How to measure the ‘messy middle’

By John Clarvis, Data and Insights Director as originally featured in Performance Marketing World

The marketing funnel is a longstanding model of how customers pass from awareness to purchase. But within this funnel lies an uncomfortable truth – no-one knows what’s happening in the middle.

The difficulty with measuring mid-funnel is that by definition, it’s a stage in which the consumer isn’t really interacting with the brand. So how do we measure what we don’t know? We must briefly stop thinking like marketeers and start thinking like psychologists.

The psychological underpinnings of long-term measurement, for example brand health, are reliant on memory. Can people recall a brand? If they can, how do they feel about it? This is a cognitive line of questioning.

We can test this at the top of the funnel with brand trackers and awareness polls, allowing us to access memory via prompts.

At the bottom of the funnel, we have consumer behaviour. What did they click on and when and why did they do it? Digital data allows us to zoom in on consumer behaviour in excruciating detail and optimise for moments of direct interaction with brands. Conversion rate optimisation (CRO) and attribution disciplines offer glimpses into moments of customer interaction with a brand.

For planners however, this data offers little value as it’s out of context. We may know that in A/B tests one advert outperforms another, but this behavioural snippet can’t be taken outside of its specific environment.

So far, we have one set of memory measures and one set of behavioural experiments. In practice, this means that we have (hopefully) embedded a brand into memory, and that at some point, this will contribute to performing a behaviour in our favour, like a purchase from our brand. There is an underlying assumption in marketing that what happens between these cognitive and behavioural stages is linear. But is it?

The average human has around 6000 thoughts per day, bouncing from thought-to-thought every ten seconds, in non-linear patterns. Our attention span is around 8.5 seconds, while on average we face a distraction every 8 minutes. In short, any kind of linear long-term process would be wrecked within a day.

The sales funnel exists as a handy schematic to understand a marketing process but serves little as a measurement framework. To build a useful sales funnel metric for the middle, marketers must work with the assumption that this stage is non-linear and that brands are competing for space against thoughts, distractions and minutiae of everyday life.

Logically, the only avenue left for measurement is to try and understand what consumers are doing in this stage and what we can measure. In a strictly marketing sense, when people are aware of a brand but not ready to purchase, they can only be either not thinking about the brand at all, or directly weighing up the brand before they commit to a behaviour.

We cannot constantly track people’s thoughts (nor should we aspire to), and we cannot ask them directly without interfering with the process and setting us back to a memory test. The most fertile area to explore for understanding these thoughts is a hybrid approach of cognitive and behavioural measurement.

Searching for answers

A measurement point that is readily accessible and the most frequently used service for information gathering is search. When people want more information, they search for it. Consumers are behaviourally committing an action to cognitively process information.

Over the course of five years I set about gathering data on searches by collecting at random the frequency of brand search terms and their related searches. With a series of exclusions and nested subqueries, I have built a relational model of searches related to brands.

The resulting dataset was both enormous and text based, requiring data reduction and classification to be understandable. The key factors were extracted to classify searches in terms of what the specific aim of the search was and by staging them in order of sequence, a multidimensional map of the mid-funnel was developed.

The output resembles a map of orbits, escape velocities and diversions, giving the product its name: Orbits.

Classification of the principal components revealed five measurable stages in the mid-funnel:

1. Gathering information – consumers are broadly interested in a category. At this stage searches remain short and vague. They can become bogged down and may require some need state to arise to progress.

2. Researching – searches become more specific and focussed. Frequently brands enter the searches but not always. Strong brands will frequently accelerate searches to finalise while weak brands will often send searches backwards.

3. Comparing – when a longlist of brands and specific features are compared. The second longest stage is particularly lengthy in categories with many brands. This frequently cycles back and forth between prior stages.

4. Shortlisting – when this longlist is shortened to provide specific details. While it does occur frequently enough to be statistically significant, it occurs only on searches for expensive items. Searches focus on very specific features and rarely progresses to finalising, most frequently sending users back to research.

5. Finalising/deal-hunting – a decision has been made, time to purchase. The most price sensitive stage. When prices or availability are unfavourable, searches tend to revert to comparison for a second-best option. By applying a hybrid approach of cognitive and behavioural metrics and psychological research methods, Orbits has been developed as a mid-funnel measurement framework to measure the previously unmeasurable.