This lesson sits at the analytical heart of the course. In the previous weeks we asked what is on social media and how people engage. This week we ask who influences whom, and how those relationships form measurable networks.
By the end of this lesson, you should be able to answer four practical questions about any real-world influencer campaign.
| Outcome | The question it answers |
|---|---|
| LO1 | What is influencer marketing today, and which trends are shaping how brands work with creators? |
| LO2 | Why do network structures matter more than raw follower counts when a brand picks an influencer? |
| LO3 | Which Social Network Analysis (SNA) metrics tell us who is truly influential — and how do we compute them? |
| LO4 | How can cluster analysis segment audiences so that brands target the right people with the right creators? |
To make the concepts concrete, we will follow one fictional brand across the whole lecture.
BrewLab is a Melbourne-based specialty coffee retailer with 12 cafés and a growing online subscription business. Their marketing director says:
"We've mastered organic social. Now we want to work with influencers — but we don't know who's worth paying, or how to prove it worked. Help."
Every concept today will loop back to this brief. By the end, you'll have a defensible answer.
Before we can measure influence, we need to agree on what an influencer is — and what has changed in the last five years.
Influencer marketing is a partnership in which a brand pays or exchanges value with a content creator so that the creator recommends, uses, or discusses the brand's product to their audience.
The word "friend" is doing a lot of work in the industry's favourite quote:
"Influencer marketing is like having a friend
who's really good at recommending things."
— Francois Marchand
Recommendations from a person you trust feel qualitatively different from an advertisement. Influencer marketing is engineered to trigger that trust response — at scale.
Influencers are usually classified by follower count. Bigger is not always better — the trade-off between reach and trust is the whole game.
In 2025, 75.9% of Instagram brand partnerships involve nano-influencers with under 10K followers. Why would a global brand hand a campaign to someone with fewer followers than their local café?
With a $40k trial budget, BrewLab could:
🅐 Pay one macro foodie (~$35k) for a single sponsored post reaching 400K followers.
🅑 Partner with 20 nano baristas and café bloggers ($2k each) reaching 200K followers combined, but with tighter local relevance.
Which would you recommend? Hold that thought — we'll return to it.
Five statistics that describe the industry BrewLab is entering.
Sources: IMH, HubSpot, SproutSocial 2024 Report, SurveyMonkey, AdWeek
The rules BrewLab would have followed in 2020 no longer apply. Here are the four shifts that dominate the 2025 conversation.
We'll unpack each one, then use them to design BrewLab's first campaign in the workshop activity.
Each trend answers a different marketing question: Who makes the content? (2.1) How do people find it? (2.2) Why do they believe it? (2.3) Where does it live? (2.4)
Virtual influencers — entirely computer-generated personalities like Aitana Lopez (350K Instagram followers) — are increasingly used by real brands.
AI influencers complement rather than replace humans. They work for aesthetic and aspirational categories; they struggle where lived experience is the whole point (parenting, mental health, fitness journeys).
Gen Z increasingly types product questions into TikTok, Instagram, and YouTube instead of Google. This changes what content earns discovery.
Influencer content must now be searchable, not just scrollable. Agencies optimise:
An unboxing video captioned "☕✨ love this!" is invisible. The same video captioned "BrewLab subscription review — best specialty coffee delivery in Melbourne" surfaces on TikTok search for months.
A creator who understands social SEO is now more valuable than one with 10× the followers who does not.
Modern campaigns thrive on relatability, not polish. Three content types dominate — each with a distinct source of credibility.
Content created by paid creators integrating the product into their own life. Builds trust because the creator's audience already believes them.
Example: A barista influencer brewing BrewLab beans in her home kitchen.
Content from ordinary customers — reviews, unboxings, testimonials. Highest credibility because there's no financial incentive.
Example: A subscriber posting their unboxing to their 200 followers.
Content from the brand's own team — humanises the company by putting real people in front. Especially powerful for B2B.
Example: A BrewLab head roaster showing the roasting process on Reels.
All three are experience-led. Both social search algorithms and human viewers now discriminate against content that looks like a traditional ad. The visual language of the campaign has changed.
One creator, one platform, one post — that formula is dead. Modern campaigns flow across platforms, using each for what it does best.
Design one campaign message that flexes to each platform's strengths. Consistency of story, adaptation of format.
Different platforms attract different audiences and serve different marketing purposes. Match the platform to the goal.
| Platform | Users | Core age | Purpose | Best for | Watch out for |
|---|---|---|---|---|---|
| 2.7B | 25–54 | Relationship building | Brand loyalty, community | Limited organic reach | |
| 706M | 30–49 | Thought leadership | B2B, professional networks | Limited organic interactions | |
| X (Twitter) | 1.3B | 18–29 | News & conversation | PR, news sharing | Character limits, volatile tone |
| 2.0B | 18–29 | Brand engagement | Retail, art, food, beauty | Visual-only format | |
| YouTube | 2.0B | All | Interaction & storytelling | Awareness, education | Resource-intensive to produce |
Source: prismglobalmarketing.com
You are the strategy lead for a digital agency. A new eco-friendly fashion startup wants to launch a bold, data-driven brand campaign. Goal: authentic engagement across platforms with a limited budget but high creativity.
Justify your choice using the demographics, purpose, and pros/cons from slide 2.5.
A trend-informed campaign is only half the job. The other half is proving — with numbers — that it worked. Which metrics matter, and which mislead?
Influencer marketing analytics is the practice of collecting, analysing, and interpreting data from influencer campaigns to produce actionable insights — decisions about who to work with, what content to produce, and where to invest next.
| KPI | What it measures | What it tells BrewLab |
|---|---|---|
| Engagement Rate | Interactions ÷ reach | Are people actually responding, or just scrolling past? |
| Reach | Unique viewers | How many different people saw the campaign? |
| Impressions | Total content views | How much total exposure did the content receive? |
| Conversion Rate | Desired actions ÷ visitors | Did the campaign actually sell subscriptions? |
Impressions grow with follower count. Engagement rate does not. A campaign that boasts 2 million impressions but a 0.1% engagement rate may have reached fewer meaningful people than a smaller one with 200K impressions and 8% engagement.
Reach and engagement tell you whether people responded. Audience insights tell you who they were — critical for deciding whether the right people saw the campaign.
Two BrewLab campaigns each reached 100K people:
Campaign A: 60% female, 25–34, Melbourne/Sydney urban, interests: sustainability, wellness.
Campaign B: 70% male, 45–60, regional, interests: sports, business news.
Same reach, wildly different value. Campaign A is BrewLab's actual customer.
To gauge true campaign impact, measure both what you can immediately count and what shows up later.
Measurable, short-term outcomes attributable to the campaign.
Return on Investment (ROI) is calculated primarily from these.
Longer-term, harder-to-quantify benefits.
Often called "brand equity" — small per campaign, huge over time.
Judging influencer campaigns on direct sales alone systematically undervalues them. Brands that only measure this way tend to stop investing right before the compounding brand effects kick in.
Even the best strategy fails if you cannot trace results back to the right influencer. These three tools do exactly that.
Unique URLs (via UTM parameters or short-link services) traced through the user journey from influencer content → website visit → purchase.
brewlab.com/?utm_source=nina&utm_campaign=spring
Each influencer gets a personal code (e.g. NINA20) that attributes every sale using it to that partnership.
Bonus: the discount incentivises the audience to use the code, giving you cleaner data than links alone.
Tools like Google Analytics, Meta Business Suite, or purpose-built platforms (GRIN, Aspire) aggregate conversions, traffic sources, and user behaviour in one view.
Enables real-time optimisation mid-campaign.
Attribution models are analytical frameworks that decide which marketing touchpoints get credit for a conversion. They answer: "Which part of the campaign had the most impact on the customer's decision?"
A customer named Alex buys a BrewLab subscription. Their path:
Question: which of those four touchpoints "made the sale"? The answer depends on your attribution model.
For BrewLab's four-touchpoint journey, multi-touch is the honest choice.
Work in groups of 3–4. Study these four influencer profiles. You'll match them to campaign scenarios in Part B.
| Name | Type | Platform | Followers | Audience | Content Focus |
|---|---|---|---|---|---|
| Sasha Luxe | Mega | Instagram, YouTube | 2.1M | Global, women 18–30 | High-end fashion, luxury travel, lifestyle |
| JayFit | Macro | TikTok, Instagram | 400K | Urban millennials, fitness enthusiasts | Fitness routines, healthy living, motivation |
| Nina's Nibbles | Micro | YouTube, Instagram | 35K | UK food lovers, 25–45 | Local food reviews, cooking tutorials, sustainable eating |
| TariqTech | Nano | Twitter, LinkedIn | 3.2K | Early-career tech professionals & students | Tech trends, coding tips, career advice |
Match the most suitable influencer(s) to each scenario. Justify your choice using type, platform, audience, and content fit.
| Scenario | Brand | Product / Campaign | Target Audience | Marketing Objective |
|---|---|---|---|---|
| 1 | Élan Couture | Luxury handbag line | Women 20–35, high disposable income | Increase visibility and desirability in fashion circles |
| 2 | Green Bean Café | Plant-based meals with local ingredients | Urban dwellers 25–45, eco-conscious | Drive local awareness and in-store visits |
For each scenario, pick one primary influencer (and optionally a secondary support influencer). Write a 3–4 sentence justification citing the concepts from Sections 1 and 2.
So far we've counted followers, engagements, and conversions. Now we ask a deeper question: what does the shape of the network tell us about who really has influence?
Two influencers can have identical follower counts and produce dramatically different results. The difference is where they sit in the network — who they connect to, and who those people connect to.
Same follower count. Radically different real influence.
Nodes (or vertices) are the actors — people, accounts, brands.
Edges (or ties) are the relationships — follows, mentions, replies, collaborations, tags.
Every social network is just a set of nodes with a set of edges between them. Once we can describe a network this way, we can measure its structure.
Social Network Analysis (SNA) is a methodological approach that maps and measures the relationships between people, groups, or organisations. It treats interactions on social media as a mathematical graph: a set of nodes (people) connected by edges (interactions).
Instead of just picking creators with the most followers in "food and coffee", BrewLab can use SNA to find creators whose audiences overlap with sustainability, wellness, and local Melbourne communities — the exact intersection their subscription box targets.
Degree centrality counts the number of direct connections a node has. In social media terms, it is essentially how many people directly follow, mention, or interact with this account.
The most intuitive measure — high degree = high potential immediate reach.
Fast to compute, easy to explain, good first filter.
Doesn't tell you whether those connections lead anywhere. An account followed by 10K bots has the same degree as one followed by 10K engaged fans.
Betweenness centrality measures how often a node lies on the shortest path between two other nodes. High-betweenness accounts are bridges that connect otherwise separate communities.
These are the gatekeepers of information flow between distinct groups.
A bridge influencer with 20K followers can spread a message across three communities. A mega-influencer with 2M followers who sits entirely inside one community may reach fewer distinct sub-audiences.
A creator active in both Melbourne food scene AND sustainability activism would be a bridge — the ideal single point of contact for BrewLab's positioning.
Eigenvector centrality weighs a node's importance by the importance of its neighbours. You are influential if you are connected to influential people — not just to many people.
Think of it as a quality score for your connections, rather than a quantity score.
A journalist followed by three Nobel laureates has more eigenvector centrality than one followed by 3,000 random accounts.
For BrewLab, this metric surfaces creators who are quietly respected within the coffee industry itself — the influencers other influencers follow.
PageRank is a refinement of eigenvector centrality that evaluates influence based on both the quantity and quality of incoming links. Originally invented by Google to rank web pages, it works the same way on social networks.
A node's PageRank score reflects the probability that a random walker jumping through the network would land on it. High PageRank = trusted authority.
PageRank penalises spam and pure follower-count games. It rewards accounts other credible accounts choose to engage with.
| Metric | Answers the question |
|---|---|
| Degree | How many direct connections? |
| Betweenness | Do I bridge separate groups? |
| Eigenvector | Are my connections influential? |
| PageRank | Do trusted accounts trust me? |
Professional SNA reports show all four. Different metrics highlight different candidates — the story emerges from the pattern.
Before SNA can happen, someone has to collect the raw graph. Three main routes, each with trade-offs.
Official platform interfaces (Meta Graph API, X API, LinkedIn API) provide structured data on posts, likes, follows, and shares.
✓ Legal, high-quality, structured
✗ Rate limits, approval required, often paid
Programmatically extracting data directly from public webpages, without going through an API.
✓ Access to data APIs don't expose
✗ Often against terms of service; ethically fraught
Commercial services (Brandwatch, Meltwater, GRIN) aggregate and license network data from multiple platforms.
✓ Ready-to-use, compliant, cross-platform
✗ Expensive, opaque methodology
Platform capabilities, budget, ethics review requirements, and how fresh the data needs to be. For coursework, we typically use pre-collected datasets or public API sandboxes.
Think about one social media platform you've used recently (Instagram, LinkedIn, X, TikTok). Write down 5–8 people you interacted with most in the past week — likes, comments, shares, or DMs.
Now we shift from analysing the network to segmenting the audience. Cluster analysis groups similar customers so brands can match the right creator to the right person.
Cluster analysis is a technique that groups entities — customers, influencers, or posts — based on similarity across measurable characteristics like demographics, behaviour, or content themes.
Within Social Network Analysis, cluster analysis groups nodes into communities. In marketing analytics more broadly, it groups customers into segments. Both use the same core idea: things that are alike belong together.
Cluster analysis unlocks four decisions marketing teams have to make.
| What clustering identifies | Decision it enables |
|---|---|
| Influencer communities — groups of creators with similar audiences or content themes | Which sub-scene should we partner with? |
| Niche vs central clusters — some clusters are small and specialised, others are broad | Are we going after depth or breadth? |
| Audience segments — customers grouped by behaviour and values | Which creator matches which segment? |
| Smart creator matching — AI-ranked lists of creators aligned to brand values | Which specific influencers should we shortlist? |
Without segmentation, all marketing is one-size-fits-all. With good segmentation, the same budget can be split across the right creators for each distinct customer group — dramatically improving relevance.
For our worked activity, we'll segment a real dataset for a fictional Australian retailer using K-means clustering in Python.
NaturaLife Australia is a mid-sized retailer of eco-friendly lifestyle products — organic skincare, reusable home goods, sustainable fashion, and ethically sourced wellness supplements. Flagship stores in Sydney, Melbourne, and Brisbane. Rapid growth since launching e-commerce and a mobile app in 2021.
NaturaLife's leadership wants to shift from one-size-fits-all marketing to hyper-targeted, data-driven campaigns. Better understanding of customers will improve:
NaturaLife collects data across store sales, app usage, surveys, and email — but the marketing team has no clear picture of who their key customer groups are.
"Can we segment our customers into meaningful groups based on their demographics, shopping behaviour, lifestyle preferences, and values like environmental concern?"
A clean dataset of 300 anonymised customers has been extracted for you, with the following variables:
What is a data dictionary? A reference document that describes each variable — its name, type, units, valid range, and meaning. Check Additional Resources for the NaturaLife data dictionary.
The Python notebook has been prepared for you. All you need is a Google account.
The K-means algorithm identified three distinct segments in the NaturaLife customer base. Each demands a different influencer strategy.
Why they fit: Digital-first, value authenticity, actively follow ethical creators.
Creator type: Micro / nano eco influencers who live the values, not just endorse products.
Why they fit: Trend-sensitive, appreciate exclusivity, trust expert tastemakers.
Creator type: Macro / luxury lifestyle influencers with cultural credibility.
Why they fit: Prefer traditional channels and personal trust; low social media engagement.
Creator type: Not recommended. Reach via email, in-store, and word-of-mouth instead.
Not every customer segment should be targeted with influencer marketing. A defensible strategy tells you where not to spend the budget, as much as where to spend it.
The four learning outcomes, revisited — with the one-line takeaway for each.
| LO | Key idea | The takeaway |
|---|---|---|
| LO1 | Influencer marketing & trends | Match the creator type (nano/micro/macro/mega) to the campaign goal — and pay attention to AI creators, social search, authenticity, and multi-platform flow. |
| LO2 | Network structures | Influence lives in network position (bridges, connectors, gatekeepers) — not in follower counts. |
| LO3 | SNA metrics | Degree, betweenness, eigenvector, and PageRank each answer a different question — use them together. |
| LO4 | Cluster analysis | Segment audiences before choosing creators — and know which segments don't want influencer marketing. |
The defensible recommendation isn't "hire one mega" or "spam 50 nanos". It's: segment your customer base, pick the segments most receptive to social influence, then choose creators whose network position reaches those segments — and measure with multi-touch attribution.
Next lesson we shift to generative AI for social media and marketing analytics — how tools like GPT and image generators are reshaping content production, campaign personalisation, and the cost economics of the influencer trends we discussed today.
If AI influencers become 10× cheaper and 10× more common over the next two years, what happens to the metrics from today's lecture — engagement rate, authenticity signals, network centrality? Which still work? Which break?
Questions? Post in the DATA4500 discussion board or bring them to office hours.
Press T or Escape to close