For some numbers on this I used the methods from this thread to make this table:
| Year | Observations | Identifications | Ratio |
|---|---|---|---|
| 2014 | 521033 | 469634 | 0.9014 |
| 2015 | 854229 | 1006366 | 1.1781 |
| 2016 | 1686482 | 2260729 | 1.3405 |
| 2017 | 3269227 | 5374616 | 1.6440 |
| 2018 | 6986623 | 10801043 | 1.5460 |
| 2019 | 13348605 | 20699723 | 1.5507 |
| 2020 | 23638564 | 30716193 | 1.2994 |
| 2021 | 29261242 | 39813280 | 1.3606 |
| 2022 | 33696966 | 48831850 | 1.4491 |
| 2023 | 41352759 | 58999228 | 1.4267 |
| 2024 | 50929555 | 71091183 | 1.3959 |
| 2025 | 62959261 | 82234281 | 1.3062 |
If my understanding of @matthewvosper’s numbers is correct then we’d need the ratio of IDs-for-others to be above around 1.5 to keep up, which was only the case from 2017-2019. On top of that when the raw number of observations is higher, then the ratio being a bit off means a difference of millions of observations rather than thousands as it was back in 2015. There’s a good chance it takes years for a new identifier to go from being someone who makes thousands of identifications in a year to someone who makes tens or hundreds of thousands of identifications a year, and by then the growth of observations will be even higher (although I imagine iNat’s growth should start to plateau at some point…). Then there’s the additional issue that having more identifiers to help confirm American Alligator and Blue Jay observations that already have 3-4 identifications doesn’t contribute to the frustrations being expressed in this thread.
But @matthewvosper’s 2.5 IDs-per-obs value is based on a certain rate of error in the initial IDs, which would often come from CV errors. If the CV is improved then that ratio can be brought down closer to 2. Then maybe the 2025 rate of 1.3 IDs-for-others would be sufficient (not accounting for backlog).