RESUMEN EJECUTIVO

Digital marketing: each sector has a different audience, and generic metrics

can produce biases. This analysis extracts and creates generic metrics, then generates hypotheses from them, and finally tests those hypotheses with real data from 320 publications from 10 IT consulting firms operating in Chile.

Using Apify to extract data, an LLM to classify topics, and statistical analysis statistical analysis with non-parametric tests, I found that there is no significant evidence that videos outperform other formats (Mann-Whitney p=0.49). In fact, 77.8% of posts in the top quartile of engagement are images, not videos.

The most decisive factor in engagement is not format or topic, but account size (correlation -0.21) and posting frequency (correlation -

0.24): posting a lot does not guarantee results, and large accounts have lower engagement rates.

PROBLEMA DE NEGOCIO

Social media content is often based on "best practices" that are

generic. An analysis based on specific sector data allows you to optimize your organic content strategy, identify what works before investing in paid advertising, and benchmark against direct competitors.

Answering questions: What factors determine the success of an Instagram post for the IT consulting sector in Chile? Are the insights we obtain with metrics such as engagement real?

METODOLOGÍA

Extracción de datos

PHASE 1 - Data extraction 320 posts from @tipddy, @jhedai, @EY, @Accenture Chile, @PWC Chile, @BACKSPACE, @Cognitiva, @ACL Chile, @Thoughtworks, @Entel Digital, followed by cleaning and standardization.

PHASE 2 - Normalized KPIs Metrics such as engagement and hook to view standardized by followers — to compare accounts of different sizes.

Classification with LLM Each post tagged by main topic (Innovation, Cybersecurity, Culture, AI, etc.)

Market trends Google Trends analysis to identify industry keywords: one anchor word ("Process automation") and emerging words ("AI for companies," "AI for my business," "AI for businesses").

Hypothesis generation Dashboard construction and hypothesis generation based on exploration.

Statistical analysis

• Shapiro-Wilk: confirmed non-normal distribution (p<0.001)

• Mann-Whitney U: video vs. non-video comparison

• Spearman correlations: identifying factors associated with engagement

• Quartile segmentation: characterizing top performers.

FASE 1

Hallazgos

The difference of 0.06 percentage points is practically nil. The boxplot shows that both distributions are almost identical, with the same dispersion and similar ranges. There is no statistical justification for prioritizing videos over images in this sector.

1.Videos vs No Videos — Diferencia insignificante

2.Formato Sidecar domina

If we analyze the TOP 10 publications with the highest engagement in the sector, we find that Sidecar (carousel) with 39.4% of the Top 10%, videos are underrepresented in the Top 10% — they go from 45% globally to only 30.3% in top performers. This means that videos have WORSE relative performance than other formats.

Between medium-sized accounts (~1k followers) and large accounts (~5k), engagement drops by 34%.

This is the critical moment when companies must decide: do we grow our follower base or maintain engagement? Companies such as tipddy also demonstrate that posting infrequently but with high impact massively outperforms volume strategies. Their ratio of 13% with ~60 posts vs. 0.3% for Thoughtworks with ~1,200 posts is a 43x difference in efficiency.

3.Existe un "punto de quiebre" de escala y La estrategia "menos es más" tiene evidencia

DASHBOARD INTERACTIVO

Dashboard: Tres niveles de análisis

Vista 3: Google Trends

- Palabra ancla: "Automatización de procesos" (volumen estable)

- Palabra emergente: "IA para empresas" (+25 puntos crecimiento)

- Tendencias regionales (Bolivia, Uruguay, Brasil lideran interés)

Vista 1: Panorama del Sector

- Líderes por métrica (engagement video, no-video, hook-to-view, frecuencia)

- Promedios del sector

- Comparativa por tipo de post

View 1: Industry Overview

• Leaders by metric (video engagement, non-video engagement, hook-to-view, frequency)

• Industry averages

• Comparison by post type

View 2: Analysis by Company

• Filter by company

• Engagement by post topic

• Video performance (play, view, interactions)

• Benchmark vs. sector

View 3: Google Trends

• Anchor word: "Process automation" (stable volume)

• Emerging word: "AI for businesses" (+25 points growth)

• Regional trends (Bolivia, Uruguay, and Brazil lead in interest)

CONCLUSIONS AND RECOMMENDATIONS

Conclusions

tipddy breaks the scale—with only ~60 posts, it achieves 13% engagement, 6x more than the runner-up. This suggests that there is something specific about its content strategy that works extraordinarily well.

The Big 4 (EY, PWC, Accenture) have mediocre results despite resources and frequency. Their engagement is between 0.5-2%, well below smaller competitors.

The correlation of -0.36 is moderate to strong — posting actively REDUCES the engagement rate. Each additional post has diminishing returns.

Disconnect between content supply and demand

What people are looking for What consulting firms publish AI for businesses (+25 pts) Corporate culture Process automation Recognition and partnerships (Real demand)

(Corporate content)

Consulting firms are publishing what THEY want to show, not what the market is LOOKING FOR.

Strategic timing ignored The seasonality of "Process Automation" (peaks in April and September) is probably NOT being taken advantage of.

Consulting firms publish with constant frequency throughout the year instead of intensifying at times of high demand.

Recommendations

Publish carousels (not videos) about "AI for businesses" in March-April and August- September, when search demand is at its peak. Prioritize quality over quantity: tipddy shows that 60 well-executed posts are 43x more effective than 1,200 generic posts. If you're a small consultant, your size is your competitive advantage — accounts with fewer followers have twice the engagement.

that the big firms, and the Big 4, are losing out in the digital arena to boutiques that understand this. Finally, if you are looking for a viral post, the format that most increases your chances is video.

Recomendaciones

LIMITACIONES Y TRABAJO FUTURO

Limitaciones

|1 .Muestra de Limitations

1. Sample of 320 posts | Limited statistical power to detect small effects.

2. Instagram only | Does not capture LinkedIn (more relevant for B2B).

3. Short period due to sample size does not allow for seasonality analysis.

Future Work

1. Expand to LinkedIn | Main channel for the B2B sector.

2. Text content analysis | Which words/phrases correlate with engagement?

3. Predictive model | Estimate expected engagement before publishing.

Trabajo Futuro

1.Expandir a LinkedIn | Canal principal del sector B2B.

2.Análisis de contenido de texto |¿Qué palabras/frases correlacionan con engagement?.

3.Modelo predictivo | Estimar engagement esperado antes de publicar.