I read Duncan Parfitt’s view and as interesting as it was, I believe the story is incomplete and focused on a small set of unique clients. As a data-centric marketer, I’d love to understand the methodology and see the data behind the article.

The reason search marketing (paid or not) is such an effective medium is that it’s demand based, where customers actively tell us they are interested in our services/products. Most other media, including radio, target audiences. You can’t tell me that someone sat in their car listening to the radio is more engaged than that same person searching for my product. As both a consumer and marketer I don’t buy into that.

Am I suggesting it’s not worth testing? Not at all, if the hypothesis is sound of course it’s worth testing. What I am suggesting is that other tests and other clients create a different picture.

Clients tested

Maybe the clients Duncan makes reference to are the select few this would work for: An airline in the middle of a global pandemic and a huge brand such as Dulux in a country that’s going renovation crazy and being fuelled by government grants. Chat to any decorator or renovator and they can tell you that all paint brands are doing well. Google trends shows us that interest in ‘paint’ grew during the last lock down and never receded.

Testing timeframes

With tests like this, the length of the test are often not long enough. With a strong brand, especially one with a virtual monopoly such as Airbnb, it would take months if not years to feel the effect of turning any advertising off. Duncan makes reference to B&D Group but looking at SEM Rush (a globally respected tool), it seems they are back spending on paid search.

In 2015 Indago ran two concurrent attribution tests for two large travel brands. One was relatively new with limited brand awareness, the other well established with a solid brand. Our findings was that with the new entrant it took over six months for their brand adverting to have a significant affect on sales but with the established brand, as soon as they were in market – clicks, conversion rates and sales all went up. When setting up a test, getting the length of test correct is critical.

Attribution

I know that search marketeers throughout advertising (Indie and agency groups alike) use sophisticated digital attribution modelling and generic paid search is often one of the most effective assisting channels. Knowing Ian Perrin I can’t imagine Speed is any different. The issue I believe they have had is attributing conversions that happen either offline or via a re-seller, but this issue would be the same across all their media. I’m also not sure how you can fairly compare an attributed sale from radio to paid search, could the methodology be the same?

For clients where the transaction happens online it is easy to understand how paid search helps drive a sale from an attributed perspective. A one click Google Analytics report (such as the one below) can help you start to understand attribution. Within GA there is a whole plethora of reports to help us understand the role our various channels play in a sale. Let’s remember that Google purchased the then market leading attribution tech, Adometry back in 2014 to help build the accuracy of attribution within GA.

Measurement

The quality of a search campaign is typically measured by one metric alone – performance. Hence 'performance marketing'. I’m not sure of another medium where this is the case. Display also has impressions, and TV TARPs. When conducting these types of tests we need to include the positive affect paid search has on a brand and the missed opportunity of not being there. If you sell garage doors, someone searches for an acquisition intent based keyword and can’t find you, they’re buying from your competitor.

Point of diminishing return

I couldn’t agree any more with Duncan's point on this. How often we see clients over-invest in paid search but as the medium is so transparent it is an easy thing to spot and rectify. (Not so easy when the sale happens offline or through a re-seller though).

If you do go down this route, I suggest you think about:

• How strong is your brand – for most of us we don’t have the brand and or budget to make this work.
• How reliant are you on paid search – it is such a sizeable part of your business that turning it off would be disastrous.
• Start small – test small facets of your paid search campaign and adapt based on learnings.
• Data reliability – are we in a seasonal or unprecedented period where external influences will cloud findings.  
• Measurement – make sure your attribution is solid and measure all your channels identically.

All that said, hats off to the team at Speed for running these aggressive tests – it’s exactly how your agency can find ways to drive great results.