GB794 Unit 10 Marketing Research Paper Example

New Orders or Moved Orders? A Unit 10 Marketing Research Paper on a Grocer's Delivery Rollout

[Student Name]

Purdue University Global

GB794: Advanced Topics in Marketing

Unit 10 Marketing Research Paper

[Instructor Name]

[Date]

Thornbury Market, its rollout and every result are composite and illustrative; the methods and the studies cited are real.

Abstract

When a grocer adds home delivery, the orders it receives are simple to tally and simple to misjudge. Some come from households that would otherwise have shopped at a competitor or not at all; others come from households that would have walked into the grocer's own stores. This paper estimates the net effect of Thornbury Market's delivery launch on total sales, using the staggered timing of a three-wave rollout to compare areas that had launched with areas that had not yet launched. Total sales rose by about 3.1 percent, while in-store sales fell by about 5.4 percent, implying that roughly two-thirds of delivery revenue replaced sales the stores would have made anyway.

Introduction

Thornbury's board has seen one figure about delivery: that it accounts for 8.5 percent of sales in areas where it operates. That figure answers the question of how much is sold through delivery. It does not answer the question that matters for investment, which is how much more the grocer sells because delivery exists. This paper's claim is that the incremental figure is about 3 percent, not 8.5, and that the difference is the cannibalization of in-store sales. The claim would be wrong if in-store sales in launched areas did not fall relative to areas that had not yet launched, or if they fell before delivery began.

Literature

Research on multichannel retailing has found both substitution and complementarity between channels. Pozzi (2013) studied the introduction of an online shopping service by a large supermarket chain that also ran physical stores and found that the online channel increased overall revenue by about 13 percent with limited cannibalization of store sales, in part because it reduced customers' travel costs and drew business from rival supermarkets. Wang and Goldfarb (2017), studying a retailer that opened physical stores, found that a new channel substituted for the existing one where the retailer already had a strong presence and complemented it where the retailer was little known, because the new channel then served mainly to inform customers that the brand existed.

The second finding suggests that results like Pozzi's need not apply to Thornbury. In its trading areas, Thornbury is well known and its stores are close to most of its customers. Delivery offers convenience, but not information about the brand. On the logic of Wang and Goldfarb, substitution should be stronger for Thornbury than for a chain entering areas where it was less established.

Design

Thornbury launched delivery in three waves: fifteen stores' trading areas in the second quarter of 2022, fifteen more in the fourth quarter of 2022 and the remaining fourteen in the second quarter of 2023. The order of waves was set by the readiness of each store's back room for order picking, which depended on renovation schedules fixed before the delivery decision, not on expected sales. That ordering matters because it makes the timing of launch plausibly unrelated to trends in each area's sales.

The outcomes are weekly in-store sales per store and weekly total sales, in-store plus delivery orders from the store's trading area. Because every store eventually launched, the analysis compares stores that had launched with stores that had not yet launched at each point in time. Goodman-Bacon (2021) showed that the conventional two-way fixed effects regression, applied to staggered adoption, can produce misleading estimates because it uses already-launched units as comparisons for later ones. For that reason the paper relies on the approach of Callaway and Sant'Anna (2021): each wave is set only against stores still waiting to launch, and those wave-by-wave contrasts are then averaged.

Results

Across the three waves, total sales in launched trading areas rose by about 3.1 percent relative to not-yet-launched areas, and the 95 percent interval runs from 1.2 to 5.0 percent. In-store sales fell by about 5.4 percent, with an interval from 3.7 to 7.1 percent. Delivery orders averaged 8.5 percent of total sales in launched areas over the first year.

The arithmetic links the three figures. If delivery accounts for 8.5 percent of sales and in-store sales fall by 5.4 percent, the net gain is about 3.1 percent, and roughly 64 percent of delivery revenue, 5.4 of the 8.5 points, replaced sales the stores would otherwise have made.

An event-study version of the analysis shows differences in in-store sales between launching and not-yet-launched stores by quarter relative to launch. In the four quarters before launch, the differences are close to zero and show no trend. After launch, in-store sales in launched stores fall within two quarters and remain lower. The absence of a pre-launch trend supports the design's central assumption.

Delivery added about three percent to Thornbury's sales; the other five and a half points it reported were sales the stores used to make.

Robustness

Three checks support the result. Excluding the last wave, which has the shortest comparison period, changes the estimates little. Assigning placebo launch dates a year before actual launches produces estimates near zero. And excluding stores whose trading areas saw a competitor opening during the study period, which could have depressed in-store sales for unrelated reasons, leaves the in-store decline within the original interval.

Loyalty data offer a further check on mechanism. Of delivery revenue in the first year, about 78 percent came from loyalty households that had shopped in Thornbury's stores in the six months before launch. That share is consistent with substantial substitution.

Why Thornbury Differs From Earlier Findings

The gap between this result and Pozzi's 13 percent gain with limited cannibalization deserves explanation rather than dismissal. Thornbury's stores sit close to most of its customers, so delivery saves them little travel and mostly changes how they shop rather than where. A chain whose stores are farther from customers, or which faces many nearby competitors, gains more from a channel that reduces travel and can draw shoppers from rivals.

Limitations

Three limits apply. The comparison relies on not-yet-launched stores, so effects beyond the first year can be estimated only for early waves. The analysis cannot observe purchases at competitors, so the source of the net gain, whether new customers or larger baskets from existing ones, is inferred rather than measured. And the result is specific to a dense, well-known regional grocer; a chain entering new areas might see more of the complementarity Pozzi and Wang and Goldfarb found.

Implications

Delivery's profitability should be judged against its incremental sales, about 3 percent, not its gross share, 8.5 percent. Because delivery orders carry picking and transport costs that in-store sales do not, sales moved from store to delivery may reduce margin even when total sales rise. A budget review should value delivery on the 3 percent net gain and the margin on moved orders, not on the gross figure.

References

Callaway, B., & Sant'Anna, P. H. C. (2021). Difference-in-differences with multiple time periods. Journal of Econometrics, 225(2), 200-230. https://doi.org/10.1016/j.jeconom.2020.12.001

Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal of Econometrics, 225(2), 254-277. https://doi.org/10.1016/j.jeconom.2021.03.014

Pozzi, A. (2013). The effect of Internet distribution on brick-and-mortar sales. The RAND Journal of Economics, 44(3), 569-583. https://doi.org/10.1111/1756-2171.12031

Wang, K., & Goldfarb, A. (2017). Can offline stores drive online sales? Journal of Marketing Research, 54(5), 706-719. https://doi.org/10.1509/jmr.14.0518