Thanks to you both for your quick answers, much appreciated!
Indumon, indeed I see how easy that can be but it doesn't quite give me what I was after. It is giving me the percentage of the blue line compared to the overal value of each week. But each week actually the overall total value changes, and I wanted to stick to dividing by the total in the first week. The biggest example is on day 4 where for the blue line I want the calculation of 8/(130+127) = 3%, whereas I think that approach is doing 8/(8+96) = 8%
But thanks, it's a useful technique for other purposes!
Dan, that solution does indeed work for me, thanks. And I don't think I would have come up with that in a hurry either, very clever :-)
[I had a little trouble implementing it as my data source is a cube so I can't change aggregations in Tableau ("sum is not a recognised function" etc. apparently ) but it works 100% on non-cube data, and I got the logic you used for "First day total" implemented in the cube itself and now it works great even on the cube.]