Predicting which visa applications to fast-track

Speeding up visa processing to save staff time and money

We built a prediction service to allocate visa applications to either fast track or review track streams, based on previous decisions. Handling applications appropriately brings faster and more reliable decision-making.

Demand for visas increasing

A labour intensive application system at Immigration New Zealand (INZ) was requiring growing numbers of staff to meet an increasing demand for visas. We were contracted to build a front-end triage system to automatically assess each visa application upfront. This initial risk assessment grades applications into those that can be safely fast tracked and those where a more thorough investigation of the applicant is warranted.

Ongoing value from data

Since all applications are graded, management can decide where they place the cut- off between fast track and review track applications. This feature of the system allows INZ to choose a level of risk for the visa processing that meets their business criteria. It also enables workflow and resources to be managed.

How we did it

Analysis and predictive modelling were carried out by Dragonfly, with technical review from staff at Victoria University. We then ported the model into the INZ operational environment, running it against a near real-time copy of their database, and delivering predictions within 10 minutes of applications being entered into the system. The prediction service makes use of data from past applications, delivering ongoing value to Immigration New Zealand.

What Dragonfly were able to drag out of our database was pretty impressive – they did what we thought couldn’t be done.

This triage system lets us decide our level of risk. We can change the settings to an acceptable level, but we still get to choose where we want the trade-off between cost and risk to sit. We haven’t lost that control.

We worked out the cost saving from this project, in terms of the extra people we wouldn’t have to hire to cope with the increased visa processing. It was significant.

Steve Cantlon
Principal Business Adviser
Immigration New Zealand .

You might also like these projects

work
Better report production one step at a time
Our work with the Ministry for Primary Industries replaced a manual production process with an automated system. New data can now be piped in to recreate each year’s National Exotic Forest Description – this is significantly faster, cheaper and more accurate than building the report from scratch using Word, Excel and InDesign.
work
Open data on protected species
We worked with the Ministry for Primary Industries and the Department of Conservation to make information with high public interest available online.
work
Automating data input and verification
Each year, detailed data is collated by the 40,000 or so resource consent applications made to New Zealand’s district and regional councils. This information is critical for understanding how the resource management act is functioning in New Zealand.
work
A new language processing app
Machine learning methods are ideal for classifying streaming data rapidly. We developed Kōkako, a language monitoring app to identify different languages broadcast on radio stations in New Zealand. Kōkako won a Best Design Award in 2016.