COVID Alert

2020 - 2022
dead canada
six million nine hundred and forty eight thousand people downloaded it. sixty three thousand of them ever typed in the one-time key that told the app they had tested positive, which is under one percent of the installs. canada's exposure notification app ran for twenty three months on the google and apple bluetooth framework, collecting no location and no identity of any kind, and health canada's own evaluation named the gap between real positive cases and keys entered as the biggest missed opportunity. roughly eighty percent of the twenty million dollars went on advertising it.

~ the obit ~

COVID Alert launched in Ontario on 31 July 2020 and did one thing. Your phone broadcast a random identifier over Bluetooth Low Energy and kept a fortnight of the identifiers it had heard from other phones nearby. If somebody you had been close to later registered a positive test, your phone worked that out by itself and told you. The matching happened on the handset. The federal server saw nothing but a list of anonymous keys and an IP address, kept nothing past 21 days, and had no way to know who or where you were.

The code was not the government's idea. Volunteers at Shopify had built an exposure-notification app called COVID Shield in the open because nobody official was moving quickly enough, and released it free under a permissive license. The Canadian Digital Service took that codebase and built COVID Alert on it for Health Canada. The Canadian Centre for Cyber Security ran the security assessments, BlackBerry did an independent review, and the federal and Ontario privacy commissioners published a joint statement of support on launch day. Every repository stayed public on GitHub.

It never became national. Nine provinces and territories plus the Canadian Armed Forces signed on, but Alberta kept its own ABTraceTogether and Premier Jason Kenney formally refused the federal app on 6 November 2020, while British Columbia's officials said it would add significant challenges and never joined. Yukon and Nunavut stayed out too. An app whose entire value depends on the density of participation was missing two of the four largest provinces by the end of its first autumn.

Health Canada decommissioned it on 17 June 2022, effective that day, with the wording "Following a thorough review of COVID Alert, the Government of Canada has decommissioned the application effective today." The exposure-notification service went off immediately and people were told to delete the app. The reason given was that provinces had wound down widespread PCR testing, so almost nobody was being issued a key any more. The last commit to the repository landed on 23 June and it was archived.

~ the rap sheet ~

Born31 July 2020, Ontario only, announced by Justin Trudeau on 18 June and beta tested from 23 July
Peak6,948,717 downloads and 456,349 exposure notifications sent; 3.23 million were still running it in July 2021
Killed17 June 2022, decommissioned effective the day it was announced
Lifespan1 year, 10 months
Made bythe Canadian Digital Service for Health Canada, on COVID Shield code written by Shopify volunteers
Killed bythe end of widespread PCR testing, which cut off the only input it had

~ the one-time key ~

Everything the app could accomplish ran through a single step. Testing positive did not notify anyone by itself. A provincial health authority had to issue you a one-time key, by phone call or text or through a results portal, and you then had to open the app and type that key in. Only then did your phone upload its last fourteen days of random identifiers for other phones to match against. One chokepoint, entirely dependent on a person who had just received bad news choosing to do an optional administrative task.

The conversion numbers are the obituary. Health Canada's evaluation, covering participating provinces to 28 July 2021, found that of all the people who tested positive in those jurisdictions, keys were generated for 5.0 percent and only 3.7 percent were actually entered into the app. Over the full run, 63,117 keys were entered against roughly 6.9 million downloads. The evaluation's own term for this was the biggest missed opportunity, which is unusually blunt language for a departmental review of its own product.

There is a smaller figure buried in the same document that shows the whole chain working as designed and still not amounting to much. Between April and July 2021, more than ten thousand people who tested positive entered a key, which triggered over 66,000 notifications, which resulted in 640 people getting notified, going for a test, testing positive and entering a key of their own. That last number is the app doing precisely what it was built to do, counted honestly.

~ what turkey did instead ~

The design decision COVID Alert is proudest of is the one I find hardest to imagine from here. It collected no identity whatsoever. Turkey's app required my national ID number plus the last four digits of the serial on my ID card in order to issue a code at all, so the state knew exactly who every code belonged to from the moment it was generated. Canada built a system that structurally could not know that, and then spent two years unable to persuade enough people to use it.

Turkey dropped the HES requirement on 4 March 2022, a little over three months before Health Canada switched COVID Alert off. The difference shows up in what survives. Canada's App Store and Play listings both return 404 today. The Turkish app is still sitting on the Play Store, unremoved, years after the code it generates stopped meaning anything.

~ the evaluation ~

Roughly $20 million went into COVID Alert, split as about $15.9 million on promotion and advertising against $3.5 million on development and maintenance. Reduced to the number of confirmed cases the app is documented as having caught, 2,446 between April 2021 and May 2022, that works out around eight thousand dollars a case. The proportions are the part worth sitting with: four fifths of the budget for an app that 0.9 percent of its downloaders ever used went on telling Canadians it was there.

A McGill modeling study estimated the app averted somewhere between 6,284 and 10,894 infections across six provinces from March to July 2021, which is between 0.44 and 0.76 percent of recorded cases, and between 57 and 101 deaths. England and Wales got a far better return from the NHS COVID-19 app, which reached 37.8 percent of smartphone users against Canada's 28 percent and was credited in Nature Communications with roughly a million fewer cases in its first year. Germany's Corona-Warn-App got similar uptake and cost about €150 million against an original €20 million budget.

Health Canada's evaluators made one observation that goes wider than this app. They could not properly quantify effectiveness because nobody had set benchmarks or targets or public-health impact indicators before launch. The government shipped a national public-health tool, advertised it for $15.9 million, and had no agreed definition of what success would have looked like.

~ leave a rose ~

an anonymous tribute. one rose per visitor per day.

@}-->-- ... roses left here

~ leave a wreath ~

a rose fades by morning. a wreath stays on this grave for good, with your name on it.

from $1, paid once through stripe. it keeps the graveyard running: the hosting and the nightly digging. adopt it and the headstone reads "maintained in memory by" you. the groundskeeper plants new wreaths at the next tending, usually within a day.

» one email a week: new graves and the anniversaries of the dead. subscribe

~ leave a note ~

visitors before you have left these graveside notes. anonymous welcome.

other software: MECC, Wink Communications, MeeGo, Microsoft Producer for PowerPoint

also died in the 2020s: MTV Music Channels, YouTube Doubler, Blanja.com

« previous  ·  random  ·  next »