A colleague suggested Watson might be a good, easy tool to use our IATI data. He says he saw a demo by the Watson team who downloaded our data and were able to do some basic stuff (graphs etc) within half-an-hour. They said some configuration work would be needed to do it properly (like filtering our data for language) but that it would be quite easy.
Has anyone tried it? I intent to look at it, if their website lets me create an account for the free trial (no luck after 1 hour watching the “processing your request” wheel).
So I’ve had a play with the free version of Watson, and it can pretty much do the kind of stuff that Google Fusion can do, plus some more advanced dashboard and infographic type layouts. If your data is nice and clean, it’s quick to use to create basic charts and visualisations (this one is Global Giving’s disbursements to organisations in Kenya over time, for example).
Watson chart example.png1349×626 65.6 KB
and this one is a treemap of the organisations by value of disbursements:
Watson heatmap example.png1349×626 81.2 KB
You can do more stuff with the full version I’m sure and I couldn’t see that it imports XML straight in, so you’d still have to convert to CSV or import to Cognos, SPSS or your own database and export into Watson.
I just saw this thread. Any updates on using Watson to analyze IATI data?
It would be useful to hear from the Watson team about how they’d ideally like to see IATI data prepped for input into Watson.
Brent Phillips I had a quick look at the documentation for the new(ish) version 2.1 of Watson Analytics, and it looks like it still only accepts flat files (ie .csv, .xls etc) and not files with hierarchical data such as XML. This means that you’d need to run the XML data through a conversion process in order to produce the flat file prior to uploading to Watson. Once converted, I also found it really useful to only include the fields (columns) that I needed for the data viz, as once flattened, IATI produces a lot of columns! Also helps to assess data quality issues before it’s uploaded to Watson so you can fix or at least be aware. Watson provides a data quality score once your dataset is uploaded, but I haven’t found it that useful.