Test and Trace - A failure in IT procurement?

Regardless of where you live on the planet, you will be aware of Covid-19, the novel Coronavirus strain. It's affected all corners of the globe. Depending on how good your Government is will also determine how well your country has dealt with it, and at what cost (a notable well-done to New Zealand, based on UK media, for getting things back to normal quickly).

Here in the UK, the response has been abysmal; money has been thrown at the problem like cost doesn't matter. A great example of this is the 'Test and Trace' system, regularly billed by UK politicians as "world leading". It's also been named 'Track and Trace' and 'Test, Track and Trace'.

Disclaimer: This post is a political discussion centred around the UK Government’s IT procurement, and the checks which should have been taken in relation to the ‘Test and Trace’ system. Please act responsibly to prevent the spread of disease.

It was announced on 16th January 2021 by the Public Accounts Committee that by March 2021 the system will have cost the UK taxpayer £23 billion.

What this highlights is why Governments need impartial IT experts to look at the costing of any IT projects, to provide a realistic insight into the whole thing. Yes, that would remove the nepotism which comes from knowing a prominent MP when something like this happens. It would remove the exorbitant profiteering which seems to run rife within Government contracts which, I speculate, comes as a result of MPs having significant shares in the bidding companies; large profits make for good dividends!

I digress; though it seems to be the only plausible explanation for the cost.

Why?

Looking at the cost-per-day of the project (based on the first 365 days), that's £63,013,698.63. Sixty Three Million Pounds. Per day! That works out around £1 per day, per person (population estimate for 2020 is 67,886,004 on Wikipedia).

According to the UK Government (based on data from ONS), the median weekly wage in the UK is £585; this works out at £30,420 per year. The amount paid for the Test and Trace system could give over 750,000 people the median salary.

Okay, giving money to people wouldn't have solved the crisis, but it helps to show the scale of the cost.

To further illustrate why this has been a monumental failure in cost management, we can look at some of the tech giants out there.

Netflix

Netflix is huge; they serve content to 204 million people in 190 countries. They make their own content, licence content from other studios, and generally keep people entertained. It's well known that they run their infrastructure on AWS, and that AWS is not exactly cheap.

However, being a business with shareholders, Netflix has a responsibility to keep costs down so they can make money. I looked at their financial records from 2019. It shows that their 'Cost of Revenues' specifically relating to 'Technology and development' was $1,545,149,000 (with an exchange rate of $1 = £0.71), that's £1,120,619,312.25 - about 5% of the cost of Test and Trace for a year.

Facebook

Facebook has something everywhere. Many different technologies and various avenues to make revenue. It is estimated that they have 1.8 billion active users per day on their platform. Serving content from a complicated network of friends, businesses, fan pages and other content. It takes a lot of resources to do that, particularly at the speed they do.

Their accounts from 2018 (when they only had an estimated 1.4 billion active users per day) show the cost of running Facebook in 2018 to be just under $31 billion. Of that, $4.15 billion was for share based compensation, so the actual running and other associated costs was more like $27 billion. At the same 1:0.71 exchange rate for the Netflix comparison, that is £19.41 billion.

That covers the cost of Facebook, Instagram, Messenger, WhatsApp, and Occulus.

Twitter

The social media giant, which was the preferred battleground of Donald Trump during his Presidency, has around 330 million active users per month. Again, it has some of the complicated personal networking and "relevant" post integration as Facebook to make money, but it does well.

Twitter's annual account for 2019 show the total costs and expenses (not just technology) at Twitter to be around $3.1 billion; £2.27 billion.

Reality Check

For the £23 billion cost of the UK's test and trace program, you could have run all of Twitter (2019), all of Facebook and associated companies (2018), and covered all of Netflix's 'Technology and development' costs (2019). A total of £22.8 billion, if you were wondering.

Instead, the UK has a system which has repeatedly failed to hit its goals. The same system which requires contact tracers to contact the same household for each individual member of a family. And the very same system which encouraged people to self-isolate, but which prevented them from claiming financial help.

If anyone with a sensible head was approached regarding this program and told the annual cost to run track and trace would be £23 billion, there is no way it would be signed off . A single program which would cost as much as the combined Government spending on "Environmental protection" and "Housing and community amenities", or 57% of the nation's defence budget, would never pass a sense check.

The numbers simply aren't grounded in reality.

A private business has to answer to, and provide value for, its shareholders. A Government should have the same accountability to its citizens as a business does to its shareholders. In the case of Test and Trace, it is the UK taxpayer who has to foot the cost of the system, despite the failings in cost management.

A possible reason for the excessive cost of the system may be related to some misconfiguration in the tech stack used. Without actually pointing fingers, Google’s Firebase has been known to generate large bills if used incorrectly, and AWS misconfigurations can be similarly alarming. If that is the case, then it's a solid case that the contract was given to someone who doesn’t know what they are doing, and the process needs to be reviewed.

What the UK government (and Governments of other nations) needs is a panel of IT experts to review the requirements of projects, and to vet them for inflated prices. Even in times of a pandemic, there should still be a proper tendering process, and things like Test and Trace shouldn't just be handed out lightly with a blank cheque.

Personally, I would be shocked if the actual cost to suppliers of the Test and Trace system came to 10% of the billed cost. The cost of manning call centres is £720 million for a year (18,000 trace staff indicates an average £40,000 per year, assuming cost is purely for staff, not other overheads). I can't imagine the cost of any software hosting being over £10 million a month (£120 million/year). The cost of developing the app shouldn't have topped £500 million for a rushed job with a lot of overtime. Even doubling the £10 million a month hosting cost; with the other estimates, the whole program should have come to around £1.5 billion for a year. It would still be a cost of £4.1 million per day, but a lot easier to imagine.

It would be interesting to see what the project actually cost to run vs. what was actually billed; and how the numbers came about. That's never going to happen because it would show some seriously underhanded tactics, or conflicts of interest in the process. But from an academic side, understanding how it cost so much would be great.

Getting Closer to the Truth

As you dig further into the whole ‘Test and Trace’ system, it eventually becomes apparent that the £23 billion is not all down to IT expenditure. Actual estimates for the cost of the mobile app put that at only (but still high) £35 million. That includes the initial app which was abandoned after spending £10 million. That appears to include the running of the app for a financial year, putting the cost of the app and staffing of the trace contact centre staff at under £1 billion.

What about the other £22 billion?

A lot of that comes down to understanding what ‘Test and Trace’ actually covers. It actually has four objectives:

1. increase the availability and speed of testing
2. identify possible close contacts of those who test positive, and asking those close contacts to isolate
3. rapidly identify and contain outbreaks
4. enable government to learn about infection rates and respond appropriately.

The speed and availability of testing is something the Government has been pushing to increase. New laboratories, more testing centres, generally a greater percentage of the population tested each day. They have a moonshot goal of testing 10 million people per day, which is estimated to cost an additional £100 billion per year. The cost of which would be attributed to ‘Test and Trace’.

Any positive cases identified will then have the system try and identify the close contacts of those who are infected, and those who may have come into contact with them. This is where the IT side, and the contact centres come in. As we’ve seen, this is only around £1 billion of the cost. But because the Government made a huge deal about the tracing app and the infrastructure around it, that’s all people think of when they hear about ‘Test and Trace’.

Identifying and containing outbreaks is where costs will spiral. In June 2020, local authorities in England were given £300 million to help with Test and Trace. Another £57 million was allocated to the devolved Governments of the Union. That was on top of the £3.2 billion which had been allocated previously to local authorities.

The Government scheme where low income families would be eligible for support of £500 if they had to isolate is also part of the test and trace system. There are around 4million people who would be eligible for that support, totalling £2 billion allocated to it.

Taking the amount allocated to local authorities to help with Test and Trace measures (£3.5bn total), plus £2bn allocated to low income families, and the £1bn for the app and contact centres, that’s £6.5bn. There’s £16.5bn which isn’t easily accounted for. I’ve not yet been able to account for the actual cost of testing, but the cost can’t fully be there either. Around 68 million tests had been conducted between April 2020 and 27th January 2021, which would mean each test cost £242.65 (staff, transportation, testing and returning results).

If the government is to expect 10 million tests per day to cost £100bn per year, then the cost for the test has to drop by several orders of magnitude. Or they are already a lot cheaper than that amount. Which begs the question; where is all of the money for the Test and Trace actually going? I cannot accurately say at this point.

What is apparent, is that it’s not all gone on IT overspending. Which, given the history of Government spending on IT, is actually refreshing.

Unfortunately, because any mention of “Test and Trace” is always shown alongside a picture of the app on a mobile phone, the insinuation that the cost is purely IT related is present in people’s minds. Mine included until I wrote this.

Yes, the UK Government - and probably others around the world - jumped into deals with suppliers without vetting or sense checking them first (PPE, anyone). Despite being critical to everyday life, IT and the associated costs are still seen by some as being excessive, unnecessary, and undesired. That makes the industry as a whole an easy direction to point the finger, deliberately or not, when there’s a high cost involved somewhere.

Is the Test and Trace program a failure in the IT procurement process? Possibly. But the overall cost of battling against the Coronavirus pandemic has brought about far higher expenditure which isn’t as easy to work through, and likely won’t be fully known or understood outside of very tight circles.

Conspiracy theory? Probably. Covering up high-profit backhanders? Likely. Just don’t blame it all on IT, please.

Special Thanks

Special thanks to Faye Sipiano for editing this article (up to the ‘Getting Closer to the Truth’ - she is blameless for typos from there!). You can get in touch with Faye via her website, or on Twitter (@JavaScriptCoff1).