Friday 22nd May, 2015
The project concerns the bits of the GOV.UK website which ask users a series of questions in order to show them an answer specific to their situation. Within GDS these are known generically as Smart Answers.
Rather than try to explain everything we’ve been doing on the project over the last few weeks, I thought I’d try to describe the situation when we arrived at GDS and a bit about what we’ve been asked to do. Hopefully we can then gradually bring you up-to-date over the next few weeks.
The Smart Answer “format” covers a wide range of domains. To give you a bit of a flavour, here are some examples:
- Calculate your employee’s statutory sick pay
- Getting married abroad
- Student finance calculator
- Calculate your leave and pay when you have a child
Some of the Smart Answers are purely decision trees, some are calculators, and others are hybrids of the two. They’re currently implemented in a number of different ways:
Smart Answers App
The majority of Smart Answers have been built using a Ruby DSL in the Smart answers app, a non-ActiveRecord Rails app. The DSL describes the possible journeys through the questions and possible outcomes. It makes heavy use of the Rails I18n API with the default YAML backend to store the question & outcome content.
A couple of these Smart Answers have been converted to use, or written using, the internally developed Smartdown library. The latter extends the idea of the Govspeak library, which itself is an extension of Markdown.
The idea behind the Smartdown approach is to make it easier for content designers to create and update the Smart Answers without needing to involve developers.
Publisher App / Simple Smart Answers
Simpler Smart Answers have been authored using the Simple Smart Answer format in the Publisher app. As with other formats in the Publisher app, the data is saved to MongoDB. However, in contrast to other formats, Simple Smart Answers save a graph of objects representing the user’s possible journeys through the questions and the possible outcomes.
The Frontend app retrieves this graph of objects from MongoDB via the Content API and displays the questions to the user, accepts responses to the questions, and eventually takes them through to the appropriate outcome.
Although Simple Smart Answers are implemented separately, they’re intended to look and feel the same as Smart Answers. Here are a couple of examples:
Smart Answers in this form can be created/updated in a draft state which allows them to be previewed before publication.
One other Smart Answer has been built in its own non-ActiveRecord Rails app, the Calculators app. This was built in a separate app, because it needed some extra functionality e.g. the ability to change the number of questions asked on a page depending on the response given to an earlier question.
These Smart Answers have proved very popular with users, but it’s turned out that in their current forms, it takes a considerable amount of time and effort to develop new ones and even just to keep the existing ones up-to-date with the latest rates, changes in policy, etc. In particular it takes a lot of developer effort which is in short supply.
Our brief is to take a bit of a step back and work out how the development and maintenance of Smart Answers can be put on a more sustainable footing. We really appreciate that GDS has presented us with a problem to solve, rather than a solution to implement. It’s definitely going to be a significant challenge, but hopefully we can come up with a good solution.
Monday was a bank holiday, so we spent Tuesday-Thursday working on the Smart Answers project (see above), although we worked remotely from GFRHQ on the Wednesday.
Over the bank holiday weekend I finally got round to a few things:
- Processing the AIS data I’d collected from a ferry on the way back from the Isle of Wight.
- Publishing a Tax Calculator I built to help us decide on the mix of salary versus dividends to pay ourselves.
- Playing with Google App Scripting.
We spent Friday mostly working on GFR admin, although we did manage to fit in lunch with Tom and Murray at the Strongroom Bar & Kitchen and a couple of beers at the Shoreditch Works mixer at the end of the day.
Chris spoke to PRS to update the address on our license, but discovered that frustratingly they are non-refundable and non-transferrable, so we had to buy a whole new license.
Anyway, that’s all folks. I’m sorry these notes were so late. Until next time.