In most cases, the cost-of-carry is harder to observe than the spot price of the underlying asset, which leads market participants to ignore the cost-of-carry, and compare the profit and loss of futures contracts to the spot price of the underlying assets directly as the roll yield.
[1] In most practical cases, it is difficult to confirm that the roll yield is equal to the cost-of-carry, as the convenience yield portion of the cost-of-carry is only observable in perfectly efficient futures and spot markets, which rarely exists in the real world.
However, in the less-than-perfectly efficient markets in the real world, that difference might contain price biases that do not reflect the convenience yield.
Roll yield is often characterized as an extra gain or loss that a futures investor captures in addition to the change in the spot price of the underlying asset.
However, market participants may face obstacles to conduct the above arbitrage; they might have difficulty obtaining physical oil storage, they might lack to scale to purchase oil at the spot market price, and they might face other obstacles.
The contango exhibited in Crude Oil in 2009 explains the discrepancy between the headline spot price increase (bottoming at $35 and topping $80 in the year) and the various tradeable instruments for Crude Oil (such as rolled contracts or longer-dated futures contracts) showing a much lower price increase, because of the strong negative roll yield.
[3] The USO ETF (using futures contracts) also failed to replicate Crude Oil's spot price performance.
Levine, Ooi, Richardson, and Sasseville (2018) found that the roll yield, after adjusting for interest rates, made up the majority of average returns for a long-run index of commodity futures going back 140 years.