To test shredders we tried to recreate all of the common shredding scenarios the average person would encounter: a stack of papers at tax time, envelopes of junk mail, credit cards, CDs, stacks of paper stapled together, and more. We also got into the mindset of a thief to determine which models yielded some opportunity for recovering information, and which negated any risk of identity theft. In the end we organized our testing into four metrics: shredding quality, speed, ease of use, and noise.
The most salient aspect of our shredding quality testing was security. This directly corresponded with the size of the shreds produced. The smaller the shred, the higher the security. This resulted in micro-cut models scoring better than cross-cut models. For more on the relative security of these models, and how to make cross-cuts more secure, see our buying advice article. We also tested whether or not each model could handle its claimed maximum sheet capacity. Models that could received a perfect score on this test, and models that couldn't got various lower scores depending on how far off they were. We constructed the test this way, rather than trying to push shredders beyond their advertised capacities, because we found pushing limits could compromise security well before a shredder would actually jam. Most models will accept more than their advertised capacity, but may not shred these stacks completely, leaving valuable information intact and legible. So we felt it was more important to be able to adequately shred at the advertised capacity than being able to exceed it. We also tested robustness by shredding tougher things like credit cards, CDs (if the model was rated for CDs), and fully stuffed junk mail envelopes. Finally, we ran each model continuously for 15 minutes, putting in a few sheets at a time, to see if any would overheat or jam. All the models passed this final test.
To evaluate speed we timed how long each model took to shred a stack of paper equivalent to its maximum capacity. For models that could not meet their advertised maximum capacity in our shredding test, we used the maximum number of sheets that we found it was able to effectively shred at once. We then took this time and calculated a maximum pages per minute figure. We also time how long each model took to shred 20 sheets feeding them in a few sheets at a time, and how long each took to shred a single sheet. However, we felt the maximum sheets per minute figure was much more useful. It better reflects how an average user would complete a large shredding job: feeding close to the maximum sheet count, instinctively adjusting the size of the stack based on whether the shredder sounds like its struggling or not. We then awarded a score to each model based on their relative speeds.
Ease of Use
Shredders are not complicated to use. Most have a single switch that toggles between on, off, and reverse (to help alleviate jams). So, much of our ease of use scoring was based on how easy our testers felt it was to empty the bin of each model. These opinions were based off of emptying each model many, many times throughout the course of our testing. Models did get extra points if they had nice features, like a light to indicate when the bin is full, or a sensor that turns the blades off if your finger gets too close. We eventually removed jam clearing from our scoring as every model had a reverse setting that worked well for spitting out material that had jammed the blades.
Originally our noise testing involved holding a noise meter a foot above each model as it shredded paper and recording the decibel level. After analyzing the data we found that all the models produced fairly similar volumes, and that the ones that were technically louder didn't match up to the ones that we felt were the most annoying to listen to. Therefore we scrapped our original decibel data and had multiple testers rate the noise of each model subjectively, based on which they felt were more grating and which were more bearable. We then averaged these scores to arrive at our final noise score for each model.