A nightmare on Downing Street

A cautionary tale of how to fail at data storytelling – with a few jokes thrown in

During the hastily cobbled-together press conference to announce and justify lockdown 2.0 on Halloween, I could tell something strange was happening. At the time, I was watching Liverpool nervously climb back to the top of the Premiership in a stuttering 2-1 win against West Ham.

I wasn’t dual-screening, watching the game on TV and the announcement on my phone. But my phone was buzzing every couple of minutes. Friends and collaborators were asking with increased interest if I – one of the only data storytellers in their circle, for we are a rare breed – was watching. Was I aware of the catastrophe of hapless narrative by numbers unfolding on screen?

The supposed A-team had been wheeled out to tell us why we were going to have to stay at home for four weeks: the Prime Minister and his Chief Medical and Scientific Officers, the normally redoubtable and trustworthy Whitty and Valence.

In the early days of the first lockdown, there was something almost reassuring about this gang of three appearing nightly on our screens at 5pm, demonstrating how they were “being led by the science”. A familiar slide deck would be presented with a gloomy, doomy talk track: hospital admissions, ventilator bed capacity, deaths, miles not travelled on different forms of transport. In a Kafkaesque, Rumsfeldian world of unknown unknowns, the daily press conference became destination TV. It provided fodder for virtual watercooler moments with colleagues, clients, and collaborators on Zoom the next day.

When Johnson caught COVID and was sidelined for weeks, the central role was shared among cabinet members. Some of the performances – from Gove, Hancock, Rabb, Sharma, Shapps, and Patel – revealed two things. One, they were taking it less seriously. And two, the de facto PM, Demonic Cummings, truly has assembled a D-grade confederacy of dunces to front his algorithomocracy.

Often they’d stumble over the data and defer to their expert wingmen. This included Gove, whose most famous Brexit soundbite was “We’ve had enough of experts … with their acronyms”. Patel’s “for one night only” appearance even created a whole new category of numbers, when she smirk-boasted that the U.K. Government had conducted “three hundred thousand and thirty-four nine hundred and seventy-four thousand” tests for COVID-19. The satirical response ensured that Patel’s first mingerlin’ appearance at a daily press conference briefing was also one of her last. Here it is, in case you missed it.

Britain stayed at home and we started to get to know our Rs from our elbows. As new cases and deaths dropped, the daily presser became an unpredictable, when-we-feel-like-it event, sometimes not even once a week. To get the economy moving, we were encouraged to go on holiday and – in a slogan every bit as thought-through as Kraft’s rebranding to Mondelēz (get a Russian friend to translate for you) – “eat out to help out”. Two initiatives that seemed counter-intuitive and which, subsequently, have been shown to be principal components of the second wave. Have a look here and over there. Just like former president Trump’s ultimately pointless rallies, indeed, which generated an estimated 30,000 infections and 700 deaths.

The Halloween press conference was put together under extreme time pressure. Some cad had sneaked out news of the lockdown announcement to The Times on Friday night and they’d splashed big on Saturday. The leak has led Downing Street to launch a leak inquiry, no less. Johnson, the man who had wanted to be PM all his life but who looks like he’d throw it all away tomorrow – if only Cummings would let him – had been planning a “Monday at five” event … until the Thunderer wrecked their weekends.

The resulting slide pack – retrospectively tarted up and published on the ONS website – is here. It’s fair to say that the data storytelling given by Johnson, Whitty, and Valence truly was a nightmare on Downing Street.

Ping! First message. “Are you watching this? It is an object lesson in how NOT to do narrative by numbers!”

Bzzz! “These slides are a car crash.”

Parp! “Constant change of reference points. No idea what the axes refer to.”

Bong! “No understanding of who they’re talking to or how this will go down.”

Quack! “Can’t we spend just £60K of the £12bn for track and trace on decent data visualisation?”

Ting! “You could retire on what you’d make from training this bunch of clowns.”

In the ten days or so since the horror show of data storytelling was unleashed on the world, the UK Statistics Authority has come down hard on the way that the Government presented data to justify the new lockdown, citing issues of transparency, sources of data, and clarity of presentation about the predictions it made. The most egregious error or exaggeration included using since-discredited worst-case scenarios to scare us into staying home. This was the worst offender in the slide pack.

The statistician’s statistician and one of the world’s best data storytellers, Professor Sir David Spiegelhalter, said the whole episode had been “really unfortunate”. Genius in its moderation and economy of language. There’s a fuller explanation of his critique here. Sir David Norgrove, the chair of the U.K. Statistics Authority, said: “I recognise the pressures faced by all those working on decisions related to coronavirus. But full transparency of data used to inform decisions is vital to public understanding and public confidence.” This isn’t Johnson’s first brush with Norgrove. When he repeatedly trotted out Cummings’ “£350m a week to the NHS” number long after they’d used it to gerrymander the Brexit vote, Norgove sent the then Foreign Secretary this terse note.

Don’t get me wrong. I’m no second wave denier, and I was with the body of evidence that called for a circuit-breaking firebreak some weeks ago. My quibble – like those of my text, email, and WhatsApp correspondents – is with the data storytelling. The presentation of poorly-presented, poorly-sorted, poorly-ordered data in an attempt to browbeat a punch-drunk populace into submission. Cummings and his acolytes might have been irritated to have their weekends interrupted by the need to get the charts together two days early, but that was no reason to delegate the task to a barely numerate year five student with a flair for PowerPoint and little grasp of how to tell stories with data.

Now, I’m also no data visualiser, and my complaint and that of my correspondents wasn’t so much about dataviz per se. (Aside: the best COVID data visualisation comes from David “Information Is Beautiful” McCandless at https://bit.ly/36ahkup). We were much more frustrated at the gang of three’s fundamental inability to abide by any of the six Golden Rules of data storytelling.

  1. Keep It Simple: completely violated. As the dream-team that brought us Brexit and have been trotting out three-word slogans ever since, Johnson and Cummings should have realised that trying to drown us in a barrage of data is counter-productive. Cognitive psychology 101 tells us that we dig in when people go over the top with too much data. It’s what led Cummings to label Cameron and Osbourne’s half-arsed, data-heavy defence of EU membership “Project Fear”. Er, Pot? Meet Kettle!
  2. Find and use relevant data: much of the data presented is relevant to the pandemic, but how much of it is relevant to the decision to call a new lockdown? In the panic of Halloween, the team in charge of the slides made the cardinal error of believing “more is more”.
  3. Avoid false positives: anyone who’s had a single lesson in statistics knows that “correlation is not causation”, but the glum trio seemed perfectly happy to make that cardinal error time and again, drawing causal inference from chronological coincidence.
  4. Beware the Curse of Knowledge: in his book A Sense of Style: The Thinking Person’s Guide to Writing in the 21st Century, the Harvard psychologist Steven Pinker defines this common affliction thus: “The main cause of incomprehensible prose is the difficulty of imagining what it’s like for someone else not to know something that you know.” And the principal offenders? Step forward academics, financial advisers, lawyers, scientists, and … Government officials.
  5. Know your audience: the fundamental key to better data storytelling is empathy – understanding who’s going to be listening to what you’re going to say, putting yourself in their shoes – their mind, their mindset – and tailoring your message for them. The charity National Numeracy has found that more than three quarters of the U.K. workforce have mathematical skills that wouldn’t scrape a GCSE pass, and half of us wouldn’t pass SATs at the end of primary school. How Johnson’s Great British public was expected to make sense of such a tsunami of data – wrong, badly-presented, off-the-charts shambolic – beggars belief.
  6. Talk Human: analytics and storytelling, when put together with skill, compassion, and empathy, are not the fire and ice bedfellows you might assume. In fact, if you adopt that rarest and most refined of dialects (Human) then you come to understand that the fundamental equation of the modern knowledge economy is Analytics + Storytelling = Influence. Meanwhile, Johnson + (Whitty * Valence) ^ COBR’s awful slides = Confusion.

In August, two professors from Seattle – Carl Bergstrom and Jervin West – published a brilliant book called Calling Bullshit: The Art of Scepticism in a Data-Driven World. The book, which also exists as a MOOC, should be required reading and viewing for everyone in Downing Street. It shows clearly and simply how to build and strengthen a critical faculty in the age of Big Data. It clearly hasn’t made even a ripple in Whitehall just yet, and on the strength (or weakness) of the nightmarish performance on Halloween, I’m calling bullshit on the whole shabby shower.

And I didn’t even have to bring up Barnard Castle …