A history of Emergency Alerts in the UK. Part 3: The final push.
Following on from Part 2, we’re now in late 2020 and the national project team is flat out across a number of work streams to get the service in place. At this point we are still targeting an early 2021 launch and in the thick of testing.
First end to end tests
The infrastructure all needed to be set up on the mobile test networks initially in order to successfully send end to end messages. I recall the two Pete’s, Pete Herlihy (GDS) and Pete Davies (Fujitsu) working through some serious complexity with what became known as the ‘translation work’. Essentially mapping the delivery of CAP based messages to proprietary interfaces with Nokia and One2Many CBC software. Both the transmit and receipt function needed to work seamlessly.
The first successful Emergency Alert was sent on the test networks from the GDS Notify CBE on the 8 October 2020.
Once the work had been completed on test networks, configuration could move to the live networks. This meant getting ready to send live messages on test channels initially. In order to play it safe the tactical use of a mast located at Santa Claus Village in Finland enabled us to safely send messages on the production service without alerting the public.
Whilst the team was perfectly aware of the seriousness of what we were building and it’s purpose, there were moments of humour which any team needs working at the pace and duration we’d set ourselves. One example I recall is how the GDS team introduced animal references into the prototypes. The first iteration was even known as the ‘fainting goat rental’ and we had test messages for Alpacas, Capybara and in this case, overly curious llamas. None of these messages were created or sent on the production service of course.
We also created mission patches that I’d first started with the original cell broadcast trials in 2019 and to add a bit of theatre to the proceedings a launch controller was skilfully converted into a mouse by the talented Andrew Armstrong and served as essential kit for generating broadcasts.
Usability Testing
The team now started to focus on usability testing, awareness and communications. A significant amount of work on the mobile network side had been completed and GDS were building out the central messaging engine (CBE) with real code. We now needed to get this into the hands of users for testing.
As the team wasn’t truly working in the open, this made it harder to engage others, particularly users that would be sending the messages. We approached contacts from the initial forums in the earlier part of the year. I also made contact with the Emergency Planning Society to recruit those involved in emergency response to test the service and to help shape its development through their feedback.
Once the service is live, GDS will publish blogs on the research and design as there were some valuable insights from this work so I’ll not cover too much detail here. The design of the CBE evolved after every usability session. We covered a range of scenarios including Covid-19 messaging as well as other incidents such as industrial accidents, flooding and civil nuclear scenarios. The initial design of the CBE reflected the need to alert by authority areas as well as by country and UK-wide.
There was a need to include mapping to allow the user to validate that their selection was accurate together with the ability to add/remove targeted areas quickly and easily.
How to communicate the accuracy of any broadcast was also a challenge. The perception from users was this was targeted and by offering a selection down to ward level initially, this gave the impression of a level of accuracy which isn’t possible to achieve without device-based geofencing. In order to provide an indication of accuracy, GDS included a feature to show the likely number of recipients in the target area and those who may receive the alert. This certainly had the desired effect when during testing a user selected the whole country and was presented with figure in the many millions…
There were a whole suite of design changes that GDS introduced such as cancelling broadcasts, the authorisation of broadcasts, the use of templates etc. Fundamentally, the design was kept simple with the aim to make the service as straightforward as possible to access, compose a message and for this to be broadcast (the 2am..test). All the users we tested with were able to successfully send an Emergency Alert unaided.
Usability testing with the public
Several rounds of public research were carried out with a broad mix of individuals who were recruited to take part. They were not aware of the nature of research prior to the sessions. How the research was delivered and the results are worthy of a whole research paper in itself.
With the restrictions of having to test remotely and not being able to send real alerts (tests or otherwise) to recipients handsets meant the team had to simulate messages. This was achieved by users accessing a browser-based article on their phone through which an emergency alert would appear after 20 seconds. The research tested responses to the alert, recall of the alert, where they would go for further information and finally their thoughts on prototype content at gov.uk/alerts
After several periods of testing that included different message types representing a range of hazards and threats plus a national ‘welcome message’ the findings suggested that:
- For many users receiving an emergency alert can be quite startling.
- Users were initially mistrusting of the message using language such as scam and pop-up. The inclusion of GOV.UK helped provide some reassurance.
- The sound was generally the first thing that attracted their attention and often their first response was to silence this.
- Users have poor recall of the message after initially dismissing it, shorter messages performed better and words users most commonly recalled were emergency, alert, gov.uk
- The link to gov.uk/Alerts provides authority although users can still be wary of clicking this.
- Explosions, fires, terrorism, severe flooding were the use cases most commonly mentioned.
- Users would seek to validate the Alerts by checking the local news, speaking to friends and family, searching for information on gov.uk and visiting their local authority website.
- Many users expected to hear about the service in advance through an awareness campaign.
Emergency Alert Public Awareness Campaign
The requirement for an awareness campaign was confirmed in the early stages of the project and validated through the user research and experiences from other countries. The Cabinet Office, already heavily engaged with Number 10 comms in relation to the Covid-19 response brought in the creative agency 23Red. Huge thanks to Wendy Manuel, Nick Corney and Michael Paul in particular for their efforts on this.
23Red set to work identifying key themes, messages, creative content and designs for both the public testing phase and building for the national launch, which was now scheduled for March 2021.
The headline message is to raise awareness of the service and to ensure the public is ‘alert ready’. Supporting messages describe how the technology has been used to save lives in other countries, how Alerts are unlike other types of notifications, how they do not require a telephone number, how and when they might be used, and to direct people to gov.uk/alerts for further information.
What’s in a name?
There were a number of discussions on the name of the service which included testing with focus groups. This was ultimately narrowed down to four:
● Local emergency alert
● Public alert service
● Neighbourhood alert
● Local warning
Local Emergency Alert was favoured by various governance groups and scored more highly with the focus groups; “a name which captures the local nature of the service would be most effective”
I have to say I was never really warmed to the ‘local’ part and felt it would be too closely associated with smaller scale incidents that wouldn’t necessitate use of the service. Thankfully, towards the latter stages of the project, sometime during testing, the ‘local’ part was dropped. I recall Pete Herlihy pushing for similar simplicity when it came to suggestions of a logo/brand and that ultimately we need people to know what to do rather than recognise a brand. In the same way you don’t look for a logo/brand for a fire alarm, you focus on what to do when the alarm goes off. The decision to drop the logo idea and to just call the service what it is ‘Emergency Alerts’ felt like a victory for content designers everywhere.
The Demos Begin
By this point in the new year we moved from lab testing to planning for the public tests. This was initially on test channels which we named ‘Demo A’ followed by testing on public channels, perhaps unimaginatively called ‘Demo B’. The latter being heavily influenced by the timing of the Apple iOS update.
I led on the planning and delivery of Demo A which was held in Reading on 9 March 2021. The objectives were to test the alerting infrastructure end to end and to ensure successful transmission of a message from the alerting engine, through the networks, onto handsets.
The test HQ was the Environment Agency office in Reading with an additional site at Brock Barracks. Over 80 individuals from across the mobile network community together with Environment Agency staff and project team members support the testing and were based ‘in the field’ with a range of handsets including Apple Engineering devices.
As we were using the test channels, there was a small chance of members of the public receiving messages. This would be people who had found the emergency alerts settings on their phone and opted in. As a precaution we included a URL to a page on gov.uk which included a survey as well as a telephone number with a recorded message in case of public receipt. Subsequently, based on the visits/responses, around 30 members of the public received a test alert.
Three broadcasts were made which enabled us to test bilingual alerts, a range of different modes across a range of handsets, international roaming on a UK network, the expiry of broadcasts, cancellation of broadcasts, phones in-calls and data sessions during broadcasts and so on.
Using the production service, alerts were successfully sent to handsets in approximately 10 seconds from the point of issue to receipt. We were able to validate handset behaviour and potential ‘bleed’ of message receipt beyond intended target area boundaries. Importantly, we identified an issue which would have caused problems in sending messages at scale e.g. a national message which was essential learning and proved the importance of testing prior to any service launch.
Planning and delivering this test was a huge team effort. Particular thanks go to Pete Davies, Bob Wood, Pete Herlihy, Chris Hill-Scott, Ian Dewhurst, Oliver Sutherland, Kevin Knappett, Nigel Brown, Mike Carr, Steve White, Sean Walsh, Ryan Thorley and Heledd Quaeck for the support with the Welsh translations. Thanks also go to the teams at O2, Vodafone, Three UK, EE, Virgin Mobile, Giff Gaff, Tesco Mobile and Samsung.
The successful delivery of Demo A was my last contribution to the Emergency Alerts project. I was ready for some time off after working solidly for a year without any real break. I’d accepted a new role as a Service Owner at the Ministry of Justice Digital Prisons team starting in May. I’d already delayed the move as long as reasonably could with the hope the service would have launched in March/April 2021 but sadly that was not to be the case. Just before I did leave, I was able to make a visit to O2 in Slough (thank you to Michelle Macdonald) where we successfully sent Emergency Alert messages on the public channel to iPhones for the first time in the UK outside of a lab.
Demo A+ (not quite Demo B) — 25 May 2021
With the announcement of Apple’s release of iOS 14.5 which would enable emergency alerts on Apple devices in the UK, the team started to prepare for the full public test. It was felt the move from Demo A to Demo B was perhaps too big a step and an interim test was planned on the Suffolk Coast near Sizewell. The community here was already used to receiving text messages for annual testing and the authorities had also been involved with the original Mobile Alerting Trials back in 2013.
The team kindly extended an invite to this Demo knowing i’d be interested in attending. The focus for this test was more on the communications and public responses to messages. It was also an opportunity to check the successful receipt on Apple devices given these had not been tested at scale on publicly available handsets.
Three broadcasts were made using the Severe, Extreme and Government Alert channel which was included in order to test the override. This video taken on the day of the test shows the Government Alert being received despite opting out of all Alerts.
Demo B — 29 June 2021
With the successful test in Suffolk completed, the team now moved on to the ‘full’ Demo B. Working with the local emergency planning team this took place on the 29 June 2021 in Reading. I didn’t attend this one in person having only recently started a new role but through colleagues and seeing the public responses, the test seemed to go as planned.
Next steps & the future
Throughout the development of the UK Emergency Alert service, driven by COVID-19, I’d always sought to ensure use of the service for flood risk was high up on the agenda. This meant the Environment Agency was essentially in second place once Emergency Alerts was operational. The devastating floods in northern Europe in July 2021 once again put a spot light on the efficacy flood warnings prompting Germany to signal its intention to adopt a cell broadcast based alert service. It is perhaps only matter of time before we see a destructive event like this in the UK.
Given the EA already has a flood warning system, having to log into two systems felt sub-optimal. We worked with GDS to develop a system-system integration so that warnings triggered from one system, would trigger a message in the Emergency Alert service where it is appropriate to do so. This approach would allow other organisations with existing alerting systems to connect up using the Common Alerting Protocol. I’m grateful to colleagues at GDS, Steve White from the Environment Agency and Pete Davies from Fujitsu for continuing this work. I have no doubt it will save lives in the future.
So with the successful delivery of Demo B, the infrastructure being ready and awareness campaign materials agreed, DCMS was in a position to handover to Cabinet Office to take forward the launch. By this point we’d seen the departure of Pete Herlihy from GDS and Dr Nigel Brown retiring from Cabinet Office CCS. It would be down to the remaining team at CCS to launch the service.
There continues to be political interest in the service, Lord Harris recently published a follow up review on London’s preparedness for a terrorist attack some 6 years after his original review. The report suggests that CB is already in use by the Environment Agency which it is not although as the previous Service Owner I was keen we adopted it. The report also describes how CB could conflict with the run-tell-hide strategy in response to terrorist attacks. However, CB can be configured so that a specific channel is enabled for those type of alerts without the requirement for audio alerting. Even though this is possible, the fact no other country has opted to do so suggests they haven’t felt the need.
Once the service is live there may also be pressure to extend the service to include LBSMS. All I will say on this point is there are strong commercial interests in LBSMS. Implementation and operational costs are significant.
With regards to the Notify CBE build, this presents an opportunity for other countries around the world that have signalled their intention to adopt CB-based alerting platforms to re-use and benefit from this development. It de-couples the need to procure both CBE and CBC from a single supplier.
Coming back to the future launch, it has already been re-scheduled a number of times, firstly for Summer 2021…then moved back again to Autumn 2021 and then in the latter part of 2021 it moved to early 2022. Whilst a few operator tests have been carried out in-between there hasn’t been much visible progress since June 2021. [We now know the service was formally switched on in March 2023 with a national message 23 April 2023].
The difficult position is that there will never be a perfect time to launch. Whilst the service is technically operational, using it without an awareness campaign in advance would, as the research suggests, be a mistake. There will have been pressure to use Emergency Alerts ahead of the Boxing Day ‘Get Boosted’ message which prompted speculation that the service could be used. Instead the default was once again SMS messaging, sent in batches which given the nature of the message was probably the right call.
There is of course the risk that an incident will occur whereby the service could have been used and it wasn’t due to it not have officially launched. The longer we wait, the greater the likelihood of that occurring. The ongoing war in Ukraine is again putting a spotlight on the fact the Emergency Alert service is not currently in place. In the words of Nigel Brown, with one of his favourite phrases…CCS will:
“need to practice skill and judgement as they say on the side of cornflake packet competitions”.
A thank you to all those involved getting us to this point, it has genuinely been a huge team effort and a pleasure to have been involved since those early days. I shall await my national welcome message and hopefully won’t ever receive an Emergency Alert for dog poo on Keswick High Street….
