Results in IATI? Or observations? Or sources?

The IATI Technical Advisory Group is kicking off, and one discussion I’m looking forward to is on how to publish “results”.

There are proposals by Herb Caudil of DevResults (1, 2) and earlier reflections by Bill Anderson (1). Herb’s proposals are good: we need indicator references (or even URIs), and it makes sense to have a separate standard to describe the definition and logic of indicators.

After a few days of discussing standards, and with some concerns of NGOs in the back of my head, some scattered thoughts:

Results: outputs, outcomes, …?

One discussion is on the types of results: you can share outputs, but are outcomes or impacts linked to individual activities?

And maybe we also want to look at process indicators, performance indicators?

Why not turn it around?

NGOs (hopefully) collect data to analyse the results of their work. The actual results may be mostly qualitative rather than quantitative (especially when working on lobby, advocacy, political change), but their still often is underlying data: surveys, statistics, assessments, and so on.

There should be a way to share the research and sources you use in an activity. Whether your own, or external sources. And whether you use that in one or in multiple activities.

And there should be a way to share the “logic”: what is your theory of change, your log frame, your indicator framework?

Inspired by HXL

The HXL standard shines in simplicity: a set of tags and additional attributes you can use in a spreadsheet, to alleviate (but not eliminate) the efforts needed to aggregate or combine data from multiple sources.

It’s meant to be as close as possible to what people in an emergency situation actually use: spreadsheets, often in offline devices. Adding one header row of standardised tags can help a lot in sorting and reorganising the data.

For many NGOs, such a simple method would make it easier to share more of their research, and promote working towards more common terminology along the way.

As a rough sketch:

<actitivity>
  ...
  <result>
    <indicator ref="..." type="8">
      <target>
        <narrative>...</narrative>
      </target>
      <actual>
        <narrative>...</narrative>
        <source  href="https://data.worldbank.org/..." type="1"/>
        <source  href="https://ourwebsite/ourspreadsheet.csv" schema="...HXL..." type="2"/>
        <source  href="https://ourwebsite/anotherspreadsheet.csv" schema="https://ourwebsite/ourtags" type="2"/>
      </actual>
    </indicator>
  </result>
</actitivity>

Concluding

I hope we can move to a separate standard on indicators, results, observations, and programme logic, allowing incorporation of existing standards to live side-by-side (using XML namespaces?).

It may make it possible to combine more elaborate and semantically complete standards on the one hand, with an easier way to publish existing spreadsheets in a light-weight manner.

As a parallel: the XBRL (eXtensible Business Reporting Language) standard can incorporate the IFRS (International Financial Reporting Standards) but does not require it. XBRL is more of a container format, IFRS more of a content format.

IATI has come quite some way as a container format. It is tempting to add more content-specific or use-case-specific semantics to the standard, but wouldn’t it be great if IATI focuses on the exchange of activity information, and makes it easy to embed other standards for more content-specific or context-specific applications?

Posted in Aid Transparency, Conferences, Ideas, International development, Open data and tagged , , .

3 Comments

  1. I’ve just put the notes from our discussion of light-weight standards on Wednesday here: https://pad.riseup.net/p/IODC2015-preday-lightweight

    One of the bits in there I got a lot from was understanding a rough hierarchy of:

    (1) Vocabulary
    * (2) Fields
    * (3) Structure / Model
    * (4) Serialization / API

    And that we need to think about working through these different steps, without jumping straight to (4). This matches I think with what you are discussing above: if we start from sharing vocabs, and then fields, and then only get to particular serializations.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.