This reference defines the DPO as
DPO = N_defects / (Number of units * Number of opportunities per unit).
I'll base my answer on this definition.
The number of opportunities by itself already contains information about the number of units, since
N_opportunities = Number of units * Number of opportunities per unit.
In your example, one paperclip (one unit) has a certain probability of containing a defect. There may be, say, 3 kinds of defects (bent, broken, brittle), so
N_opportunities = 1 unit * 3 defects per unit = 3 opportunities
If you found your one paperclip was bent then the DPO would be
DPO = 1 defect / 3 opportunities = 0.33 (2 decimal places)
Of course, this metric is only useful for considering samples which are sufficiently large that the probabilities of the defects manifest in a meaningful way e.g. so that expectation values start to become useful quantities.
For example, consider 2 samples (A and B), each of 10000 paperclips. If sample A had 20 bent paperclips, 3 broken paperclips, and 7 brittle paperclips from a sample of 10000 then
DPO_A = 30 defects / (10000 units * 3 opportunities per unit) = 0.001
whereas, if sample B had 10 bent paperclips, 2 broken paperclips, and 3 brittle paperclips
DPO_B = 15 defects / (10000 units * 3 opportunities per unit) = 0.0005
The DPO now tells you something interesting about the samples and, perhaps, the way in which the samples were created.