In 1997 I worked for a Member of Parliament of an Ottawa riding (The Boss) and I would go to work each day to do my job as a Member's Assistant and then, since my job was as much on the line as The Boss's, I would go volunteer at the campaign office in the evening.
Whenever we went canvassing there was always one member of the team who was responsible for recording whether a person we spoke to would support The Boss or not. They recorded this on a piece of paper on a clipboard. After each canvass that paper would be handed off to the data entry team at campaign HQ so that they could enter it into the computer. That was our database for pulling the vote on E-day. A few days ago when the Conservative candidate visited my neighbourhood I noticed someone with a tablet filling this role.
I provide this little anecdote to demonstrate that data and its analysis has been an important part of election campaigns for decades. With the increase in computer power and capabilities it has become one of the keys to winning a modern election.
With modern computing power, some of the extremely powerful data analytics software available on the market and well trained data scientists a political party can determine the probabilities of how an election will turn out at the riding level. Forget about the national, provincial or regional levels. Think about what all of the seat aggregators produce but on steroids.
So before this election was called the Liberal data team would have had reliable probabilities for every riding in the country. Since 170 is the magic number they would have identified at least 180 ridings where the Liberals had at least a 75% chance of winning (A large chunk of them would have been in the 85%+ range). To allow for some slippage they would then have identified another 20+ ridings where they had at least a 60% chance of winning. If they were unable to meet these objectives the PM would not have called an election. Being in power with a minority is better than risking an election unless you believe you have it essentially locked up.
The data team developed these probabilities using a whole host of data collection methods. Quantitative methods such as polling, but at the riding level and qualitative methods such as focus groups just to name two. They then would have crunched the numbers to come up with the probabilities and data analysts would have provided the "picture" to the party decision makers.
In addition these same teams would have tested all of the different policy proposals that have been released by the party and they would have tested the potential reaction to the early election call.
Once the election is called the data team has to switch to monitoring progress. For the ridings where they identified lower probabilities of winning they would continue to collect data, from polling and other data sources to look for changes in those probabilities. For the safe ridings they would probably not regularly poll them. They would use that money on less safe ridings. However, remember the tablet I mentioned in my anecdote? All of the data collected during canvassing would be uploaded to party HQ where the data analysts would see if there were any serious signs of trouble in those 180 seats with the 75% probability of winning. If they did that would be flagged. In addition they would be testing the reaction of voters to the daily announcements and actions and words of each of the party leaders.
All of this would be done and a daily dashboard would be produced to be presented to the people running the campaign, which would include the PM. Now here is the rub. This data analysis would not just tell the decision makers the probabilities of victory for their party but it would also provide the probabilities for the other parties. So the Liberals know where they stand each day but they also know where the other parties stand. The Conservatives, who have a similar operation, have the exact same information.
I do not know all of the intricacies because I am not part of the team but this is a general description of the importance of data to modern elections.
Of course, there are limits to what data can tell you and it is not uncommon for people to misinterpret what the data is telling you. As well, although these data can provide probabilities that does not constitute a guarantee. There is always a chance the probabilities could be wrong. We can never discount that possibility.
However, I will continue to point out that one of the reason why the Conservatives were so hysterical about the early election call was their data teams were showing the same things the Liberal team was showing. The Liberals had at least a 60% probability of winning more than 180 seats. As well, I will also continue to point out that although the public polls are saying the Liberals are losing support neither the Liberals nor the Conservatives are showing signs that their own data analysis teams are in agreement with those polls.