The spotlight is once again on polling companies who failed to predict the outcome of the presidential election, unable to see the swell in support for Donald Trump.
Before election day, they were showing a sustained but narrowing lead for Hillary Clinton following the FBI resuming their investigation into her use of a private email server.
As the polls shifted towards Trump, they moved stock markets and influenced people's betting behaviour. American data journalism site FiveThirtyEight gave him a 33 per cent chance of victory going into the final few days but this was generous compared to other forecasters.
Our poll tracker, which was based on an aggregate of polls taken from RealClearPolitics, showed the race was tightening between Clinton and Trump in the final week of the race, leading to a final gap of two percentage points between the Democrat and the Republican.
On Tuesday morning the poll tracker's final predictions showed that the election would be very close but that Hillary Clinton had less to do in order to win. She was predicted to be about two percentage points ahead of Trump.
This was out by three percentage points, with Trump leading Clinton by one percentage points.
A week before the election, the Telegraph's poll tracker was showing that Clinton could rely on 247 electoral college votes, while Trump had secured 164 votes from Republican states.
This picture changed as election day approached and the FBI re-opened its investigation into Hillary Clinton's use of a private email server.
On the morning of the election Clinton was predicted to win 206 electoral college votes while Trump was expected to get 164. The remaining 168 were toss-ups with big states like Florida, Ohio and Pennsylvania too close to call.
This compared to a final result of Trump gaining 266 electoral votes compared to Clinton's 218.
He managed to claim key swing states such as Ohio, Florida, North Carolina and Pennsylvania, jumping through the path that pollsters claimed was so small.
Previously, the New York Times worked out that even one day before previous elections, a simple polling average has differed from the final result by about four percentage points.
Which polls were right?
Not all polls were wrong. Over the last two months, 10 polls published on Real Clear Politics gave Trump the lead.
Nine of these were from from LA Times/ USC, and one was from Fox - out of a total of 93 polls between 8 September and 8 November.
Questions will now be about about why the LA Times was the only polling company to consistently pick out Trump's lead.
Polls in swing states
When it came to individual battleground states the performance was mixed.
Florida, for example, which has been on a knife edge, has divided pollsters. Trafalgar Group and Remington gave Mr Trump a lead, three other organisations, including Opinion Savvy and Quinnipiac put Mrs Clinton slightly ahead.
Our aggregate of all the polls put Trump and Clinton on equal footing going into the final day of the race. However, Trump won Florida by 1.4 percentage points after the votes were counted.
Ohio, which has maintained its record of siding with the winning candidate in every presidential election since 1964, was projected to vote for Trump but North Carolina and Pennsylvania were expected to go to Clinton. They both voted for Trump.
Based on the swing state polling it was hard to see how Trump could forge a path to the White House. It looked as though too much would have had to have gone his way in order for it to happen. However, Trump's support base turned out in greater strength in these areas than was forecasted.
