AAU Student Projects is unavailable between June 15th 1.30pm and 17th 1.30pm due to planned system maintenance. The projects cannot be downloaded during this period.
AAU Student Projects - visit Aalborg University's student projects portal
An executive master's programme thesis from Aalborg University
Book cover


From Qualitative Content to Quantified Insights: LLM-Based Analysis of LinkedIn Communication Strategies and Engagement

Authors

;

Term

4. term

Publication year

2016

Submitted on

Abstract

This thesis examines how Danish business-to-business (B2B) companies craft their LinkedIn posts and how different ways of framing messages relate to audience engagement measured by likes, comments, and reposts. LinkedIn is less studied than consumer-oriented platforms, and manual content analysis is time-consuming. The study also shows how large language models (AI) can classify communication strategies at scale. The study begins with a systematic mapping of prior research to identify common strategy types. These are turned into five practical categories: 1) Thought leadership and expert positioning (showing expertise), 2) Product- and solution-centric communication (presenting offerings), 3) Storytelling and personal communication (human-centered stories), 4) Interactive and dialogic communication (inviting questions and dialogue), and 5) Value-based and CSR communication (values, responsibility, and society). A manually coded gold-standard dataset is then used to test different models and prompting setups. GPT-5.4, Claude Opus, and Qwen3-30B are evaluated with zero-shot (no examples) and few-shot (a few examples) prompts. Claude Opus with a preliminary few-shot prompt is selected because it delivers the most balanced performance across categories. This setup is applied to 1,200 LinkedIn posts from Danish B2B firms. Findings show that product- and solution-centric content is most common, while interactive/dialogic posts are used only sparingly. No single strategy performs best on all metrics: storytelling/personal posts attract especially many comments and also perform well for likes; value-based/CSR posts perform strongly for likes and reposts; and product-/solution-centric posts have the highest average level of reposts. A regression analysis assesses whether these differences are statistically meaningful. It finds that strategy is significantly associated with likes, comments, and reposts, but with limited explanatory power—framing matters for engagement, yet other factors also play an important role.

Afhandlingen undersøger, hvordan danske B2B-virksomheder (der sælger til andre virksomheder) udformer deres LinkedIn-opslag, og hvordan forskellige måder at indramme budskaber på hænger sammen med engagement målt som likes, kommentarer og delinger (reposts). LinkedIn er mindre undersøgt end forbrugerrettede platforme, og manuel indholdsanalyse er tidskrævende. Studiet viser også, hvordan store sprogmodeller (AI) kan bruges til hurtigt at klassificere kommunikationsstrategier i stor skala. Først kortlægges tidligere forskning systematisk for at finde gennemgående strategier. Disse oversættes til fem praktiske kategorier: 1) Thought leadership og ekspertpositionering (vise viden), 2) Produkt- og løsningsfokus (præsentere tilbud), 3) Storytelling og personlig kommunikation (menneskelige historier), 4) Interaktiv og dialogisk kommunikation (stille spørgsmål og skabe dialog), og 5) Værdibaseret og CSR-kommunikation (værdier, ansvar og samfund). Et manuelt kodet guldstandard-datasæt bruges derefter til at afprøve forskellige modeller og prompt-opsætninger. GPT-5.4, Claude Opus og Qwen3-30B testes uden eksempler (zero-shot) og med få eksempler (few-shot). Claude Opus med en foreløbig few-shot prompt vælges, fordi den samlet set klarer sig mest balanceret på tværs af kategorier. Denne opsætning anvendes på 1.200 LinkedIn-opslag fra danske B2B-virksomheder. Resultaterne viser, at produkt- og løsningsfokuseret kommunikation er mest udbredt, mens interaktiv/dialogisk kommunikation kun bruges i begrænset omfang. Ingen strategi klarer sig bedst på alle mål: Storytelling/personlig kommunikation giver særligt mange kommentarer og også mange likes; værdibaseret/CSR-kommunikation klarer sig godt på likes og delinger; og produkt-/løsningsfokus har i gennemsnit flest delinger. En regressionsanalyse vurderer, om forskellene i engagement er statistisk meningsfulde. Analysen viser, at strategi hænger signifikant sammen med likes, kommentarer og delinger, men forklaringskraften er begrænset. Det peger på, at indramning betyder noget for engagement, men at andre forhold også spiller en vigtig rolle.

[This apstract has been rewritten with the help of AI based on the project's original abstract]