Paddy Power’s fake kit promotion – evaluated

I love the site PR Examples, it’s a great source of some of the most creative campaigns undertaken and its round ups can give great ideas. But do these creative ideas work?

Over the coming months we’ll be running a semi-regular feature that looks at the effect of some of the campaigns it highlights to determine the effect it has on the business. Some will (I expect) have an amazing, demonstrable effect. Some will have a negligible effect.

First up is Paddy Power’s fake-football shirt campaign, listed as the top campaign in PR Example’s best of June / July roundup. It had declared aims of highlighting the number of gambling firms sponsoring Premier League teams. But we’ve also (and principally / cynically) measured the campaign against financial / salience metrics.

Copyright – PR

So, what are the effects of these campaigns for the company? 

The PR campaign

The campaign has so far taken four phases (see graph below): announcing sponsorship (15th July); unveiling the shirt – with rumours it might be a stunt (17th), confirming it is a fake shirt (19th July); and continuation – initially via a charity auction (22nd) but also mocking of other kits and a TV campaign.

Paddy Power shirt coverage graph during the campaign – please excuse use of corporate colours – becomes useful when competitor is introduced.

From a coverage level, the campaign has been spectacular – and the people tasked with maximising coverage for the announcement should be commended.

Meltwater picked up nearly 270 items of coverage of Huddersfield sponsorship and the campaign. For a comparison, this is 11 times the level of coverage Meltwater recorded for 32Red’s sponsorship of Derby County in May 2018 (just 25).

PR Week has understandably described it as a “masterclass in PR” and the campaign has also been covered positively in most of the marketing press (see the Drum).

Paddy Power has said its was “blown away” with the earned coverage – and I’d agree with that sentiment.

The campaign got so much more coverage than similar announcements … and it was only the fake-shirt (or at least ugly shirt) element that allowed it to get national coverage (eg in the BBC, the Indy, the Mail).

In short – the people behind it created a campaign that got amazing coverage for what would, otherwise, have been a very small story and should feel rightly proud. But, was it worth the effort?

Evaluation of the PR – part 1, did it meet its stated goal?

Obviously, coverage is only an intermediate goal – with the aim of the coverage being to affect a change in behaviour (See AMEC’s Barcelona Principles).

So, looking first at their quoted goal (no pun intended). There has, during the campaign, been a spike in daily searches for football sponsors – including a 172% (2.5 standard deviations*) increase on the mean daily searches on the 19th, the day Paddy Power announced it was a fake shirt. Coupled with this was a 132% (just 1.9 standard deviations) increase on the 17th, when rumours of the fake shirt circled. 

But the Google Trends graph** and Google Ads suggest the numbers searching are low (Google Ads estimates 590 per month). So this represents a small increase in salience – adding c.60 searches (34 on the 19th, 26 on the 15th). 

We also looked at terms such as football sponsorship (similar pattern), and football sponsor gambling / football sponsor betting / sponsor gambling (all having not enough data for Google Trends to provide a graph.

Conclusion – it did not meet its stated goal.

Evaluation of PR – part 2, did it have a return on investment?

Here comes the cynic in me. Let’s assume they need to see some form of return on investment for the sponsorship and this campaign isn’t just because they think there are too many logos on shirts / gambling companies… a reasonable assumption given:

  • it’s a gambling company sponsoring a football team, and 
  • it has been accused of trying to cheat rules relating to the highly addictive / criticised fixed-odds betting terminals) what was the actual effect of the campaign?

From a coverage perspective, taking this approach definitely caused more coverage vs the 32Red / Derby County announcement (see above).

The coverage generated also got shared, with 165 of the 270 articles being shared on Twitter / Facebook, with a combined 253,640 shares across those two platforms. But coverage and shares don’t always equate to money.

I’ve therefore used share price and searches as evaluation metrics.

Share price:

Share price is one of the most instantaneous measures that can be tracked and correlated with PR. It only (obviously) works for limited companies, but is a valuable metric to show the value of a campaign.

Here there is a definite spike on the day the news broke that Paddy Power were sponsoring Huddersfield Town, rising from 6726 GBX on the 14th to 7010 GBX (+4%) on the 15th when the sponsorship was announced. Coincidentally, this is the highest share price this year.

Clearly the news of the sponsorship coincides with a spike, and arguably had an effect … but, did the fake-kit news have an effect? I would argue it didn’t.

Coverage levels on the 15th appear to unaffected be news of the sponsorship, being purely factual on the 15th. As far as I can see there was no mention of anything suggesting the kit was ugly. There was no national news that Meltwater picked up. And it was almost exclusively local / trade press that covered it, none of the articles (that Meltwater picked up, at least) used anything with the fake shirt.

The Huddersfield Examiner used the Paddy Power logo. The Yorkshire Post  used a photoshop of the logo with Huddersfield Town’s crest. SBC (gambling press) used a photo of the team in the old OPE-sponsored kit. Casino Beats used a Chelsea-linked image. Even Paddy Power’s release used a photo with the OPE shirt.

And, coverage levels were on a par (on the day of the sponsorship announcement) with those of Derby County / 32Red – even with it being both the largest sponsorship deal (£1.5m per year) outside of the Premier League, and it being (reportedly) the first time an Irish bookmaker has sponsored an English club.

The big spikes in coverage instead came on the 17th and 19th, where the share price was remaining roughly static (dropping slightly on the 17th, and up again by 1.5% on the 19th).

It’s also worth noting that this 4% (while earning more than I make in several years) jump on the 15th followed a much larger spike in late June, which took the share price from 5696 GBX to 6890 GBX (21%) over 9 days and also coincided with a notification of acquisition / disposal of voting rights. Similar spikes in share price for Paddy Power typically last two weeks and then fall – and this spike has following that pattern.

Conclusion – the fake shirt campaign (probably) had no effect on share price.

Searches

If not share price, how about on customers and revenue? 

Paddy Power operates both online and through a little over 100 stores in the UK. According to the FT, Paddy Power had a Q1 UK revenue of £152 million (£11.7m per week – albeit this includes a skew from the Grand National). Google Ads suggests that 1.22m (0.28m per week) searches are done in an average month for the term Paddy Power.

Not all of its business is done online (I couldn’t easily find a breakdown of web vs bricks and mortar transactions). Also, only 26% of its web traffic comes from searches. So search patterns aren’t perfect, but they tend to give a reasonable approximation of changes in visits to a company’s websites / stores. I will highlight that I have made a (decent-sized) assumption that an increase in Google searches correlates with an increase in store visits / bets placed.

I’ll also state that, obviously we cannot see data for the amount spent per customer and if this increases.

From this we can however see that Google Trends data has most web searches taking place on the weekend (when most bets / sporting events happen) so it’s a fairly (#caveat needed) safe to assume the correlation of search / visits is correct in this case.

News of the sponsorship itself appears to have had little to no effect, with traffic on the 15th down by 9.3% vs the average Monday. 

But on the 17th (once rumours the kit may be breaking rules / be false spread), searches were up 15.8% vs the average Wednesday (c. 5700 more searches). And was 14.1% up vs the average Friday on the 19th (c. 6000 more searches).

But by Saturday the 20th (when most searches, sporting events, and bets placed happen) these searches had dropped – with searches on the 20th 12.7% down vs the average Saturday (c. 7100 fewer searches).

Still, over the 7-day period from the 15th (to include the weekend) there was a net increase of 1.0% (c. 2800 searches). 

Putting a financial figure on this 1% increase in searches would be hard. We don’t have any data to go off other than the quarterly revenue figures.

If (and this is an unbelievably big IF) searches increase correlate precisely with revenue (both from online and in-store bets), based on the Q1 results (which included a Grand National skew, so isn’t a true representation) this would have added c. £115k.

Again, this figure will be more than the cost of the campaign. But this figure is also likely to be a spectacular over-estimation, especially given the drop in searches on the Saturday, when most bets are made. And it is also pocket change to an organisation pulling in £11.7 million per week.

How significant were these increases?

By looking at the standard deviations from the mean for each day of the campaign – measured against the mean for the specific day of the week, we see that, despite the high level of coverage, the PR didn’t create a significantly greater salience for Paddy Power.

As the graph shows, the spikes that were seen are well-within normal ranges**. The 12.4% increase on the 19th is just 1.04 standard deviations (which takes into consideration the range of spikes) from the mean. This is not considered a significant jump, with it having an almost a 1/4 chance of it being due to random chance.

That said, this doesn’t factor out the external events that would create spikes – the one in mid June is likely to have been related to Queens / Ascot. And the one in late May / early June to the Champions League. No sporting event took place during the week of the campaign /the weekend after it (albeit England had just won the Cricket World Cup, possibly leading to bets on England in the upcoming Ashes series).  

Eliminating the effect of these sporting events?

Another way of analysing these spikes is via comparison with its rival, Ladbrokes. According to Google Ads, Ladbrokes has roughly the same number of monthly searches (1.22m) and its Google Trends graphs are nearly identical.

And here, the data suggest the campaign had a small effect during the week, but nothing remarkable. In fact, for the days after the launch and the news that the shirt was a fake, Ladbrokes searches spiked too (bets on England for the Ashes?) suggesting Paddy Power’s spike was not completely associated with the sponsorship news).

So was it successful.

Based on similar sponsorship deals (Championship club / betting company) it did give significantly more coverage than the simple / standard news announcement.

And the PR agency / in-house team without-doubt did a spectacular job in both coming up with an idea that gets coverage / implementing it. The campaign got 11 times the coverage of 32Red / Derby County’s announcement, so serious kudos are needed for the team/s.

But, coverage is only an intermediate goal, and the real question was did it affect a change in behaviour – be it revenue, investment or (if I remove my cynicism) the company’s stated aim of removing logos from shirts?

I think it was, as PR Week put it, a “masterclass”. But in coverage generation, and sadly (unexpectedly) this doesn’t appear to have translated to share price / search gains. 

I’d obviously like to see the financial data, and the PR team should be rightly proud (they were, after all, tasked with getting coverage for a minor story) but I think it may be wiser to question the true value of shirt sponsorship, or at least for a team outside the premiership.

Can anything be learned?

I work with companies and agencies to help them show off what has worked well against objectives and make adjustments when things haven’t. This helps them be more efficient – putting forward strategies that include more of what works, and less of what doesn’t.

So what can be learned from the campaign – both to be mimicked in the future, and to be improved upon?

Firstly, I love the concept and the execution of this. I’ve said it several times in this blog but the PR team involved did a spectacular job of getting coverage (plus in-bound links) for what would otherwise be a dull, minor story. 

Commenting on this story in Campaign Live, Mark Lloyd at Dark Horses said that he applauded Paddy Power for its bold thinking but “in today’s game, shirt sponsorship feels faceless and corporate, the brands completely interchangeable [and instead] clubs, brands and governing bodies to create a plan to make the shirt sponsorship iconic again.”

In short, shirt sponsorship may be a poor return on investment (especially outside the Premiership) and there was nothing else Paddy Power could have done better.

I suspect this may be close… but one minor improvement seems obvious. The day of announcements should, to me, have matched the key revenue-generating day (ie Saturday) for Paddy Power.

The team announced the fake shirt on the Wednesday, confirming it in the Friday press. By shifting it just one day (to Thursday / Saturday), or one / two weeks weeks (to when biggest sports events were on), i feel, would have helped salience when people were more likely to make a spontaneous bet.

Footnotes:

*A change is considered significant if it has a less than 5% chance of happening randomly.

In a normally distributed population, 75% of the population will fall within one standard deviation from the mean, 95% will fall within two, 99% will fall within three. 

This allows us to say that the chance of anything lower than 1/ 2/ 3 SDs have a greater than 25% / 5% / 1% chance respectively of happening by random accident.

**Because Google Trends gives a percent of the max (and only gives whole numbers) I’ve looked at traffic from the weekend after the Grand National to show more variation.