How to measure brand activity in a performance marketing world


Brand and performance marketers can argue around the houses of attributed sales vs uplifts in sales.

Let’s consider why!

People used to shop in the high street, and see a range of shops. Brand marketing would then increase the chances of you buying that product rather than another. Great example, Coke or Nike!

Those brands would not have direct access to their customers details, they would only really see the overall sales volume in a given area; maybe some surveys etc or maybe some insights from their retailers; but largely speaking they just see their marketing spend and increase in sales.

Performance marketers, by that we mean digital, have no high street. One way or another people need to find their store/website, and given digital tracking we can see where/how people do this.

As more channels have come online, we have the challenge of knowing that people have complex journeys; so, we build attribution models to try and split that up. The challenge always remains though, that if we didn’t have 1 channel (or spent less) would that customer who used it have purchased anyway?

Over the years there have been more approaches to measure brand spend than ever before:

  1. Marketing Mix Modelling
  2. Structured media tests
  3. Brand tracking studies
  4. TV spot time analysis

The challenge with all of these is that they are often run and viewed independently & not aligned with the performance measurement tools.  So, lets consider how we can more closely align these two worlds.

Marketing Mix Modelling (MMM)

MMM looks at the spend in all of your channels over a period of time, and the overall level of sales. It tries to find and define relationships between the spend & the sales.

The challenges are:

  1. If you are running small or localised tests MMM cannot see this
  2. It is difficult to see trends at smaller segment levels
  3. If there are no changes to budgets, it will struggle to explain the “base line” and may miss some value

A good econometrician (who would build these models) can adjust their model to reflect other drivers that you may have or know. For example, if you did a structured test on media, you may find a better base line and bake that in.

This means that we can also be more creative, and start to consider passing more complex patterns for the model to explain.

For example, you could provide the overall number of specific customer journeys that use 1 vs multiple touch points into the model. I.e. how many single visits happened via PPC vs a visit and retargeted return?

We can then feed these into the time series model to measure the impact on sales as we increase the single vs join activity. This can then give us a ppc strategy of focusing on net new traffic or existing.

Structured Media Tests

Structured media tests are a great way to validate that a given channel is working. In many cases larger organisations who give up on MMM will revert to zero based budgeting using tests.

The idea is fairly simple, you turn off your media and turn it on again and measure the uplift. Or use a control group (like a control postcode).

Structured tests are not so common in performance marketing, as it is the equivalent to having road works outside of your store – it physically blocks people from coming.

However, we can use brand media tests to better understand the relationship of brand and performance marketing.

For example, one hypothesis is that brand advertising makes a consumer more likely to choose you when they are in market.

If this is the case, when brand advertising is on you may find that you get more engagement from position 2-3 in PPC search campaigns, than when it is off. I.e. although you are at a lower position people may choose you if your brand is stronger.

When conducting structured media tests for brand, it is worth trying to list & consider all of the behaviours you think that brand supports; and then find measurements that performance marketing can help solve.

Brand Tracking Studies

Brand tracking studies are research panels that ask questions around the recall of adverts and brands, and the consideration that consumers have for using them.

In the absence of being able to directly track AD exposure, they enable us to measure the overall mood or impact of the market; as well as indications of the overall market size.

Like in the example above, where people are often offered a range of brands to choose from (in search engines or other places), there are many places where improving consideration is the main goal.

If you look at almost any digital account, there will be peaks and troughs throughout the year; seasonality will be one key factor. However, there will also be times when performance changes and we don’t really know why.

Brand tracking can provide that layer of context to explain ‘why’ performance measurement has changed. Once we understand the drivers we can decide on resolutions.

TV Spot time analysis

An off the shelf spot time analysis report can be fairly cost effective. It is usually a tracking code you put on the website which looks at the uplift in visits when a TV ad is on. For one brand we work with they knew they would get 50k people visit their site on the XFactor advert break.

However, these products usually have limited tracking and won’t tell you:

  1. How many new customers there are
  2. How many people buy
  3. How many people come back and buy again
  4. How people are getting to the site
  5. How these visits work in the overall marketing or attributed mix

By investing in enhanced analytical products, we can easily answer all of these questions. For one of our clients, for example, we noted that in TV spot times:

  • People used specific search terms that entered to the site on unexpected changes (which led to SEO optimisation)
  • The ad would trigger people to engage with CRM comms that they have seen
  • Most people wouldn’t buy within the spot time, but would then comeback at a later date

For most mature brands, they are likely to be using a number of these brand tracking approaches already; but they will still suffer from:

  1. How to action them
  2. Combining the insight from one to the other
  3. Having a range of questions that are still not answered

When looking to solve these challenges, it is usually not a totally new method or approach that you want or need, instead you need to think of subtle ways you can align these approaches with your other teams and existing approaches.

Impact of AI


AI has the potential to process vast quantities of information, whilst also overlaying additional context.

For many of these approaches, aligning brand and performance will require better access to data and ensuring the fundamentals of tracking are working.

 If you have a long-term ambition to use AI to help with advanced modelling, then you need to ensure that you invest in your data infrastructure today so that you have the right level of detail for the future. Otherwise AI will try to infill results with assumptions – which can quickly go wrong.