Something in this technically well-crafted distribution diagram caught my attention, and at first I didn’t really know what it was…

This is from the Start Fund Report 2017. The Start Network is made up of 42 humanitarian aid agencies and “enables member NGOs and their partners to respond quickly to crises that would traditionally slip under the radar of humanitarian response”. Get more information on the Start Network here and on the Start Fund here.

This is a breakdown of “activated alerts” by geography, by crisis type and by activities that were triggered in the reporting period 01 April 2016 to 31 March 2017. Out of 74 alerts that year 43 resulted in activations (i.e. release of funds for immediate response to a humanitarian crisis).

So what made me stumble and raised my scepticism while looking at this diagram? Can you spot it?
[If you wish to find out yourself, stop here and do not read my thoughts below!]

I had intuitively tried to figure out the number of activated alerts in Africa. But how many alerts correspond to 50.7% out of 43? So I got out my pocket calculator and tried to get some integer numbers for the other values represented by the streams (or ‘bands’). I figured that the percentage values would probably be rounded to one decimal digit, but nevertheless could not get to any meaningful numbers for alerts per geography (nor per crisis type or per activities).

Next I thought that maybe this Sankey diagram doesn’t visualize absolute number of alerts, but rather the money that the fund distributed for these cases. But then, the sub-headline clearly states that it “shows activated alerts” and not “money distributed in response to activated alerts”.

The note “data is from alerts 77 to 150” under the diagram made me think: Hey, what if that was just a small typo and they are actually showing the distribution of 73 alerts instead of 43 alerts. Well, unfortunately none of the percentage values match with a base of 73 cases either and would yield absolute numbers.

My best guess, at present, is that we are actually looking at 150 alerts (the total number of cases the Start Fund has actually been activated over the last three years) and not at 43. Taking into account that the percentage values are rounded (50.66% > 50.1%) this seems to be the best match producing mostly integer numbers. But I might be wrong…

Maybe the author of the graphic or the report wishes to shed some light on this…

2 Responses to “Alerts distribution, Start Fund Network”

  1. Ian Simcox-Heath Says:

    Dear Phineas,

    Thank you for your inquisitive post and for your patience in my reply as I’ve only just come across this page. To clarify, this Sankey diagram is based on expenditure of activated alerts as opposed to number of activated alerts, which explains the issues you encountered when doing calculations. Our MEAL team decided on this approach because we felt it more accurately portrays the regions, crises and sectors in which the Start Fund operated that year, given that funding amounts (and therefore, expenditure) differed so greatly from one activated alert to another. Expenditure is mentioned in the notes below the diagram, but I agree with you that it should have also been included in the title to prevent any readers from misinterpreting what the diagram represents. We’ve just started our work on the next Start Fund Annual Report, so your comments are really appreciated and have already been taken on board as we think of new and better ways to show the work of our membership.

    All the best, Ian (on behalf of the Start Fund MEAL team).

  2. phineas Says:

    Hello Ian,
    thanks for your input and clarification. Hope to again see a Sankey diagram in the next Start Fund Annual Report

Leave a Reply