We all know that h-index is crap.
There are plenty of reasons for this. There is of course the idea that it is an indicator and that research can not be assessed only by indicators. This is true, but since we are evaluated by indicators, they might as well be useful.
1/n

In this thread, I will try to give you a real-life argument against the use of h-index.
2/n

Let me introduce you 2 anonymous researchers, Dr Blue and Dr Red. They are the same generation and have a career of about 15 years of research.

You can see on this plot the number of papers per year each published.

3/n

Dr Blue has a steady career.
Dr Red has had a big change in his career in 2016 that made them change gear.
And it really shows!!

4/n

If we look at the evolution in the h-indices of Dr Blue and Dr Red as a function of time, this should completely obvious right. These 2 researchers clearly don't play in the same category anymore.
Let's have a look at their h-indices:

5/n

Nope, they are not very far, and they follow a very steady and close growth...

How is it possible ?

6/n

That's because h-index:
1) is an integral value
2) assumes the career was completely homogeneous in terms of "impact"

7/n

Bc of 1) it bundles together both parts of Dr Red's career and anyway, bc of 2), it has no way to cope with such a different publishing pattern.

8/n

Here is an alternative 'indicator'. It is the average num of cites per year due to papers published that year. Let's call it "annual impact". Here is how it looks for Dr Red and Dr Blue.

9/n

You see that it does capture the change, and it does show that in 2020 they are not in the same category anymore.

10/n

This is an instantaneous (per year) indicator. You could make it integral if you wished, and plot the cumulative impact.

11/n

Which alongside h-indices shows that h are really useless.

12/n

I am not trying to _improve_ h-index, my message is really to show that it really IS crap.

13/n

And no, I don't think we can summarize a career by a number, whatever that number is.

14/n

But as a scientist, I can not simply say that metrics are useless. Metrics (good metrics) are useful. Summarizing someone by 1 metric is crazy.

15/n

It's the image of the body temperature: it is not enough to say if you are well or sick, but it is still useful AMONG THE REST of the examination.

16/n

But if your thermometer is crap, you need to throw it away!

The end.

17/n

PS: after reading some of the comments, I also add their number of cites per year and total number of cites. These metrics are a bit less straightforward to interpret, but even these show a slope change around 2016...