4 min read
14 Mar
14Mar

Introduction


I’m slow at writing up client work and make no secret that I’d rather be doing more good work instead. This post has only come to fruition because I contracted covid and so I’m enduring a period of effective house arrest. 

What you won’t see here or in the short-form are lots of numbers that you can use as benchmarks. Here’s why: 


  • Like all decent data modelling providers, I take client confidentiality very seriously. Imagine how you’d react if a partner you worked with shared your business return on investment and budget history for anyone to find


  • After many years sensitivity wanes. My business is young and we’re not there. Besides, especially right now, the wider business context is key. Old cases can be less relevant unless chosen very carefully to match circumstances


  • There’s a whole chunk of my project experience that’s not technically mine to share, so I don’t. Eighteen years’ worth of corporate modelling projects belong to the companies I was employed by, not me


  • With large numbers of cases, you get emergent patterns that make benchmarking useful. These exist for certain sizes and types of brands, but smaller businesses often have a very different profile on many key metrics


There’s a limit to what can be learned about what action to take next simply by studying what’s happened to those who have already trodden that path. As you probably communicate to your customers, you are unique.

 

Businesses behind the cases


If you’re after the TL; DR bullets click here. For deeper context:

Both businesses fall into the UK government SME definition but are quite different sizes by turnover and employees. While each predominantly provides B2B services they both sell to individual consumers too.

They each provide something that their customers would consider necessary rather than discretionary. That influences the length of the purchase cycle and their pricing models. Predominantly ecommerce now, both adopted PPC ads in the early days.

Both are UK registered companies with UK as the flagship market, but they each have a global presence too, predominantly in Europe. They have competition. Within their respective competitive sets, one of them is much bigger than the other.

They innovate. The core of their services doesn’t change but they grow by keeping pace with the industry trends to roll out the same service to more groups of people in their marketplaces. Covid forced each to change their delivery system for a time.

The CEOs of each business approached me after seeing me talk about marketing mix modelling on LinkedIn. They are both all over their performance but had not managed internally to disentangle the impacts of their marketing from other drivers.  

Scope for each business


Marketing mix modelling is the gold standard approach for measurement of marketing return on investment especially if you do more than just digital marketing. And even if you don’t, MMM is not reliant on cookie tracking. Just saying. 

Because each business spent a large proportion of their marketing budget on PPC, two KPIs were modelled for each. Website visits, to capture PPC (and other drivers) ability to fill the top of the purchase funnel. Another measure to reflect conversion. 

These models can stand alone but are even more powerful if the link between visits and conversion is quantified as part of the modelling. As well as understanding the role of marketing activities at different stages of the funnel you get an overall view. 

Modelling the UK was a no-brainer because it drove the largest share of global revenue for each brand. Each business agreed with my recommendation not to try to apply findings from the UK to other countries. 

There are usually too many structural differences for this to be accurate, even if you try to adjust the outputs first. One company added a second country for strategic reasons, to give them another angle on an important long-term decision. 

The other company have a very ambitious roadmap for marketing investment decision-making to eventually manage in-house. We settled on a representative country from each core region to lay the foundations for that. 

You should understand if you’ve got sufficient good quality data on all the factors that drive sales before you sign an engagement. That’s why each project began with a feasibility study. Read more about why these are a wise first step

I don’t then disappear for a few weeks and come back with my singular view of your world unless you really want it that way and accept the risks. The most useful data, most actionable findings and arresting takeaways only come from collaboration. 


Findings and recommendations


Because of the way it works, marketing mix modelling will often give you answers to questions that you didn’t even ask. The depth that you can get depends on what’s happened during the modelling period and how you have scoped the models. 

As a minimum you should get a quantification of how much each driver explains of each metric that is modelled. All the detailed stuff derives from these building blocks so make sure that you understand them, so you have faith in the figures. 

For paid marketing, standard outputs are return on investment (at revenue or profit depending on whether you share financials.) These can be cut over different time periods. Differences over time or campaign can be tested. 

Be wary of making decisions based solely on ROI without understanding the broader context. Was it only high because you spent a tiny amount? Was it only low because that execution included was poor? Or it was run at an expensive time? 

If you have some variation in spend levels for each channel it will also be possible to measure carryover effect and diminishing returns, two more handy metrics that will help with decisions. You can read more here about MMM outputs

Covid was an externality that needed accounting for even though neither business had many specific questions about it. (They had traded through it, adapted and were surviving so knew in detail what the top and bottom-line impacts were.) 

Both firms pulled back on paid marketing spend in the early days of the pandemic but neither slashed their budgets to nothing. There was no evidence that marketing during lockdowns was less efficient, or that the efficiency changed much over time. 

But the company in the more cluttered industry saw some competitors ramping up PPC spend since the start of the pandemic. They became more vulnerable over time to what their competitors were doing. 

There’s a lot of “what” findings that come from marketing mix modelling. But it’s adding the “why” to the what that generates insights. Even better, articulating the “so what” as a business recommendation – sometimes this becomes clearer later. 


This happened …
Because of these things …
Which suggests that you should …


Wrapping it up


While case studies are useful, it’s also risky to only be making decisions based on someone else’s business results. It’s also tricky to benchmark from one or two specific case studies, so I have tried instead to focus on what the possibilities are. 

One thing that is clear is that while two business may both be classified generally as SMEs once you get into their individual contexts they are as different from one another as snowflakes. It’s then less of a surprise to find that you can’t generalize.

 Despite that, the same analytical approach was used for each business, demonstrating how versatile marketing mix modelling is. But it works best when the scope is tailored to each business. A cookie cutter approach doesn’t, well, cut it. 

This post didn’t go into what marketing mix modelling is, but if you’re intrigued or need a refresher, try this out. It only takes a coffee break and there’s plenty more if you want to go into more depth. 

Don’t hesitate to drop me a line if you have any questions. It is one of my favourite conversation topics :) 


© Jo Gordon Consulting Ltd 2022

16Jun
Comments
* The email will not be published on the website.