If you've followed Britain's #COVID19 saga you've prob noticed some serious issues with the data. No figs for how many people have been tested cumulatively. Big revisions in recent weeks. Questions from the @UKStatsAuth. So what's REALLY going on beneath the surface?... 🧵
Over the past few weeks @rowlsmanthorpe & I have had dozens of conversations with people within the testing system to understand what's going on. We've written a long-read. You might need to sit down to read it - not just because it is indeed quite long news.sky.com/story/coronavi…
What we discovered: major irregularities in terms of how testing data was collected. Some of it collected by hand on notepaper, some of it compiled by ministers phoning round to get numbers and put them into Excel. Statistical codes of practice were ignored.
Data on testing is central to our ability to manage and withstand a second spike of the virus. If we don't know who has #COVID19 or don't trust the numbers then how will local lockdowns work? How will we target care without more disruption? So this matters.
The picture that emerges from our investigation is of a testing system which struggled from the very beginning to collect robust, detailed data, struggled to link it back to people. That problem was particularly acute in the so-called Pillar 2 testing system in England
One source involved said: "It was barbaric. There were hundreds of people working on this new system, yet the national test results were being literally hand-written." Here is one of those hand-written notes - in this case showing national tests for parts of pillar 2 in mid-May 
Govt insider: "Normally when govt publishes data it takes three months... We didn't have that kind of time... We said: 'Forget the conventions, we're putting things out'. We didn't want to be accused of a lack of transparency. In doing so we probably broke some of the rules."
It is a complicated story and no single person or institution is to blame. And in @DHSCgovuk's defence the data is now much improved from those early days. After months of being unable to say how many people were being tested each day it can now produce a weekly estimate.
But those early problems still haunt it. The other day it revised the total number of cases by about 30k to eliminate some of the double-counting we document in our long-read. Even now @DHSCgovuk admits the data is not altogether reliable (see note from latest data release)
Upshot of those early months where, we reveal, testing data was thrown together in, let's say unconventional ways is that there are still black holes. Likelihood is we will never know for sure how many people, cumulatively, have been tested for #COVID19 news.sky.com/story/coronavi…
