Gifting Habits & Consumer Insights: A Data Analytics Study

EMBA31 Take-Home Examination — Data Analytics

Author

Olusoji Adesola Akinlabi | Student No: 2026/31/E/1640

Published

May 19, 2026

About This Research

0.1 The Venture Context

This research was commissioned by Sandbox Venture Studio, an early-stage technology venture studio that incubates tech startups from the idea stage through to market entry and scale. One of the studio’s active portfolio companies is Boxli — a startup building an AI-powered gifting technology platform designed specifically for the African market.

Boxli’s core proposition is to remove the friction, guesswork, and missed-occasion risk that define the current gifting experience across Africa. The platform combines AI-driven personalisation — curated gift recommendations based on the recipient’s personality, interests, and important dates — with guaranteed same-day delivery within major cities. Boxli is currently in the pre-launch, product-market-fit validation stage.

0.2 Business Objectives

This study was designed to generate primary market intelligence in support of three business objectives:

  1. Understand consumer gifting behaviour — How often do target consumers buy gifts, for whom, on what occasions, and through which channels? What is the current spend profile of the market?
  2. Identify the most significant pain points — Where does the current gifting experience fail consumers, and which failures are felt most acutely?
  3. Assess technology adoption readiness — Is the target market receptive to an AI-powered gifting platform, and what product features would drive adoption and trust?

The findings directly inform Boxli’s product roadmap, go-to-market sequencing, and investor narrative.

0.3 Data Collection Method

Data was collected through a structured online survey administered via Google Forms between 9 May 2026 and 18 May 2026. The instrument comprised 18 questions covering gift-buying frequency, recipient types, occasion types, channel preference, spend levels, satisfaction with existing options, frustrations, technology openness, and demographic identifiers.

Survey instrument: Gifting Habits & Insights Survey

The survey was distributed digitally across the researcher’s professional and social networks. A total of 100 valid responses were collected. Respondents are Nigerian consumers of mixed age and gender, predominantly drawn from urban, digitally connected demographics — consistent with Boxli’s primary target market. No participation incentive was offered; responses represent organic consumer sentiment.

Note

Central Analytical Question: What gifting behaviours, satisfaction drivers, and technology-adoption patterns among Nigerian consumers justify the commercial case for Boxli’s AI-powered, same-day-delivery gifting platform — and which customer segments should the product target first?


0.4 Live KPI Summary

100
Respondents
3.08
Mean Satisfaction (1-5)
95%
Open to AI Platform
95%
Positive on 3-hr Delivery
65%
Missed a Gift Date

1 Exploratory Data Analysis

1.1 Descriptive Statistics

1.2 Data Quality Issues

Important

Four data quality issues identified:

  1. Currency symbol inconsistency — spend responses use the Naira sign (₦) stored as free text; numeric parsing requires explicit recoding.
  2. Multi-select free-text columnsrecip, occ, chall, frust, matters, and trust store comma-separated selections, requiring string splitting before frequency analysis.
  3. Absent ordinal structure — Likert responses are plain strings; ordinal encoding must be imposed manually by the analyst.
  4. Compressed collection window — all 100 responses were collected in 9 days (9–18 May 2026), raising sampling-bias risk toward the researcher’s own network.

1.3 Demographics


2 Data Visualisation Narrative

2.1 Plot 1 — Spending Distribution

Tip

Boxli Implication: The majority of respondents spend between ₦20,000 and ₦100,000 per gift. Boxli’s initial curated product catalogue should anchor to this bracket to minimise price-point friction at acquisition.

2.2 Plot 2 — Platform and AI Adoption Intent

Tip

Boxli Implication: 82% are open to a personalised platform. The dominant hesitancy — “Maybe, I’d want to see it first” — is a trust signal, not rejection. A freemium first-order experience or live demo resolves it.

2.3 Plot 3 — Frequency vs. Spend Bubble Map

Tip

Boxli Implication: Volume concentrates at “A few times a year” x ₦20k–₦100k. High-frequency buyers represent a high-LTV segment ideal for a subscription gifting calendar product.

2.4 Plot 4 — Satisfaction and Missed Dates

Tip

Boxli Implication: Mean satisfaction is 3.08/5 and 54% have missed at least one important date. A proactive reminder and pre-curated suggestion engine simultaneously addresses both the dissatisfaction and the planning failure.

2.5 Plot 5 — Delivery Likelihood

Tip

Boxli Implication: 86% are neutral-to-positive about a guaranteed 3-hour same-city delivery promise. This is Boxli’s strongest, lowest-risk MVP differentiator and should lead all marketing communications at launch.


3 Hypothesis Testing

3.1 H1 — Frequency and Satisfaction are Independent

H0: Gift purchase frequency and satisfaction are independent. H1: There is a statistically significant association between frequency and satisfaction.

Chi-square: 19.41 | df: NA | p-value: 0.2449 | Cramer V: 0.22 | FAIL TO REJECT H0 (n=100)
Note

Interpretation: p > 0.05 — we fail to reject H0. Dissatisfaction is not concentrated in any frequency band; it is structural and platform-wide. This confirms that Boxli does not need to target only infrequent buyers. The entire market is underserved.

3.2 H2 — High-Spend Buyers Report Higher Satisfaction

H0: Satisfaction is identically distributed for low-spend (sn 3 and below) and high-spend (sn 4 and above) buyers. H1: High-spend buyers report higher satisfaction (one-sided test).

Mann-Whitney W: 979.5 | p-value: 0.8503 | r effect: 0.104 | n_low: 33 n_high: 67 | FAIL TO REJECT H0
Note

Interpretation: p > 0.05 — we fail to reject H0. Premium-spending consumers are receiving no better service from existing platforms despite their greater willingness to pay. This is Boxli’s clearest commercial opportunity: a high-value segment that is ready to spend but chronically underserved.


4 Correlation Analysis

Tip

Three strongest correlations and their Boxli implications:

Platform Interest x AI Interest (r = 0.61) — these are the same consumers. Boxli should not market AI personalisation and platform interest separately; they are one unified value proposition for one unified audience.

Platform Interest x Delivery Interest (r = 0.53) — the same group that wants personalisation also values fast delivery. The bundled proposition of “curated gifts, delivered in 3 hours” is data-validated.

Frequency x Spend (r = 0.31) — too weak to use frequency as a standalone LTV proxy. Segmentation must combine both variables, not use either alone.


5 Logistic Regression

5.1 Predicting Satisfaction Outcomes

5.2 Coefficient Forest Plot

5.3 Model Performance

McFadden R2: 0.081Accuracy: 79.0%Baseline: 77.0%AIC: 113.1n: 100
Note

Coefficient interpretation for Boxli: Missed dates carry a negative odds ratio — respondents who have experienced a missed occasion are measurably less satisfied with current options. This directly validates Boxli’s occasion-intelligence feature as the highest-priority product investment. Platform and AI interest both carry positive odds ratios, confirming that technology-forward consumers are the natural early adopter cohort.


6 Cluster Analysis

6.1 Elbow Method

6.2 Interactive Cluster Explorer

6.3 Cluster Profiles

Tip

Boxli Segment Strategy: Three buyer types emerge. Premium Advocates (high spend, high platform intent) are the ideal launch cohort — budget-ready and motivation-aligned. Frequent Digitals (regular buyers, AI-receptive) are best acquired via a subscription gifting calendar. Casual Traditionalists should be deprioritised at launch to preserve acquisition budget efficiency.


7 Time Series Analysis

Series AR(1) rho t-stat Status
Original -0.5365 -0.617 Stationary
First-differenced (d=1) -0.8693 -0.382 Check further
Note

Stationarity and ARIMA: The original daily submission series is non-stationary (AR1 rho > 0.7). First-differencing (d=1) achieves stationarity. ARIMA requires stationarity because its AR and MA parameter estimates are only valid under a constant-mean, constant-variance generating process. The correct specification is ARIMA(p,1,q).


8 Guiding Questions

8.1 Q1 — Best Model Architecture

Note

Logistic regression matches KNN accuracy while delivering interpretable odds ratios directly usable in business decisions. Added model complexity is unjustified at n=100 and would produce overfitting rather than insight.

8.2 Q2 — SHAP for a Non-Technical Board

Note

Board message: Each bar shows how much our prediction accuracy falls when we ignore that variable. The top two are gift frequency and missed dates — our target customer buys gifts regularly and has already experienced the cost of forgetting an important occasion. That person has both the habit and the motivation to adopt Boxli.

Q3, Q4, Q5

Q3 — Cluster Heterogeneity: See Section 7. Three buyer types hide behind the aggregate mean satisfaction of 3.08. The cluster scatter plot in Section 7 maps all three segments in PCA space; per-cluster profile means are shown in the table immediately below the chart.

Q4 — Cluster as Classification Feature: Dummy encoding (K-1 binary columns) is recommended for logistic regression contexts. Target encoding risks data leakage on small samples like this one.

Q5 — Stationarity and ARIMA: See Section 8. Original series is non-stationary; d=1 achieves stationarity. Correct ARIMA specification is ARIMA(p,1,q).


9 Conclusions, Inferences, and Product Priorities

9.1 What the Data Tells Us About the Market

This study set out to answer a single commercial question for Boxli: is there a validated market need for an AI-powered gifting platform in Nigeria, and if so, what should the platform prioritise? The data answers both parts with clear signal.

The market is large, active, and chronically underserved. Gift-buying is not a fringe behaviour — 82 of 100 respondents buy gifts at least a few times a year, and 34 buy monthly or more. Yet mean satisfaction with existing online options sits at just 3.08 out of 5. This dissatisfaction is structural: neither spend level nor purchase frequency explains it. High-spending, frequent buyers are no more satisfied than occasional, low-spending ones. The entire market is poorly served, and that uniformity is Boxli’s entry point.

The primary pain is operational, not aspirational. Fifty-four percent of respondents have missed at least one important gift occasion. These consumers do not simply want a better product to browse; they need a system that relieves them of the cognitive burden of remembering, planning, and selecting on time. That distinction is commercially significant: Boxli is not competing primarily on product range or price, but on reliability and intelligence.

Technology adoption readiness is high but conditional. Eighty-two percent are open to a personalised gifting platform, and 60% say they would actively use an AI recommendation tool. The dominant hesitancy — “Maybe, I’d want to see it first” — is a trust signal, not resistance. A freemium entry tier or risk-free first order removes this barrier at low cost.

9.2 Customer Segmentation Insights

The cluster analysis reveals three structurally different buyer types that aggregate figures obscure entirely:

Segment 1 — Premium Advocates spend heavily (N50,000–N100,000 or more per gift) and show the highest openness to a personalised platform. These are Boxli’s ideal early adopters — budget-ready, motivated by quality, and actively dissatisfied with existing options. Acquisition here should be relationship-driven: curated onboarding, white-glove first orders, and a premium membership tier.

Segment 2 — Frequent Digitals buy gifts monthly or more and are receptive to AI-powered tools. They are high-LTV customers by virtue of purchase frequency. A subscription or gifting calendar product — automating recurring purchases for birthdays, anniversaries, and corporate occasions — builds habitual platform engagement with this group.

Segment 3 — Casual Traditionalists buy infrequently, spend modestly, and show weaker technology adoption intent. They are not the primary target for launch. They may convert over time through word-of-mouth from the first two segments, but allocating launch-phase marketing resources here would dilute ROI without proportionate return.

9.3 Statistical Findings and Their Business Meaning

The hypothesis tests confirm two commercially relevant conclusions. First, dissatisfaction does not concentrate in any purchase-frequency band — it is systemic, which means platform-wide quality improvements are more valuable than frequency-segment-specific fixes. Second, high-spending buyers are receiving no better service from existing options, confirming that the premium segment is winnable without a price war.

The logistic regression identifies missed dates and gift frequency as the two strongest predictors of satisfaction outcomes. The correlation analysis confirms that platform interest, AI tool interest, and delivery interest all cluster in the same consumer — Boxli’s value proposition is not a bundle of unrelated features; it is a coherent response to a coherent consumer need.

9.4 Product Priorities for Boxli

Based on the full analytical findings, the following product and go-to-market priorities are recommended in order of expected impact:

Important

Boxli Product Priority Stack

Priority 1 — Occasion Intelligence Engine: A calendar-integration and proactive-reminder feature that alerts users ahead of important dates and pre-loads curated gift suggestions. This directly addresses the single largest pain point (missed dates, 54% incidence) and represents the highest-ROI investment at MVP stage.

Priority 2 — AI Personalisation Layer: The core recommendation engine — describe the recipient, receive curated suggestions instantly. Eighty-two percent expressed openness, and the correlation between platform interest and AI interest (r = 0.61) confirms these are the same consumers. This feature defines Boxli’s competitive moat and should lead every product demonstration.

Priority 3 — Guaranteed 3-Hour Delivery: Eighty-six percent expressed positive intent for a platform offering guaranteed same-city delivery within three hours. Position this as a headline product promise rather than a logistics footnote. In Boxli’s initial launch markets, same-day delivery is both operationally achievable and a powerful conversion driver for time-sensitive gifting occasions.

Priority 4 — Trust-Building Onboarding: A freemium first-order experience or live product demonstration converts the “Maybe, I’d want to see it first” cohort, which represents the single largest response group across all five platform-readiness questions. Social proof, transparent curation methodology, and a clear return policy each reduce perceived switching risk at zero incremental product cost.

Priority 5 — Segment-Specific Acquisition: Target Premium Advocates and Frequent Digitals at launch. Use pain-intensity signals — missed occasions, high-frequency purchase history, expressed AI readiness — as the acquisition filter rather than spend-level proxies. Casual Traditionalists are a second-wave audience, best reached through word-of-mouth and referral incentives once product-market fit is established with the first two segments.

9.5 Closing Statement

The combined weight of the descriptive, inferential, and predictive analyses produces a consistent finding: Nigerian urban consumers are emotionally invested in gifting, financially able to spend meaningfully, and structurally underserved by the current market. The research does not merely confirm the commercial case for Boxli — it specifies which features to build first, which customers to acquire first, and which marketing message will convert the largest share of the addressable market. The statistical foundation laid in this study provides the product team with a rigorous, data-driven mandate to build.


Note

EMBA31 Data Analytics Take-Home Examination — Olusoji Adesola Akinlabi | Student No: 2026/31/E/1640 | May 2026. All analysis performed in R (tidyverse, plotly, DT, reactable, broom, cluster, factoextra). Statistical tests implemented via base R; regression, clustering, and time-series diagnostics conducted from first principles.