Embedded Files
Analysis of Digital Advertising Performance
I decided to analyze a dataset of 1,800 digital marketing campaigns to understand what factors could improve customer acquisition efficiency. Using statistical analysis, I looked at how platform, campaign type, industry, ad spend, and acquisition metrics related to Return on Ad spend (ROAS) and Cost Per Aquisition (CPA). The goal was not only to identify statistically significant differences, but to translate the results into practical recommendations for how a global marketing team could allocate budget and improve campaign efficiency.
Research question: What should a global marketing team invest in to improve customer acquisition efficiency?
The Finding
The Finding
After the analyses were completed, it was found that Platform was the only variable that significantly predicted the Return on Ad Spend (ROAS), with TikTok Ads having the highest average ROAS at 9.54 compared to Meta at 6.92 and Google at 4.11. It was also found that lowering the Cost Per Aquisition (CPA) could be the highest-leverage action for improving acquisition efficiency, while increasing ad spend did not necessarily lead to better returns.
What I Did
What I Did
To answer the research question, “What should a global marketing team invest in to improve customer acquisition efficiency?”, I analyzed 1,800 campaigns across Google Ads, Meta Ads, and TikTok Ads, covering 5 industries, 4 campaign types, and 7 countries, with no missing values. I focused on ROAS and CPA: ROAS measures how efficiently ad spend generates revenue, while CPA measures the cost of acquiring a customer.
Using JASP, I first ran descriptive statistics to assess overall campaign performance. I then conducted three one-way Analysis of Variance (ANOVA) testing using ROAS as the dependent variable, testing Platform, Campaign Type, and Industry separately to identify factors associated with ROAS differences. Finally, I used a Pearson correlation matrix to examine relationships among ROAS, CPA, ad spend, Click Through Rate (CTR), revenue, and conversions.
The Data
The Data
The data gathered across 1,800 campaigns showed that overall ad performance was highly inconsistent. ROAS averaged 6.45 but had a standard deviation of 6.59- almost as large as the mean- with individual campaigns ranging from 0.13 to 49.00. That spread is what the ANOVAs were aiming to explain. Average CPA came into be $46.61, average ad spend at $6,172, and the average CTR at 3.8%.
The ANOVA test produced the clearest results in the analysis. With F=107.7 and the p<0.001, platform was a statistically significant predictor of ROAS. The mean ROAS by platform broke down as TikTok Ads at 9.54, Meta Ads at 6.92, and Google Ads at 4.11- meaning TikTok outperformed Google by 2.3x on average. Campaign type and industry told a different story. The campaign type ANOVA returned F=1.919 and p=0.125, and the industry ANOVA returned F=0.994 and p=-.409- neither reached statistical significance, meaning the format of the campaign and the industry it ran in did not meaningfully predict ROAS.
The correlation matrix added further context. The strongest negative correlation was between ROAS and CPA at r=-0.500 (p<0.001), confirming that as the cost to acquire a customer rises, return on ad spend falls, ROAS and ad spend also correlated negatively at r=-0.225 (p<0.001), suggesting that larger budgets alone do not drive better returns. On the positive side, revenue and conversions showed the strongest relationship in the dataset at r=0.831 (p<0.001). The only non-significant pair was CTR and CPA at r=-0.037 (p=0.113), indicating that getting someone to click does not reliably predict how efficiently they are acquired.
What It Means
What It Means
The biggest takeaway from the analysis was that where a company advertises mattered more than campaign type or industry when looking at ROAS in this dataset. Platform was the only variable tested that showed a statistically significant difference, making it the clearest area for a marketing team to consider when allocating budget.
TikTok's higher average ROAS does not automatically mean that a company should move its entire budget to TikTok. Instead, it shows that platform selection is worth investigating further before assuming that changing the campaign format or industry strategy will improve returns.
The CPA result was also important because it showed that improving efficiency may be more valuable than simply increasing spend. Since ROAS and CPA had a negative relationship, lowering the cost to acquire a customer was associated with stronger returns.
The negative relationship between ROAS and ad spend also showed that bigger budgets were not producing proportionally bigger returns in this dataset. This does not mean that increasing spend always hurts performance, but it does show that spending more should not automatically be treated as the solution to improving acquisition.
Finally, the CTR result showed that getting more clicks does not necessarily mean getting more efficient customers. A campaign can have a strong click-through rate without having a strong relationship with acquisition cost.
What I'd Do Differently
What I'd Do Differently
Based on these findings, the global marketing team should invest in optimizing each platform, reducing the CPA, and controlling scaling in the future.
Increasing the investment in higher performing platforms, which is currently TikTok with the highest average ROAS at 9.54, the team should receive more data testing for this platforms and different budget considerations compared to Meta ads and Google ads.
Due to the CPA having the strongest negative relationship with ROAS, reducing the CPA would improve conversion efficiency and optimize new campaigns. Additionally, the CTR was not significant to CPA, inferring that the team should prioritize conversions over clicks when evaluating the campaign success.
Increasing the budget for campaigns should be tested slowly and gradually tested rather than immediately due to higher ad spend being associated with a lower ROAS.
Skills Demonstrated
Skills Demonstrated
Research question design
One-way ANOVA
Pearson correlation analysis
Descriptive statistics
JASP
Translating statistical output into business-relevant findings
Data-driven strategic recommendation
Google Sites
Report abuse