Terrible examples. He says "Sparklines (...) are small infographics designed to convey large amounts of information "inline" with the text" and then shows a big graph. He then continues "From the general shape the graph, the reader is able to get a sense for the general trend (upwards)" which is absolutely not true. The graph is too wide to see the upwards trend easily and since there are no horizontal guide lines I say most people don't notice it.
There is some news site that embeds tiny sparklines of company's stock history actually inline with the text. That's good use of them.
I'm sure there's great uses for these things, but currently they're a solution in search of a problem. This is wishful thinking:
"The last data point, annotated with text and a dot tells the reader the value, allowing them to intuitively set the scale for the rest of the graph."
If the HTTP return codes is the killer application for sparklines, I couldn't say. I'd have to wonder, though, if this were a mission critical issue: Would you do the same thing?
Having only 1 point annotated means that you can't tell the scale of the graph. Are those little tiny changes over time, or extremely massive ones?
I can't imagine anything I would find this useful for, and if I tried to use it for something for management, they'd ask the scale as their first question.
There is some news site that embeds tiny sparklines of company's stock history actually inline with the text. That's good use of them.