Why Segmentation Matters
Something we all know – some from hard experience – is that the wrong message to the wrong person is sometimes saying nothing.
A lot of what I’ve discussed online has been very much about the message – achievements, leaderboards etc. Marketing executives are often cited as saying “I know that I’m wasting 50% of my budget, but I don’t know which 50%” – Lets see if we can at least understand where we are going wrong.
Introducing the 3R’s
After being so negative, let us look at how we can score segmentation techniques:
Right Person: Does our segmentation technique successfully target those most likely to act on the message?
Right Time: Does our segmentation technique help with understanding the timing of when to deliver a message to the customer?
Right Message: Does our segmentation technique help with understanding the levels of segmentation
Level -1: Broadcast
We (almost) all grew up with this on television and radio – yes, we ended up all singing along to the jingles, repeating the slogans and laughing on cue at the bad jokes, but how effective was it as a segmentation technique?
Right Person: There is no control about who sees a message when it is merely, broadcast out to the world – while media sellers will tell you about the demographics they anticipate, it is a guess. (Score: 0/5)
Right Time: With no information about the individual receiving the message you have, or at best, broad-stroke information, e.g. its 7 am so its probably breakfast time. Probably. Ignore that they might be already on their way to work. And won’t be paid for another week. (Score: 0/5)
Right Message: The ability to deliver a message really suffers with broadcast segmentation since it needs to not offend at the same time as informing people. Its why kids need to be able to repeat car slogans long before they will ever decide to buy one. (Score: 0/5)
Examples: Broadcast Television, Commercial Radio
Overall Score: 0/15
Level 0: Historical Profiles
This is the level most executives will associate with segmentation – a break down of their customers based on historical behaviours, e.g. last purchase time, gender, geolocation and a lot of implied information (socio-economic bracket, likes/dislikes based on age). The info breaks down into several potential target cohorts (groups) typically up to 12 – the limiting factor is generally the ability for people in an organisation to market to the groups and their likes/dislikes. Crucially it is based on behaviours that were true in the past – typically a month ago, a year ago or even the last census!
Right Person: Often organisations report impressive jumps from even quite basic historical profiling – this is unsurprising even with inaccurate profiling with four well-presented campaigns you have a 25% probability of accidentally targeting the right person with the right campaign. With “good” results being anything from 3% plus, there’s often an unacknowledged amount of luck. (Score: 1/5)
Right Time: With coarse groups, historical profiling doesn’t necessarily help with the timing of messaging unless one of the criteria is specifically about time. (Score: 0/5)
Right Message: Historical profiling is often associated with personas which definitely helps marketers customise the message. Given that historical profiling is often mutually exclusive (a person fits in one segment but not another) the right message might still get to the wrong individual (Score: 1/5)
Examples: Targeted letter drops, specialised magazines
Overall Score: 2/15
Level 1: Micro Profiling
Micro-segmentation takes information that is available in historical profiles and expands out the number of categories that it is broken into – into the hundreds. Prominent examples of this sort of segmentation include Facebook and Google whose tools allow you to craft messages against specific intersections of behaviour to enable targeting of very specific messages.
Right Person: Undoubtably micro profiling allows you to identify the right person more accurately, but it is inherently based on historical behaviours. This might result in over-targeting or overexposure, as it won’t take into account changes in behaviour until the next reprofiling exercise, or sometimes ever! (Score: 3/5)
Right Time: As with all historical profiling, micro-segmentation techniques don’t necessarily help with getting the timing right for a message. Micro-segmentation doesn’t have current information (e.g. you’ve just ordered pizza, you don’t need another Pizza Hut ad) (Score: 1/5)
Right Message: Messaging is still primarily coarse-grained with micro profiling – specific messages for specific audiences without sufficient information to change the content of the message (Score: 2/5)
Examples: Classic Facebook and Google ads, Point of Sale Supermarket Vouchers
Overall Score: (7/15)
Level 2: Real-Time Profiling
We’ve all seen real-time segmentation at work on the Internet – browse for a vacuum cleaner, and suddenly you are confronted with a bunch of ads in the correct currency, with the distance to your nearest retailer. Its an amazingly sophisticated trick that seemingly only the most prominent Internet companies can pull off – Google, Amazon, Facebook – on specific (usually their own) content.
Right Person: Although the visual impact of real-time profiling can be dramatic, it often doesn’t take into account when the right person becomes the wrong person (Score: 3.5/5)
Right Time: Undoubtably the timing of messages is dramatically improved – the reason for not receiving a higher score is that real-time profiling usually doesn’t take into account the customer journey – research, shortlisting, purchase, etc. (Score: 3/5)
Right Message: When we researched this category, we were surprised at the lack of customisation of the message – yes, things like price and availability were included but very little customer-specific information (Score: 3/5)
Examples: Amazon in-store purchases
Overall Score: (9.5/15)
Level 3: Real-Time Behavioural Profiling
The ability to change the segmentation of a customer (and messaging, and timing of messaging) must feel like science fiction for many businesses – and yes, the best examples of customising journeys, messaging and timing of messages are best showcased in games. Quests, tournaments and challenges adapt to player progress, statistics and capabilities. A less emphatic example is for ride-hailing companies who have embedded within mapping tools – insight into your behaviour allows them to modify their messaging and pricing according to not just your behaviours but those of the marketplace.
Right Person: An element of self-selection based on behaviours nails picking the right person for messaging – we would note that even companies with real-time behavioural profiling still default to broadcast messages in the case of big promotions. (Score: 4/5)
Right Time: Because gates trigger messages, the timing of messaging is often pinpoint accurate and not excessive – since the real-time behaviour knows that an action has been taken, a message can be withdrawn and the next selected (Score: 4/5)
Right Message: Messaging is still largely pre-defined in this category – pricing for ride-hailing is predetermined (even if algorithmic), and games messaging tends to follow specific storylines without varying incentives etc. (Score: 4/5)
Examples: Online games, Ride-Hailing companies
Overall Score: (12/15)
Level 4: Guided outcomes (Machine learning)
Strap in because this is the ultimate level of segmentation – and nobody is there just yet.
Rather than setting messages and segment, the business person will set the desired outcome or destination that they would like their customer base to get to, or move towards, for example, the adoption of a new feature.
At this level, we would see the customisation of whom to target, when to target, and what to target with, all based on machine learning.
Right Person: Segmentation at this level will ultimately be a segment of 1 – an individual. It’ll be hard to get better than that (Score: 4.5/5)
Right Time: (Score: 4.5/5)
Right Message: Messaging is still largely pre-defined in this category – pricing for ride-hailing is predetermined (even if algorithmic) and games messaging tends to follow specific storylines (Score: 4.5/5)
Examples: TBC
Overall Score: (13.5/15)
Takeaways
As we can see from the different levels, data about customers is crucial to moving to a higher precision of segmentation. It is easy to focus on the quantity of data, but what is also essential is the timeliness of data – the closer to real-time, the better your ability to:
- Send the Right Message
- At the Right Time
- To the Right Customer