From the start, this campaign made a commitment to full transparency. To honour that commitment, we held a debrief session with the entire team invited to participate; analyzed the data we had gathered; and have created a report on the campaign's achievements and mistakes, found below.
It is also apparent that there is an appetite for moving forward together, and all volunteers, pledge signers, and supporters are invited to be a part of shaping what comes next - but more on that later.
Campaign Overview
Given the short timelines from inception, to launch, to election, our model was built to scale up quickly, and to give volunteers and supporters an easy route to plug in and immediately begin effective work.
For a more fulsome explanation of our model, you can read this document, but as a brief overview, we had two angles of attack: first, due to our very short timelines, we build an online amplification team that would grow our visibility on social media, create content to help drive our narrative around stopping Ford and uniting the vote to do so, and would scale up quickly by incorporating new people regularly in an expanding network. This visibility would then be leveraged to organize people on the ground in each of our target ridings, to sign up "Unity pledges" who committed to vote for our endorsed candidates, and who would themselves reach out to their neighbours, friends, and relatives to gather more pledges and later to support our candidates directly. Meanwhile, we'd gather data to plug into our model to inform our prediction calls and Unity endorsements.
Teams
All of this would be supported by four key teams: creative team to generate the content, amplification team to push it out to our channels and to rally new people to join, research team to gather and analyze our endorsement data, and organizing team to hold it all together and mobilize people on the ground. And it was all supported by a substantial tech infrastructure, including a fair bit of automation to make it quickly scalable.
The amplification side was an unequivocal success. We earned a great deal of coverage and visibility, we quickly ramped up our amplification network, and our messages were omnipresent throughout the campaign. While we are still digesting the specifics of our analytics, a good topline number is our Twitter reach - we earned over 3.3 million impressions over the course of the campaign. This, from an account that started with zero followers and without a single dollar of ad spending (in fact, our entire budget for the entire campaign was just a couple hundred dollars).
Partly as a result of this ubiquity, we were one of a handful of dominant factors in shaping the narrative of the election - we earned plenty of media coverage, but even when we weren't directly participating, our narrative was part of the discussion. Every party was forced to comment on strategic voting and the possibility of collaborating; after the election was over, we've witnessed a surge of articles discussing electoral reform, the possibility of both short-term alliances and long-term mergers between ONDP and OLP, and the relevance of uniting the vote. While there's no doubt that there would have been some talk of this regardless, the tell is in the framing - our key messages regarding the urgency of collaboration, the distance between Ford and the other three parties, and the utility of working together suffused the coverage both during and after the election.
It wasn't just amplification driving this narrative. Obviously it's subjective, but we feel our creative content was excellent - it is our view that our grassroots, unpaid operation created some of the hardest-hitting, most creative, and most memorable messaging of the cycle, which could compete with anything the other campaigns and advertisers put out. We have some independent verification of that - certainly our messages got very high uptake on social media, but on top of that, Elections Ontario expressly indicated to us that they wanted us to track our production because they felt it was clear that that it was all professional content. On the contrary - many of our best pieces were made by volunteers with next to zero experience in video editing or graphic design. That said, we did indeed have some real experts on the team too, and a professional comms plan as well as an excellent outreach operation to influencers and news media.
Our research team was also largely successful, but we'll talk about them more specifically below.
By contrast with our air war, our on the ground organizing took time to get moving, and never ramped up to the numbers we needed. We generated a great deal of resources and toolkits to support on the ground organizing, much of which did indeed get used. But it was difficult to convert people from pledge signup to active organizing in the short timeline. And our onboarding process was streamlined over time - but it could have been done more efficiently from the start.
We also held off our endorsements til rather late in quite a few cases - this may have been the right call, maybe not, but either way it made it hard to get energy behind our particular candidates. An earlier endorsement would likely have helped. That said, even if we'd endorsed out of the gate, six weeks is just not a very long time to build a complete field operation in 13 urban ridings.
Despite these challenges, the organizing team deserves tremendous credit. We had flyers, sometimes in multiple languages, in volunteer hands within 24 hours of endorsement in our target ridings. While our numbers on the ground were not as high as we'd have liked, they were very consistent - our volunteers were dedicated and unwavering.
Tech
Building a campaign like this in 6 weeks would have been, frankly, impossible, if we hadn't made excellent use of technology. We automated extensively to ensure that we could scale up quickly. In addition, our creative impact was in part due to excellent tools that overcame any missing experience. A full explanation of our tech stack can be found in our model overview document.
Prediction Model
So let's be clear: nobody likes strategic voting. We would rather not have been doing this. But if we were going to do it - and by March or April it had become clear that we had to do something about the fractured anti-Ford vote - then we wanted to do it right. Simply scanning the polling aggregators is insufficient; riding level results in the aggregators are extrapolations, not actual local polling (usually). It's easy to get things wrong.
As a result, we built a very thorough research model that incorporated an "aggregate of aggregates", pulling from multiple sources which were weighted by our outstanding data team. In addition, we looked at factors that couldn't be captured by aggregators - we interviewed candidates where possible to get a sense of their campaign operation, we looked at their history in the riding and incumbency, and tried to incorporate those qualitative pieces as well.
As an example, while all the aggregators showed a tight three way race, it was the strength of Doly Begum's campaign operation in Scarborough Southwest that tipped her over the edge for us and led us to endorse her in the first wave. Similarly, it was the strength of Mazhar Shafiq's operation that decided us in his direction in Scarborough Centre.
We're proud of that model, and largely, it delivered. But let's talk for a minute about where we went wrong (everyone loves to tell us about where we went wrong!)
First, in Humber River Black Creek, we had initially identified an NDP endorsement on the strength of incumbency. However, it was a tight three way race, and near the end of the election there was some movement in the polling numbers towards the Liberals. Ultimately, we blinked on our qualitative factors and let the numbers decide us - and that turned out to be the wrong call. This, of course, is exactly why we built the qualitative elements into our model, so it's unfortunate that we backed away from it at the last minute. But we did, and as a result, we called it wrong. Fortunately, the NDP candidate pulled it off anyway (and HRBC was not a region where we had a heavy presence in terms of pledges).
In addition, although we didn't technically "get it wrong", we did not wind up making a call in York South-Weston; we had initially identified an NDP endorsement based on incumbency in a tight three way race, then backed down in the face of the polling data and leaned towards a Liberal endorsement, and ultimately decided not to risk it at all. But the reality is, we had campaign strength data that should have put us over the top for Faisal Hassan of the NDP, and we consider our failure to endorse Faisal a miss as well. And since we did have some strength in YSW, this one hurts too.
Other than those two ridings - one wrong call, and one incomplete call - our candidate either won or came in second every time, or we (correctly) decided not to endorse due to low risk of PC victory. That's a fairly strong accuracy rating. And our data team deserves incredible credit - a lot of work and skill went into sourcing data, building the analysis model, and assessing the outcomes. Any mistakes made in the final judgment were solely the responsibility of leadership; the data team did everything right.
Our final assessment of the model: we think we're onto something, it performed relatively well, but we should have stuck to our guns and had more confidence in our qualitative side, and that's a lesson learned for next time (if there is a next time).
Volunteers
The final, and unequivocally most important, piece of the puzzle was our volunteers. We talked about the teams and their tasks above, but here we want to talk about the people who made this happen.
When our campaign launched, it was just one person with a laptop. Within 24 hours our material was all over the web; within days we'd grown exponentially, and by the end we had hundreds of people involved as volunteers in one fashion or another and many more as pledges - all in the span of just six weeks.
It was clear immediately to us that we'd tapped into real energy - people were looking for our message and were drawn to our campaign.
Our analytics also showed some interesting demographic data. Our campaign and our supporters (and the people who engaged with our content) were overwhelmingly women (around 80%). We had representation from across the city, although some areas we were stronger than others.
Crucially, our message needed to resonate across party lines and it did. We drew New Democrats and Liberals in roughly equal measure, plus a handful of Greens. Excitingly, we also drew unaffiliated people, disaffected Conservatives, and many people for whom this was their first foray into organized politics.
We drew people from all walks of life and with a diverse array of backgrounds and skillsets. One of the strengths of our model is how quickly people were able to plug in and contribute to the max of their time and ability - we gave people many ways to contribute, and put as few barriers in their way as possible. This allowed people to self-direct and move very quickly and was a cornerstone of our scalability.
Most importantly: throughout a grueling campaign, holding a diverse array of people from every partisan background imaginable, our volunteers showed tremendous courtesy, compassion, and determination. We had a code of conduct and moderation system, but we never really had to use it. Everyone was there to get things done together. We had open, frank conversations, and people felt empowered to speak their minds and then make decisions collectively. Despite our many differences, we quickly cohered into a fierce, capable, and effective team.
Conclusion
Obviously we would have liked a different outcome. Let's not mince words: we had one objective, and we failed. Despite the accuracy of our prediction model, only one of our endorsed candidates won. And despite the reach of our messaging, we weren't able to move many votes or puncture Ford's momentum.
We may have been able to have more impact with a smoother on-the-ground organizing rollout, earlier endorsements, or of course, more time to ramp up. We also may have been helped if one or more of the opposition parties had been better able to craft their own counter-narrative; and of course, the dream would have been open endorsement of our campaign by one or more of the opposition parties such that we had the kind of collaboration and unity we needed. Or runaway success by one of the opposition parties - if we'd run this campaign in 2018, for example, we may well have been able to tip the scales towards the ONDP, which would have received almost all of the endorsements.
But despite the failure to hold Ford to a minority, we did something truly remarkable in an incredibly short time. We are all deeply proud of the work we did together and the team we forged together. And the success of our message and our narrative has laid the groundwork for what's necessary for next time. Furthermore, we innovated, big time. There's never been a campaign quite like ours: fully open-source, every element shared publicly for replication (and some groups did choose to replicate it), decisions made collectively with people from every party represented, technology leveraged quickly and effectively and innovatively (in some cases we were learning how to use a piece of tech and deploying it to the campaign just hours later). To have done all that on pure people power; to have taken a couple hundred bucks and a few hundred determined volunteers, many of whom were new to politics, and gone head to head with campaigns that have tens of millions of dollars and dozens of full time staffers with sophisticated professional campaign operations - when all is said and done, everyone who was part of this campaign can feel damn proud.
What's Next
In our debrief session, there was a clear appetite to continue working together. There's a lot of ways that our core mission - uniting the opposition to effectively oppose Ford - can be advanced. Our members have talked about electoral reform, working to foster collaboration between the opposition parties, running more amplification and messaging campaigns to counter Ford's big-dollar ad spends, increasing voter turnout and support for democracy, and plenty more. We've kept our organizing space open and are meeting again in the weeks to come to determine how we can best move forward as a team.
No matter what direction things take, we know that we can count on each other. We know that our shared values and priorities are more enduring than partisan differences. We know that we can learn, and grow, and innovate together as a team with trust in our shared competence and patience for our shared mistakes. And we know that there is an appetite for a more collaborative approach to progressive politics in this province.
As we said at the beginning, there is more that unites us than divides us. We have three parties that have very real policy differences, but also share many values - and are all much closer to one another than the outlier PCs. And we also have a majority of the province who didn't even see a reason to vote in the recent election. To us, it seems clear that the time has come for a more unified approach to not only opposing Ford's out-of-touch PC government, but also to building what comes next. Whatever happens next, we are going to be part of it - and we are going to do it together.