Customer avatar research (but not the old way)

The days where marketers had to rely on pure assumptions to generate customer avatars for their buyer personas without reliable data were long gone. From fictions to facts, the world needs to move on to more productive (and free) tools.

[Author’s original digital art]

Introducing the Twitter Avatar 5-Series Tool

5 Simple Steps to Generate Avatars from Twitter Data

Whether you’re selling on e-commerce like Amazon or Etsy, whether you’re dropshipping or selling SaaS, or you’re into user-centric SEO designs — if your work relies on frequently researching the personas of new buyers or new users — this is the solution you rightfully deserve.

Developed on top of the industry-trusted, low-code, open source data analytics platform KNIME, the tool consists of 5 components representing the 5 steps to generate customer avatars using publicly available Twitter data.

I named it Twitter Avatar 5 — Version 1.0, or shortly, Twitter A5 v1.0. Here are the principles behind the novel tool:

• Worldwide-coverage: It leverages Twitter’s handy API calls to gain access to at least 100,000 tweets in each run (depending on your query). Not only is Twitter’s data publicly available, it also allows targeted customer research to be made in any locations in the world where Twitter usage is common and active, giving marketers the actionable insights they need from location analytics.
• Smarter-clustering: I believe in the power of entities. Entities are objects, events, or abstract concepts which are the center of other associated datapoints. An event could be ‘Taylor Swift concert’, ‘The US general election’ etc. An object could be any tangible item or product, such as ‘Air Jordan Shoes’, ‘outdoor ovens’, ‘gaming headset’ etc. Abstract concepts could be intangible constructs such as ‘mindfulness’, ‘mental health’, ‘grammar syntax’, ‘state of mind’, ‘level of emotions’ etc. Using entities to model the real world will help highlight the nature of relationships amongst things, which in turn, used to answer business questions. It’s therefore not surprising that the ‘Google God’ himself, leverages entity datapoints to model its Knowledge Graph. As for my Twitter A5 tool, I utilize the Twitter’s V2 API system with its own built-in entity recognition that’s able to extract entities from tweets, and use this info to cluster the Twitter users. The clustering is doable with KNIME’s Network Mining Extension, where I could score entities based on how authoritative they are within the network and keep in only those that function as hubs. This approach is built upon the belief that each person is represented by the entities of interest in their lives, and key entities from a particular population tell a lot about the members within that population — and hence which cluster they belong to. In comparison to other clustering algorithms, which assume the number of clusters beforehand, my tool uncovers the clusters naturally.
• Intuitive & Inclusive: This is what I love about KNIME, where people from non-tech backgrounds like me are able to use industry-standard tools without much fuss. With its easily comprehensible workspace, and being open source at its core, KNIME becomes the lowest common denominator uniting people from all professions looking for common solutions. If this is the first time you’ve heard of this free data analytics platform, I would recommend this Medium article, which to me is by far one of the most useful introductions to KNIME for new users.

So, how does my tool work?

Step 1 — Querying & Clustering

The first component of the series is where your queries to Twitter API are made. In order to access the data, you’ll need to obtain an Elevated Access developer account. This is mandatory, and I recommend every business to apply for one — yes, it’s free. Only one account per company is all that it takes.

With an Elevated Access, you’ll be able to extract up to 4,000,000 tweets monthly, and the rate limits for such privilege are not as restrictive as accounts without one.

This is how much of the monthly cap usage I have used when developing and testing the multiple prototype versions of the tool, before coming up with the final form of Version 1.0.

The detailed instructions will pop up once you’ve right-clicked on the component.

A screenshot of the pop-up window appearing upon right-clicking on the first component.

As you can see here, you’ll be prompted to provide the geo-coordinates file containing the target locations you’re interested to dig into, what tweet identifiers you’re querying about, as well as the radial distance covered from your provided coordinates.

Those crucial pieces of information allow precise research to be made, honing in on the right people at the right places. The rest of the information are simply the 5 different tokens you were provided with the first time your Twitter Elevated Access account was approved.

At the end of this article, I’ll provide a link to the KNIME workflow where you can access the tool. It contains a more comprehensive rubric related to the queries you’re about to make, the computer setting as well as security instructions.

Step 2 — Persona Profiling

Once the querying and clustering processes are complete, the next step allows you to profile the key personas identified in the previous step.

This is the chance for you to get to know each personas better, and to look at their tweets containing the key entities of interest.

Studying what the entities are within the tweets’ context will help you get a better glimpse of their lives and the associated entities. (Remember that the principle behind the tool is the belief that people are representatives of their entities of interest. That translates studying the individuals to studying what they are up to around those key entities.)

The detailed instructions will pop up once you’ve right-clicked on the component.

An example of what you’ll see upon right-clicking on the second component.

As you can see here, you’ll be shown the curated tweets containing the entities pertaining to each key persona. This is when you’ll go through a quick glance at all key entities identified as a result of the prior clustering during Step 1.

Once you’re familiar with the key entities, the note-taking exercise that follows later on becomes as smooth as a baby’s skin.

But before that, you’ll be answering a set of multiple-choice questions, using the Table of Reference shown above.

The first set or group of questions is Question 1, divided into 3 parts:

Each part represents the different psychological motivations and is mutually exclusive to its counterparts. The user will assign each persona to one of the 3 groups based on which one is dominant for that persona. The concept is based on academic research on neuromarketing.

Following that is the second set or group of questions pertaining to the choice of words made by each persona.

When it comes to understanding the person behind a tweet, the lingo of the tweets is really telling. Someone who’s highly educated, rich in vocabulary, and of high profession in the societal rank might be comfortable with using niched words to express themselves intelligently. Someone whose age group belongs to teens is most likely to use slang words within their social circle. Otherwise, most standard tweets will probably belong to the third option. Having the opportunity to look at the raw tweets from the Table of Reference prevents you from making assumptions, and the result of the assignment you make here will make your marketing content better informed as to how to approach the target audience from a linguistic perspective.

The third set or group of questions that follows deals with the length of texts:

Again, this opens up to a better-informed marketing content from a linguistic point of view, but this time pertaining to the length of texts. Not all people appreciate lengthy texts, unless they’re really into reading. On the other hand, some people prefer short texts because they want to quickly absorb the information they seek, or because it resonates well with how their brain is hardwired biologically.

Ending the section for Step 2’s component is the note-taking and profile removal section. Here, you’ll be taking notes of the tweet content by paraphrasing each tweet into active sentences — a common exercise in marketing practices called empathy mapping:

You may tick the box should you wish to remove an avatar profile from the equation. For avatars or personas whom you like to keep, you may double click on the respective cell to start typing your notes, guided by the on-screen instructions.

Empathizing with the target audience puts yourself in their shoes to resonate with their state of emotions. It’s an initiative to start offering better services and products, better content creation as well advertisement copies which are curated specially for them, coaxing their perception of your offerings to shift to the more positive side.

Step 3 — Cluster Profiling

While Step 2 allows you to profile the personas, Step 3 introduces you to the clusters in which the personas belong, and thus allows you to start profiling them collectively in their respective groups.

The detailed instructions will pop up once you’ve right-clicked on the component.

This section displays the info you need to know about each cluster. Some are based on your responses in Step 2, while the rest comes from the automatic clustering and querying processes in Step 1.

Following that, you’ll be given another note-taking exercise:

This is your opportunity to profile the clusters comprehensively, using the information displayed earlier, in addition to looking at the profile images of the Twitter users behind the clusters. (Yes, sometimes images do help, especially if they show an obvious pattern of demographics like ethnicity, gender or estimated age group).

If you scroll down to the bottom, you shall find these questions to navigate the cluster profiling task towards a successful completion:

These 11 awesome guiding questions ensure you’re able to cover as much important info as possible, leading to a solid profiling of the clusters!

Step 4-Humanizing the Avatars

By now, you should have the cluster profiles completed. It’s time to inject a little bit of human elements in them, in order to make them appear lively and realistic.

The detailed instructions will pop up once you’ve right-clicked on the component.

Humanizing your avatars makes you rethink the way you interact with them, whether it’s about mastering the craft of writing powerful opening lines, or about being interested first, instead of trying to be interesting — whatever it is, treating them as humans first really goes a long way!

If your avatar is stressed from working all day, you can lighten up their day by making them laugh. If your content serves a sales purpose, post something about a solution to their pains instead of bluntly pushing your products or services right to their face. Inability to relate to them typically comes from a mismatch of personalities; a misfit between your brand and the target personas. Hence is the importance of this last exercise:

What good is a humanized avatar without a name, right?

Step 5 — Generating PDF Reports

Finally, your last step brings you to the joy of generating the PDF reports for your own documentation and for showing off to your work colleagues!

The detailed instructions will pop up once you’ve right-clicked on the component.

There you go: the complete avatar profiles that you can export as PDF files to any folder location in your computer!

Apart from spitting out the date, location coordinates and associated entities, the report also includes the juicy stuff which advertisers and content creators would crave for — the target devices as well as preferred media type.

It’s never a good thing to overuse a single content type; always embrace various formats instead, to mix things up and keep the audiences hooked. Besides that, reaching out to them where most of them hang out as far as device types are concerned will be like reaching out for the low-hanging fruits off of a tree — it’s super efficient.

As the market becomes more and more saturated with the same content on a daily basis, created by both humans and artificial intelligence, the constant cry for attention in the sea of deafening marketing voices can lead to consumer brand fatigue.

Knowing your pool of audience better than your competitors will definitely help your brand stand out and keep the marketplace all freshened up with new and exciting content!

Get this free marketing tool in the official KNIME Hub repository.

Oh, and many thanks to the wonderful people whose articles I’ve cited here, Dennis Ganzaroli, Cheanu Chew, theaniconanan, Germán Coppola, Maxine Ray, Kaushik Shakkari & Abubakar Alaro — without your awesome articles to back up or supplement my points, my words mean nothing.

Found that one thing that impressed you? Now go and use it to impress your Twitter audience! Share the knowledge on Twitter.