GrantNav’s (new!) approach to currencies


#1

In the March 2018 GrantNav update, support for multiple currencies was greatly improved. Grants are grouped by currency; a funder’s most common currency is used by default etc. I confess I didn’t know GrantNav included grants in anything other than GBP, so this is really cool!

An alternative approach to handling multiple currencies would be to convert between them – i.e. allow the user to specify a currency, and amalgamate grants according to that. It’s noticeable that GrantNav avoids doing this. The current solution looks like it might have been more trouble to implement (though I might be wrong about that!) so presumably this is done very deliberately, in order to avoid processing funders’ data.

The current solution results in some weirdness:

  • the funders page includes 69 funders (rather than the 71 on the homepage). I guess that’s because it’s just showing GBP funders? (I haven’t checked)

  • the currency changes between funder pages (depending on the funder’s most common currency) which could be jarring for the user. E.g. Arcadia vs Rowntree

  • totals are all per currency, rather than an overall value total (Zing has non-trivial USD grants, but they total is hidden without an extra click)

  • the grants table on a funder page includes amounts in different currencies. Sorting by amount sorts by absolute numbers, rather than relative value. If there’s a plan to add more currencies, I suspect this will become increasingly confusing. E.g.

    47

This isn’t intended as criticism! I’m just trying to understand the reason for avoiding currency conversion.


#2

This is a great topic to pick up as will affect how people “trust” the data if discrepancies are not clearly explained. It takes a little digging to work out why there are disparities something I imagine fewer of our target users are likely to do or be able to do. May require a combination of clearer signposting, options to amalgamate (but based on what exchange rate?) or to display all currencies in certain views. Does this affect enough users to prioritise? In terms of trusting the data even if they aren’t interested in currency per se it seems quite important as could appear misleading or inaccurate inadvertently.


#3

Thanks for the feedback, Andy!

presumably this is done very deliberately, in order to avoid processing funders’ data.

You presume correctly. GrantNav is designed to be a mirror on the data - to present the data that is published in a helpful way and to encourage engagement with it. To achieve that, we’ve held to the principle of presenting the data as it was published as closely as possible. Up until now, GrantNav has actually been misrepresenting some of the data by presenting non-GBP grants as being in GBP.

The current solution looks like it might have been more trouble to implement

Currency conversion is a tricky thing to decide how to do, rather than tricky to do - there’s lots of decisions around rates to use, when to consider a conversion to have happened, and whether it exacerbates the sense of false equivalency between grants that were apparently the same size.

Fundamentally, however, we come back to the question of what GrantNav is for - and that’s to present the published data. If the publisher says that a grant of 200 USD was given, then GrantNav should say that a grant of 200 USD was given, not that a grant was given that’s equivalent to some amount of GBP. If that sort of analysis is helpful to a user, they (should) have everything they need in the download to do that analysis in a way that makes sense for their use case.

The current solution does have some weirdness that we need to work on, for sure. As @cased says, discrepancies make people distrust the data (or the tool) very quickly. We’re currently working on signposting / explaining the interface, so we’re hoping to address some of that over the coming weeks.

the funders page includes 69 funders (rather than the 71 on the homepage). I guess that’s because it’s just showing GBP funders? (I haven’t checked)

I’ve added https://github.com/OpenDataServices/grantnav/issues/456 to address this. I think you’re right, it’s because it’s only showing GBP, but we should either explain it or provide a switcher.

the currency changes between funder pages (depending on the funder’s most common currency) which could be jarring for the user. E.g. Arcadia vs Rowntree

Could you tell me more about this? It’s certainly more information on the page. I did wonder about this when we developed it, but I largely put it down to being so used to how it used to be that it was jarring for me just because it was different.

totals are all per currency, rather than an overall value total (Zing has non-trivial USD grants, but they total is hidden without an extra click)

This is a consequence of not converting. We did considering adding a totals section for each currency but had to trade off consistency of design with how much information we presented on the page. I’ll add it as a GitHub issue, if you think it’s something we should address differently.

the grants table on a funder page includes amounts in different currencies. Sorting by amount sorts by absolute numbers, rather than relative value.

This is another consequence of not converting. Our aim in the work we’ve just completed was to make GrantNav more truthful about non-GBP grants, and to that aim, it’s succeeded. The previous version just presented USD grants as being in GBP - so a 100 USD grant would be presented as a 100 GBP grant, which was just plain wrong.

However, this has, as you’ve discovered, created a whole new load of issues to fix. Ironic, perhaps, that stopping GrantNav providing incorrect information might actually have a negative impact on trust - but it’s all learning we need to do.


#4

I think I’d give the same answer as for @timdavies question about machine learning – as long as you’re transparent about how the processing was done, I’d consider it okay to do some reasonable processing that’s potentially helpful (i.e. state your assumptions / show your working). So for instance, using the value date and some trusted list of exchange rates for given dates seems reasonable to me, provided it’s stated that that’s what has been done.

As part of this update, GrantNav already processes and enhances the published data, by attempting to infer missing recipient regions, districts and wards from postcodes. This serves as a pretty good analogy, because it isn’t possible to do it with 100% accuracy. Boundaries move, new postcodes are created, postcode areas don’t map neatly to wards. But although this means it’s no longer a faithful reproduction of published data (and therefore strays from what GrantNav is for) I do think it’s a good and useful feature!