The UCOVI Blog

The UCOVI Blog


First-hand insights on data governance in the world of politics




Ned Stratton: 12th September 2022

Part 2 - Votesource

So it was Truss wot won it in the end, and though her Cabinet might be cause for Dr Dre alarm bells ringing loudly, we do at last have a Prime Minister again in spirit well as in name. Though all of this has of course been over-shadowed by Liz's namesake in the Windsor family, whose 70-year reign has ended with her sad passing aged 96. When Queen Elizabeth II came to the throne in 1952, data analysts worked with pencil, paper, and abacuses, while infographics looked something like this, which does rather put the UCOVI Blog in context.

But I want readers to cast their minds back 7 years to 2015. A fresh-faced David Cameron was Prime Minister, and an EU-membership referendum was just a clickbaity policy in his arsenal to keep people from defecting to UKIP and shore up enough votes for a second coalition with the Lib Dems, who wouldn't allow the referendum anyway. So it would be totally fine.

I was working on the IT support desk at Tory party HQ in London at the time, supporting the rollout of their new campaign software and database Votesource, brought in to make data the secret weapon for David Cameron and the Conservatives in the general election due in May that year.

To quickly recap on Part 1, political parties in the UK need campaigning databases to:

  1. blend the UK Electoral Roll with public data on who votes and doesn't (the marked register) as well as records on support previously pledged (canvassing), membership, and other useful info such as social-strata indicators (MOSAIC codes), telephone numbers and constituents' nicknames.
  2. Govern and analyse this data so it can be used by party activists, who are operating out of commitment to the political cause rather than capacity or desire to work with computers or iPhone apps. This is important.

Up until now, the Conservatives had contended with the MERLIN database for this purpose – a motley old Knight of the Round Data-Table that had an unfortunate knack of losing important data by dent of its federated network architecture design, crashing during election campaigns, and leaving its users uncertain as to whether the TPS flag on a constituent record meant "Telephone Preference Service" or "Tory Party Supporter".

In this unsatisfactory context MERLIN had to be replaced by something better and more befitting of the dawn of cloud computing and phone apps, and Votesource came online in January 2015.

First Impressions

Votesource was met within Tory HQ with a sense of loving mockery - staffers thought it sounded like something to put on their chips - but nonetheless it was user-friendly and relatively quick at generating reports and inputting data. Users in constituency offices were just relieved that, being on a single server accessible via the web, it could be used outside of a specific computer in their constituency offices.

Useful improvements included an Outlook-style calendar where constituency organisers could arrange mailings, surveys and door-to-door canvassing as events in a current-month view (linked to the data selections for each activity), as well as the replacement of MERLIN's MOSAIC codes with a cleaner corpus of 8 "Target Voter Types" for canvassing outside of already-pledged support. Overall, a sense of having made a forward step prevailed.

The Chinese calendar

The only snag was that by this stage (January/February 2015), Votesource had only been rolled out for use at Party HQ and in 80 marginal constituencies or "target seats", and substantial modules of functionality including membership management and the critical election-day "Get out the Vote" module – real-time tracking of pledged supporters' voting status - were still to be released.

When the time came to bring all 400 Conservative held or winnable seats onto Votesource (whilst smoothly releasing its new functionality in time for the election in May), its performance sadly proved to be as languid as an anteater's walk and as buggy as its diet. Users rang into the helpdesk to describe it as "hellishly show" and comment that the hitherto loathed MERLIN "shone like a beacon in a cesspit" in comparison. Constituents' voting intentions entered by volunteers and activists would take some days to update the numbers in reports, which gave organisers in marginal seats false panic about their MP not having enough support. Votesource took a special dislike to being asked by its users to run mailing-list selections with complex inclusion/exclusion criteria – something I secretly relished as it meant I could practice VLOOKUPs by exporting the inclusions and exclusions in two Excel files and checking for electoral roll numbers that appeared in both. Owing to this, nobody expressed gratitude that Votesource did at least exclude dead people from selections by default, which was an improvement upon MERLIN.

Votesource's software rollout nadir came in early March 2015, when a half-tested new release led to constituency organisers and IT helpdeskers alike logging in to witness Votesource's calendar showing in Mandarin. I remember my office neighbour suppressing a snigger as I responded to a call with "Yes we are aware that Votesource is in Chinese this morning, and are taking steps to resolve it as quickly as possible", and from then on until election day, the severity of all and any software or data screw ups to come would be benchmarked against the Chinese calendar.

Edward Stratton Nedbert

One that came close was the occasion when Votesource data was used for a direct mail piece in a few constituencies including my own, something which I found out when I discovered a leaflet in my letterbox inviting me to vote for my local Conservative candidate in south London. It was addressed to "Edward Stratton Nedbert". I could trace this back to Votesource since Edward Stratton was my full name and "Nedbert" was what I had put into the Nickname field of my own record while doing some testing. I took the leaflet into Party HQ to flag up the possibility that a few hundred thousand leaflets had gone out to constituents with their nicknames on them, which prompted an urgent investigation into other nicknames that constituency organisers and volunteers had given to people they'd met canvassing on doorsteps, some of which were on the uncharitable side. Whether this was explicitly Votesource's fault or that of a HQ staffer working with the data I never managed to find out, but it underscored how the database was at the mercy of its user-base's idiosyncrasies as much as MERLIN was.

Lessons learned

Votesource's speed did gradually improve as the May 2015 election approached, but inevitably its tired server couldn't cope with the usage demands of the big day itself. For the much of the time that polling stations were open from 7am to 10pm on 7th May, Votesource ran a degraded service and activists could not get the real-time data they needed about which of their pledged or potential supporters had or hadn't voted. A phone call I fielded from a deflated local organiser 3 hours before polls closed told the story. He reported to me that he was concerned about low turnout in a Tory-supporting ward in his marginal seat, but that he simply couldn't tell if it was one of three things: "Either people aren't voting for us there, Votesource has lost the data or crashed and isn't updating, or the volunteer I have manning the office over there isn't putting in the data correctly, and I know how monumentally thick she is."

But against the odds, the Conservatives won the election in May 2015 with a majority, and Votesource is still their campaigning database. My time there from October 2014 to May 2015 provided me with the foundational knowledge of databases, their underlying technologies, and considerations that need to be made to ensure their success. This set me up for work as a database manager and data analyst, even if it did so via a constant stream of "what could go wrong here?" anecdotes.

I also learned the following:

  1. Stack Overflow isn't just the name of a popular support and Q & A website for developers, it's what happens when a computer can't allocate enough memory to a specific task. (This was the error message given to users when MERLIN crashed).
  2. Problems with software rollout and data need to be catastrophically severe to have visibility at the very top. The angst and chaos that my team and I experienced while we were supporting Votesource would make us dread coming to work, but a mere two weeks after the Chinese calendar, David Cameron and George Osborne came into Party HQ to thank the whole office and say that it had been the most slickly run and professional campaign they had ever worked on.
  3. Finally, Conservative activist database end-users make up for their computer illiteracy with creativity and resourcefulness. One elderly volunteer was struggling to login to their MERLIN PC, so the helpdesk team asked them for a screenshot of the login window. Not knowing how to take a screenshot, they sent in the hand-drawn screen below instead which, if anything, over-exaggerated the polishedness and aesthetic quality of the MERLIN interface.

MERLIN, Myself, I - an artistic approach to screenshots for IT support


Making up the numbers (Rogue Data Analysts) (02/12/2022) ⏪ ⏩ Data in Politics Part 1 - MERLIN (02/09/2022)

⌚ Back to Latest Post