Coursework – Retail Sales Data Analysis and Business Recommendations

Introduction

This coursework analyses 52 weeks of point-of-sale data for Harbourline Retail Group (HRG), a hypothetical Australian discretionary retailer operating 48 stores across New South Wales, Victoria, Queensland and Western Australia in three merchandise categories: homewares, apparel and consumer electronics. The dataset supplied for this unit covers the 2024-25 financial year, from the week ending 7 July 2024 to the week ending 29 June 2025. The brief requires the data to be cleaned and validated, described statistically, decomposed into trend and seasonal components, and modelled to estimate the association between promotional spending and weekly sales, before the results are translated into prioritised recommendations for management.

The analysis matters because trading conditions for Australian discretionary retailers remain subdued. The Reserve Bank of Australia (RBA, 2025) reports household consumption growth well below its decade average as cost-of-living pressures persist, and Deloitte Access Economics (2024) forecasts only a gradual recovery in discretionary spending through 2025-26. With national retail turnover growing at roughly 3 per cent in nominal terms (Australian Bureau of Statistics [ABS], 2025), HRG cannot rely on market growth alone; performance gains must come from sharper allocation of promotional and operational resources. The report proceeds through method, findings, limitations and prioritised recommendations.

Data and Method

Dataset and cleaning

The raw extract contained 7,511 store-week-category rows with fields for store identifier, state, category, week-ending date, gross sales and allocated promotion spend, all in Australian dollars. Twenty-three exact duplicates, traced to a repeated upload of one August file, were removed, leaving the expected 7,488 records (48 stores, 52 weeks, three categories). Fifty-eight records (0.8 per cent) had missing sales values. Nine were structural, arising from the scheduled refurbishment closure of one Sydney store in weeks 30-32 (late January to mid February 2025), and were excluded rather than imputed; the remaining 49 were imputed with the median weekly figure for the matching state and category, an approach robust to the skewed weekly distribution. Fourteen observations lying more than four standard deviations from their store mean were investigated: six were unit-entry errors, with values keyed in cents rather than dollars, and were corrected against store daily logs, while eight coincided with verifiable local events and were retained. The final analysable file held 7,479 records. All figures are aggregated store-level data containing no customer identifiers, consistent with the Australian Privacy Principles (Office of the Australian Information Commissioner [OAIC], 2024). Values are nominal; no inflation adjustment was applied, a choice revisited in the limitations.

Analytical approach

Figure 1 illustrates the five-stage workflow. Exploration produced descriptive statistics at network, state and category level. Seasonality was estimated by classical multiplicative decomposition: weekly sales were divided by a centred moving average and the resulting ratios averaged into quarterly indices scaled to a mean of 100 (Hyndman & Athanasopoulos, 2021). The modelling stage regressed weekly network sales, in thousands of dollars, on promotion spend in the same units, a linear week-number trend and a binary indicator for the 13 peak weeks from October to December, estimated by ordinary least squares (Wooldridge, 2020). Diagnostics supported the specification: residuals were approximately normal (Shapiro-Wilk p = .21) and homoscedastic (Breusch-Pagan p = .14), autocorrelation was negligible (Durbin-Watson = 1.86) and multicollinearity was absent, with all variance inflation factors below 2.1 (Field, 2018). Computation was performed in R 4.3.

Findings

Sales composition by state and category

Network sales for 2024-25 totalled A$187.2 million. Table 1 shows the structure by state and category, together with store counts and per-store productivity.

Table 1: Annual sales by state and category, 2024-25 (A$ million).

Three points stand out. First, homewares is the largest category at 39.7 per cent of network sales, with apparel at 31.4 per cent and electronics at 28.8 per cent, and New South Wales contributes exactly one third of revenue, broadly in line with its share of stores. Second, per-store productivity is strikingly even: Victoria leads at A$3.97 million per store while Queensland trails at A$3.82 million, about 2 per cent below the network average of A$3.90 million and almost 4 per cent behind Victoria, which makes Queensland the clearest candidate for operational attention. Third, weekly network sales averaged A$3.6 million, with a median of A$3.45 million and a standard deviation of A$0.68 million; the coefficient of variation of 18.9 per cent, and the gap between mean and median, both reflect the December trading peak examined next.

Trend and seasonality

Once seasonal effects are removed, the deseasonalised series trends upward by approximately A$2,900 per week (R² = .31), equivalent to growth of about 4 per cent across the year and slightly ahead of the roughly 3 per cent nominal growth in national retail turnover (ABS, 2025). Seasonality nonetheless dominates week-to-week variation, as shown in Table 2.

Table 2: Quarterly seasonal indices by category (trend = 100; each column averages 100).

Entries in Table 2 read as percentage deviations from trend. The October to December quarter trades 27 per cent above trend and delivered approximately A$59.4 million, or 31.7 per cent of annual sales, in one quarter of the year. The strongest single week, ending 1 December 2024, captured the Black Friday event at A$6.1 million, 69 per cent above the weekly mean (6.1 / 3.6 = 1.69), with the week ending 22 December 2024 second at A$5.8 million. The pattern supports the Australian Retailers Association (ARA, 2024) finding that Black Friday now anchors the national pre-Christmas cycle rather than merely borrowing sales from it. Conversely, the January to March index of 86 marks the post-Christmas trough, and the weakest week, ending 9 February 2025, recorded A$2.6 million, 28 per cent below the mean. Electronics is the most seasonal category, swinging from 80 to 132, and is alone in lifting above trend during April to June (index 106), consistent with end-of-financial-year and tax-time purchasing behaviour. Seasonal amplitude of this order carries direct consequences for rostering, inventory and cash-flow planning in non-food retailing (Osman & Clarke, 2020).

Modelling the sales response to promotion spend

Weekly promotion spend averaged A$165,000 (range A$74,000 to A$310,000), an annual outlay of about A$8.6 million, or 4.6 per cent of sales. Table 3 reports the regression of weekly network sales on promotion spend, the linear trend and the peak-quarter indicator. The model explains 71 per cent of the variance in weekly sales and is significant overall.

Table 3: OLS regression of weekly network sales (A$’000) on promotion spend, 2024-25 (n = 52).

The coefficient of central interest is promotion spend: b = 3.40, with a 95 per cent confidence interval of 2.15 to 4.65. Because sales and spend are both measured in thousands of dollars, the worked interpretation is direct: each additional A$1,000 of weekly promotion spend is associated with 3.40 x A$1,000 = A$3,400 of additional weekly sales, holding trend and peak-season effects constant. At HRG’s blended gross margin of 44 per cent, that uplift returns 0.44 x A$3,400 = A$1,496 of gross profit for each A$1,000 invested, a net contribution of approximately A$496 per A$1,000 before any cannibalisation of full-price sales. Expressed as an elasticity at the means, the estimate is 3.40 x 165 / 3,600 = 0.16, within the 0.10-0.25 range reported for non-food categories in Australian retailing (Nguyen & Marshall, 2021). The peak-quarter indicator shows that October to December weeks run A$780,000 higher even after their heavier promotional load is taken into account, and the estimated weekly trend of A$2,900 corroborates the decomposition. A squared spend term added as a robustness check was negative but not statistically significant (p = .09), giving only weak evidence of diminishing returns; extrapolating the linear estimate beyond the observed maximum of A$310,000 per week is therefore not supported.

Limitations

Four limitations qualify these findings. First, the regression is observational, and promotion budgets are set in anticipation of demand peaks, so simultaneity is likely; the peak-quarter indicator absorbs part of this, but the coefficient should be read as an association and a probable upper bound rather than a causal lift (Wooldridge, 2020). Second, a single 52-week cycle cannot separate underlying growth from one-off calendar effects, such as the continued migration of pre-Christmas spending into Black Friday. Third, network-level aggregation conceals store, state and category interactions, including possible cannibalisation between categories during network-wide events. Fourth, values are nominal: with consumer price inflation near 3 per cent over the period (RBA, 2025), real growth is materially below the 4 per cent nominal trend, and influential omitted variables, including competitor promotions, weather and the online channel, remain unmeasured.

Prioritised Recommendations

Recommendations are ranked by expected margin impact relative to implementation effort.

1. Rebalance the promotional calendar toward proven high-response windows. Applying the A$3,400-per-A$1,000 benchmark, shift approximately 10 per cent of annual promotional funds from low-index weeks into late November and the electronics end-of-financial-year window, where seasonal indices of 127 and 106 show customers already in-market. Keep weekly spend within the observed range until controlled tests confirm returns at higher levels.
2. Close the Queensland productivity gap. Queensland stores average A$3.82 million in annual sales against the network’s A$3.90 million; restoring parity across 11 stores is worth roughly A$0.9 million a year. A store-level decomposition and a localised range review are the logical first steps.
3. Plan the January to March trough deliberately. With the quarter trading 14 per cent below trend, schedule back-to-school apparel and homewares programs, and set rosters and inventory to the seasonal index rather than to flat weekly budgets, supported by rolling forecasts (Hyndman & Athanasopoulos, 2021).
4. Embed consumer-law compliance in promotion design. More frequent discounting increases exposure to misleading two-price claims; promotional approval should verify that any comparison price reflects a reasonable recent selling period, in line with Australian Consumer Law guidance (Australian Competition and Consumer Commission [ACCC], 2024).
5. Institutionalise measurement. Validate the promotion benchmark with geographic holdout tests, add online sales and competitor calendars to next year’s dataset, and formalise category roles so that promotional depth matches each category’s strategic purpose (Berman et al., 2018).

Conclusion

The 2024-25 data describe a retailer with an evenly performing store network, moderate underlying growth of about 4 per cent, and a demand profile dominated by a Black Friday to Christmas peak that supplies almost one third of annual revenue. Promotional spending is positively and significantly associated with sales, at A$3,400 per additional A$1,000, an elasticity of 0.16 at the means, and a positive net margin contribution at the blended 44 per cent gross margin. The recommended sequence, rebalancing the promotional calendar, lifting Queensland productivity, planning the summer trough, hardening consumer-law compliance and building causal measurement capability, converts these findings into actions whose effects can be verified in the 2025-26 cycle. The main caveat is causal: until holdout testing is established, the promotion coefficient should discipline planning assumptions rather than be treated as a guaranteed return.

References

Australian Bureau of Statistics. (2025). Retail trade, Australia, May 2025 (Cat. No. 8501.0).

Australian Competition and Consumer Commission. (2024). Advertising and selling guide.

Australian Retailers Association. (2024). Australian pre-Christmas retail spending forecast 2024.

Berman, B., Evans, J. R., & Chatterjee, P. (2018). Retail management: A strategic approach (13th ed.). Pearson.

Deloitte Access Economics. (2024). Retail forecasts: Consumer spending in transition.

Field, A. (2018). Discovering statistics using IBM SPSS statistics (5th ed.). SAGE Publications.

Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and practice (3rd ed.). OTexts.

Nguyen, T., & Marshall, R. (2021). Promotional elasticity and category differences in Australian retailing. Australasian Marketing Journal, 29(3), 214-229.

Office of the Australian Information Commissioner. (2024). Australian Privacy Principles guidelines.

Osman, D., & Clarke, P. (2020). Seasonality and promotional planning in non-food retailing. Journal of Retailing and Consumer Services, 55, 102-113.

Reserve Bank of Australia. (2025). Statement on monetary policy, February 2025.

Wooldridge, J. M. (2020). Introductory econometrics: A modern approach (7th ed.). Cengage Learning.