to-date repair data for every lens will be available on its web page, so I don’t need to repeat it every six months.

Second, we’ve doubled the number of lenses we carry since the last report, so this report includes Sony, Zeiss, and other lenses that were not eligible for review previously because we didn’t have enough copies.

Finally, and this is most important: we’re doing it differently this time. Previously our numbers have been additive for an entire year–we looked at the data for one year prior to the date of the report. This report is for the 6 months period from January 1, 2010 through July 1, 2010. There are two reasons for the change:

  1. We are so much larger now that we’re actually looking at as much data during this 6 month period as the previous report looked at for a one year period, so it makes a nice then versus now comparison.
  2. A few months ago I wrote an article, This Lens is Soft and Other Facts in which I speculated that manufacturers quietly upgrade problem lenses fixing problems without really admitting the problem was ever there. Part of the reason I made that speculation (and that’s all it is, I have no proof at all) was the changes in repair rates we’ve seen with certain lenses. Showing the data this way demonstrates a couple of those changes.

What do the numbers mean?

It is an annualized repair rate of the period from January 1, 2010 through June 30th, 2010 for lenses that have failed (meaning stopped working with no obvious signs of damage). The percentage we use is simple—number of lens repairs divided by the average number of copies of that lens stocked. With this report, we then doubled the six month figures to give the annualized rate so that it is comparable to previous reports.

A dropped or physically damaged lens does not count as a repair for this list. It is probable that some lenses included in these numbers actually were damaged, but there was no overt evidence that it was so. We do not accept the factory service center’s word that the failure was “secondary to shock damage,” because a number of brand new, fresh-out-of-the-box lenses that we’ve sent back for repair came back with warranty work denied because of “shock damage,” even though we sent the lens to them straight out of the manufacturer’s shipping box.

We make no comment at all on lenses that we have less than 9 copies of (we don’t think that’s particularly useful, it’s too small of a number), or on new lenses we’ve carried less than 6 months (unless something is spectacularly bad, which has only happened a few times). All of our lenses are rented with about the same frequency (we have many more copies of popular lenses than of less popular lenses), so there’s not a great difference in the amount of wear-and-tear one lens gets compared to another.

Now, for those of you who want to reach conclusions from the data, please note the following:

  • Our median failure rate for all lenses during the last 6 months was 5.5%.
  • If the lens is on the list below, we’ve had more than 9 copies during the last 6 months, and it has a repair rate significantly above the median rate.
  • Our lenses are less than two years old, the vast majority less than 15 months old, we rotate them out of stock regularly.

There are three reasons a lens is not on the list:

  • We don’t carry it. Hence, we have no comments on the Sigma 70-200 f/2.8 OS, Canon 18-55 EF-S, and a number of other lenses, because we don’t carry them.
  • We carry the lens but have fewer than 9 copies of it, and therefore we don’t feel any comment on reliability is appropriate (all Leica, 4/3, and Micro 4/3s lenses fit this category).
  • The lens has proven so unreliable that we’ve stopped carrying it. The following three lenses have been discontinued for this reason:
  • Sigma 150-500 OS
  • Sigma 120-400 OS
  • Tamron 70-200 f/2.8

One other repair rate favorite no longer qualifies for the list because we now have less than 9 copies, the Sigma 120-300 f/2.8. Rental demand for it has dropped significantly and we just don’t carry as many copies as we used to.

If we stock the lens, it isn’t in the ‘low copy’ category mentioned above, and it’s not on this list, then its failure rate is not high in our experience.


Lens Annualized Repair Rate Typical Problems
Sony 24-70 f/2.8 ZA 30% Decentered, calibration, autofocus
Sony 70-200 f/2.8 APO 28.5% Autofocus, manual focus
Canon 18-200 IS 23% Zoom sticking, IS failure, aperture stuck
Nikon 70-300 AF-S VR 22% VR failure, autofocus failure
Nikon 50 f/1.4 AF-S G 22% AF, calibration, decentering
Nikon 70-200 f/2.8 VR II 21% Zoom sticks, VR failure, AF failure
Nikon 24mm f/3.5 PC-E 20% Loose mechanism
Nikon 14-24 f/2.8 17% Zoom sticks
Sigma 50-500 OS 18% AF failure
Tokina 12-24 f/4 PRO 14% Zoom sticks, calibration
Canon 35mm f/1.4 14% Calibration, decentered element, autofocus failure
Canon 70-200 f/2.8 IS II 13.75% Zoom sticks or catches, IS failure
Canon 100-400 IS L 13.75% IS problems, zoom sticking, calibration
Canon 28-300 IS L 12.5% Zoom sticks, autofocus
Canon 70-200 f/4 IS 12% Manual focus, autofocus, IS failure
Nikon 16-85 f/3.5-5.6 VR 12% Zoom sticks
Tokina 11-16 f/2.8 11% Calibration, zoom jammed
Canon 50 f/1.2 L 11% Calibration, autofocus
Nikon 80-400 VR 11% Zoom sticks, VR failure, autofocus sticks
Nikon 17-55 f/2.8 11% Autofocus, calibration

A Few Observations

  • Fanboys love to misuse the list above, and one of the common things I’ve seen is,“Brand X has the most (or least) lenses on Lensrentals’ high repair rate list.” Let’s keep it in perspective. There were 41 Canon, 37 Nikon, 14 Sigma, 13 Zeiss, 7 Sony, 5 Tamron, and 3 Tokina lenses (120 total) eligible to make the list. The final makeup was 7 Canon, 6 Nikon, 2 Sony, 2 Tokina and 1 Sigma lenses. Every brand seems to have some troubled and some trouble free lenses.
  • This is the first time period during which we’ve had enough data to show Sony lenses in this report. They definitely represented once given the opportunity, but the sample size is relatively small (for example we have 12 to 15 copies of most Sony lenses compared to over 100 copies of similar Canon and 50 copies of similar Nikon lenses), so a few repairs made a relatively high number. Then again, when 1/3 of all the copies fail, it makes you take notice.
  • I would also add, for those who haven’t experienced an out-of-warranty repair on a Sony lens, the price will make you spit out your morning coffee. If you want to know how much higher, let’s just say it’s in proportion to the cost of a Sony lens hood compared to a Canon Lens hood. Alpha has some of the best lenses around, but say your nightly prayers that they don’t break.
  • For those that believe early adopters face risks, this year’s data tends to support you. Three new lenses: the Canon 70-200 f/2.8 ISMkII, the Nikon 70-200 f/2.8 VRMkII, and the Sigma 50-500 OS made the list. The models they replaced were all low-repair rate lenses.
  • Several lenses that have been regulars on the list previously not only dropped off, they dropped way off. The Canon 17-55 f/2.8 IS has been a fixture on this list for IS and electrical problems since we started but not only is it gone, its repair rate has dropped to 6%, right around our average. Similarly, the Canon 10-22 EF-S has dropped off the list and now has one of our lowest repair rates. The Nikon 24-70, which gave us a horrible time with sticking zoom barrels when it was first introduced, now is virtually trouble free. We’ve eliminated the web-page warnings on all of those lenses. (I probably should point out again that we turn our lenses over pretty frequently, and in all three cases over half of our current copies were purchased in the last 6 months.)
  • The Nikon 70-300 VR, which has always been a trouble free lens, suddenly appears on the list almost entirely from electrical problems. It reminds me of a similar phenomenon we saw last year with Canon 300 f/4 IS lenses, where a number of copies bought at roughly the same time had a very high problem rate. A batch of bad circuit boards, perhaps? If so, the problems will disappear with our next batch of lenses.

One Last Thought

I wrote in another article that it seems likely that manufacturers quietly correct problems with subassemblies in various lenses. We’ve seen some circumstantial evidence of that with this year’s data. Several lenses that have had high repair rates for the last couple of years are suddenly absent. Not just absent, but they’ve gone from high repair rate lenses to very low repair rate lenses.

Lens 4.0 (2010) 3.5 (2009) Repair 3.0 (2008-9) 2.0 (2008)
Canon 17-55 f/2.8 IS 7.7% 29% 25% 20%
Canon 10-22 EFS 3% 9% 17.5% 16%
Nikon 24-70 f/2.8 5% 14% NA NA
Canon 35 f/1.4 14% 10% 22% 9%
Nikon 17-55 f/2.8 11% 20% 10% 19%
Canon 100-400 14% 16% 11% 12%

The three lenses listed at the top, the Canon 17-55 and 10-22 EF-S lenses and the Nikon 24-70 have had dramatic changes in repair rates this year compared to last. However, I also want to make clear that there is always some variation in our numbers from one report to another. I’ve put the two lenses that have had the widest variation in repair rate over the last several reports (the Canon 35 f/1.4 and Nikon 17-55 f/2.8) next. As they demonstrate, a change from 20% to 10% and back again can just be normal variation of our numbers. The last example, the Canon 100-400, is there just to show what our more typical lens looks like. There’s some variation, but nothing all that dramatic. The vast majority of our lenses are like this, roughly the same repair rate every period.

So what does it mean? It might be that something is slightly different in newer copies of the Canon 17-55, 10-22 and the Nikon 24-70 that has solved a reliability issue. Or it might mean nothing at all, it could just be mathematical chance.

Maybe we should take an old one and new one apart and see if there’s anything different. . . . But in the meantime, we’re going to move to using “last 6 months repair data”, which will be posted on the various lens pages, rather than overall repair data as our standard.

Author: Roger Cicala

I’m Roger and I am the founder of Hailed as one of the optic nerds here, I enjoy shooting collimated light through 30X microscope objectives in my spare time. When I do take real pictures I like using something different: a Medium format, or Pentax K1, or a Sony RX1R.

Posted in Equipment
  • Yes – this IS useful 🙂 – read on.
    Margin of error in a normally distributed population is
    (1) ~~= 0.5 x square_root( (P)x(1-P)/N) for N items or
    (2) ~= square_root(1/N) worst case (p = 0.5)
    Where p is the actual probability. (0-1 = 0 100%)
    Putting P = 0.5 gives the worst case easy formula (2) above
    SO For 100 lenses the margin of error (MOE) = sqrt(1/100) = 0.1 = +/- 10%
    For 10 lenses it’s sqrt(1/10) ~= 0.3 = +/- 30%
    This obviously need sto be used with intelligent case BUT gives some feel for what failure rates suggest.
    The other useful test is the “Auric Goldfinger test”
    Once = happenstance.
    Twice = coincidence.
    Three times = enemy action (Mr Bond).

    Russell McMahon

  • Jay

    Thank you so much for this information. I recently sent a lens in for repair, only to be shocked by the “shock damage” evaluation. I was pointed to this page on a popular photo gear forum, and feel vindicated for pursuing warranty repair, despite the manufacturer’s evaluation (they relented, btw, and my lens is being repaired under warranty, despite their claim that it shows shock damage). I’ll leave it at that, but will say that my lens is on your discontinued list 🙁

  • Sigurd

    I just want to say thank you very much for publishing this interesting article. Ultimately every manufacturers produce some copies that fail even most of the batch is good, but this is interesting information to see what are my odds in getting a bad copy.

  • Geoffrey

    Is “average number of copies of lens stocked” relevant? Wouldn’t the data be more representative if failures were tied to “number of days rented”? If you have 9 copies of Lens Model A sitting on a shelf you have no data, and the one copy of Lens Model B that gets rented 39 times will provide lots of useful data, and the unpopular Model A will have a perfect repair record compared to that of the very popular Model B.

    I realize that you naturally tie your stock to demand, but I would think a more accurate Failure Rate = (Number of Repairs) / (Aggregate Number of Days Lens Model is Rented), and the threshold for including the data would be some Minimum Number of Days Rented.

    There are also, of course, huge variables (conditions of use, clumsiness of user, etc.) that make the data less-than-scientific, but this is an interesting project. Thanks!

  • Bernard Ortiz

    As the lens repair data isn’t quite up yet – and it’s now been quite a while since the last report – could you re-run the report? This one could be interesting, because of the longer ‘hold’ times you have due to the shortages…

  • Dmitry

    Could you publish the full version of rating?

  • Roger Cicala

    It’s still in the works, I’m afraid. We had planned to have it done this winter, but it turned out to be more complex to automate than we had expected.

  • Brian

    I see you mention putting the lens repair data on the product pages but I don’t see this even though you’re on the new site. Is this still in the works or is it only on high risk products?

Follow on Feedly