Just like the introduction, your general discussion follows a very formulaic structure

Section 1: Summary of the Findings

One short paragraph summarizing the results of the paper broadly. No need to get into specifics, pick the 1-3 findings you really want people to walk away with. If you have run an exploratory analysis, you can group your findings into something that hopefully is sensical for the reader.

Bad Example Good Example
We found in the main analysis of millions of grocery store purchases that people tend to choose more variety later in the day. Robustness checks of this analysis show X, Y, Z. By integrating data on sunrise and sunset for each day in the data and daylight savings, we find support for a circadian arousal mechanism. Follow-up experimental studies show that physical arousal mediates the effect and that chronotype is a moderator. We tested the effect of time of day on variety-seeking behavior across four studies, including experimental studies and a dataset of millions of grocery store transactions. People choose more variety later in the day, and this is driven by circadian changes in physiological arousal. As people’s physiological arousal (e.g., body temperature) naturally increases throughout the day, so does their preference for variety, which is an external source of arousal. Importantly, morning-types prefer variety all day, as they have higher physiological arousal in the morning than the rest of the population.

Good Example 2:

Over seven studies, we show that gifts are a more effective form of acute emotional support than are conversations. That is, support receivers feel emotionally better after receiving a gift than conversational support. This effect is mediated by perceptions of giver sacrifice. Specifically, support receivers believe gifts are more receiver-focused than are conversations, leading to differences in perceived sacrifice across conditions. (source: Howe, Wiener & Chartrand, 2024)

Good Example 3:

We first show that consumers engage more with teasing content even when funniness of the content is controlled for (studies 1A, 1B, and 2). Next, we find that prosocial teasing reliably leads to increased self-brand connection through brand anthropomorphism when compared to a purely humorous baseline (studies 3 and 4). In studies 5, 6A–C, 7A, and 7B, we demonstrate that this effect is limited to prosocial teases and that antisocial teases—because they activate a negative, rather than positive, human schema—do not garner the same positive effects. (source: Oba, Howe & Fitzsimons, 2024)

Section 2: Summary of the Major Contributions

Remember how you discussed the contributions of the work in your introduction? You want to mirror that here. Usually the contributions will remain roughly the same, but now with the context of your results you can really get into the details.

<aside> 📌

Recall: Your contribution cannot be “no one has ever looked at this before.” That is not a contribution.

</aside>

You may want to separate your discussion of contributions into “theoretical” and “practical” depending on how many of each you have

Here are some examples of contributions moved from intros to GDs:

In the introduction In the GD
First, we contribute to brand humor literature by bridging the gap between the success of
real-world teasing campaigns and the negative effects observed in prior work (Ning et al. 2022; Roehm and Roehm 2014; Thomas and Fowler 2021; Warren and McGraw 2016b) We provide a framework for understanding when
brand teasing leads to positive consumer responses. Despite the rising popularity of teasing campaigns, there was no existing theoretical framework to explain when and why this strategy should be effective. Prior studies had largely focused on antisocial teases in brand-to-brand communication (Ning et al. 2022; Roehm and Roehm 2014; Thomas and Fowler 2021) and had not identified the possibility that prosocial teases lead to
positive brand outcomes, even when the target of the tease is the consumer.
Second, we contribute to a small but growing liter-
ature in consumer behavior that looks at consumption
related to taking care of others' physical and emotional needs. Consumers spend time and money taking care of others, but with the exception of charitable giving(e.g., MacDonnell & White, 2015), research on what influences their behavior in this space is just beginning (Garcia-Rada etal., 2023; Liu et al., 2019). Most existing research has examined givers' motives for spending time or money on caregiving (Liu et al., 2019) or in response to a loved one's crisis (Wiener et al., 2022). We add to this literature by examining the receivers' perspective
and what forms of support are better at helping them
feel better. We demonstrate that gifts are a highly effective form of emotional support. Prior work has largely ignored gift-giving as a form of emotional support, focusing instead on monetary gifts as a form of tangible support(Barrera et al., 1981; Zimet et al., 1988). More recent work
has looked at the trade-off between gifts and emotional support, but from the perspective of the giver (Wiener et al., 2022). Our findings suggest that social support researchers should consider gift-giving when building a full picture of the actions involved in social support, especially considering that gifts may be more effective than other commonly studied approaches
First, by focusing on joint decision conversations, we diverge from the traditional approach of inferring the process of joint decision-making retrospectively, after the decision is made (Corfman and Lehmann 1987; Fisher, Grégoire, and Murray 2011; Tu, Shaw, and Fishbach 2016). This allows us to uncover more nuance in the process of how decision partners make a joint choice instead of focusing on characteristics of the decision outcome. In doing so, we answer calls for research about “joint journeys” through the decision lifecycle (Hamilton et al. 2021), conversation during joint decision-making (Cross and Gilly 2014; Epp and Price 2008a; Qualls 1987; Queen, Berg, and Lowrance 2015), and spoken language (Packard and Berger 2024; Yeomans et al. 2023). First, by using the decision conversation as the focus of our analysis instead of the decision outcome (e.g., Corfman and Lehmann 1987; Davis 1976a; Davis and Rigaux 1974; Spiro 1983), we answer calls for more research on spoken language (Packard and Berger 2024; Yeomans et al. 2023) and uncover novel insights about the process of joint decision-making, including the role of persuasion. Perhaps surprisingly, given the relative focus of the previous literature (Davis 1976a; Davis and Rigaux 1974; Filiatrault and Ritchie 1980; Munsinger, Weber, and Hansen 1975; Spiro 1983), we find that persuasion is not ubiquitous in joint decision conversations. We found in both studies that decision partners use persuasion less, on average, than getting on the same page and building and not all decision conversations contain persuasion. Further, persuasion was not the only communication pattern that had explanatory power for downstream outcomes of the decision. These findings highlight the need for future research to focus on the other communication patterns beyond persuasion; there could be much to discover about joint decision-making in doing so.

Section 3A: Limitations

Can be combined with the future directions section. Sometimes, published academic work doesn’t even include a limitations section because informed readers will be able to pick-up the limitations themselves. We don’t need to write about them.

For a master’s or PhD thesis, you do need to write about limitations. Your committee is assessing if you know and understand the shortcomings of your own work.