Table of Contents
Type: Presentation | Subject: Research | Level: Masters | Word Count: ~1400 words
This model presentation was produced by an Essays UK specialist as reference material for learning purposes only. For support in this field, see our postgraduate research and dissertation support.
Editable .pptx · 218 KB · free to use as reference material
You are a Master’s student preparing for your dissertation viva voce. Produce a 15-minute defence presentation summarising your research problem, methodology, key findings and contribution to knowledge, structured for a two-examiner panel.
Slide 1
A Mixed-Methods Study of UK Online Retail
Viva Voce Presentation
Candidate: Alex Whitfield, MSc Marketing Analytics
Principal Supervisor: Dr Sarah Coleman | Second Examiner: Dr Michael Osei
Slide 2
Slide 3
UK online retailers increasingly rely on algorithmic recommendation engines to personalise product suggestions, yet consumers report growing scepticism about how these systems use their data (Bawack, Wamba and Carillo, 2021). Trust, not technical accuracy, has become the binding constraint on adoption.
Existing research treats trust as a single measured outcome, typically captured through a short survey item, offering limited insight into why some shoppers accept personalisation while others disengage or switch to a competitor. Retailers have invested heavily in recommendation accuracy, yet abandonment rates on personalised category pages remain stubbornly high across the sector.
This gap between technical investment and consumer scepticism motivated the study, which asks not only how much consumers trust recommender systems, but why – and what, specifically, retailers could change to close that gap.
Slide 4
Aim: to explain the drivers of consumer trust in algorithmic recommendation systems within UK online retail.
Objectives:
RQ1: Which factors best predict consumer trust in an algorithmic recommendation system? RQ2: How do shoppers explain their trust judgements in their own words, and what do those explanations reveal that a survey score alone cannot?
Together, the two questions were designed to move the thesis beyond a purely descriptive account of trust levels towards an explanatory account of the mechanism behind them.
Slide 5
Prior work shows that perceived transparency (Kim, Ferrin and Rao, 2019) and usability shape online trust generally, but rarely tests these alongside personalisation-specific variables such as recommendation accuracy. Most studies are also quantitative-only, leaving the reasoning behind trust judgements unexplained.
This study addresses both gaps by combining a predictive model with interview data that unpacks the mechanism behind the statistics, answering calls in the literature for explanatory, mixed-methods designs (Pavlou and Gefen, 2020).
A conceptual framework was developed prior to data collection, positioning transparency, usability, personalisation accuracy and prior experience as parallel predictors of trust, with the qualitative phase held in reserve to explain whichever predictors proved strongest once the survey data were analysed.
Slide 6
A pragmatist, explanatory sequential mixed-methods design was adopted (Creswell and Plano Clark, 2018): a quantitative survey phase identified which factors predicted trust, and a qualitative interview phase then explained why those factors mattered to participants.
This sequencing was deliberate – the qualitative phase was designed to interpret the regression result, not simply to add supporting anecdotes alongside it.
A pragmatist stance was adopted rather than a strict positivist or interpretivist position, on the grounds that the research questions themselves – one predictive, one explanatory – called for different but complementary methods rather than a single paradigm applied throughout.
Speaker notes: If asked, be ready to justify the explanatory sequential design over a concurrent one – the qualitative phase exists to explain the regression finding, not to triangulate it independently.
Slide 7
Slide 8
A multiple regression tested four predictors of trust. The model was significant and explained just under half the variance in trust scores.
| Predictor | B | SE | β | t | p |
|---|---|---|---|---|---|
| Constant | 1.82 | 0.24 | — | 7.58 | <.001 |
| Perceived transparency | 0.41 | 0.07 | .38 | 5.86 | <.001 |
| Website usability | 0.27 | 0.08 | .24 | 3.38 | .001 |
| Prior positive experience | 0.19 | 0.09 | .15 | 2.11 | .036 |
| Personalisation accuracy | 0.11 | 0.08 | .09 | 1.38 | .169 |
R² = .46, F(4, 209) = 44.7, p < .001. Transparency was the strongest predictor, followed by usability and prior experience; personalisation accuracy alone was not a significant predictor once the other three variables were controlled for.
This was the most surprising result of the survey phase – the variable retailers invest in most heavily, recommendation accuracy, carried the least statistical weight in predicting whether a shopper trusted the system.
Speaker notes: Have the full assumption checks (VIF, Durbin-Watson) ready as a backup slide in case an examiner probes multicollinearity or independence of residuals.
Slide 9
Thematic analysis of the twelve interviews identified four recurring reasoning patterns behind participants’ trust judgements, strongest around transparency and weakest around personalisation accuracy – consistent with the regression result.
Figure 1: Thematic frequency in interview data (n = 12 participants).
Participants with high trust scores repeatedly described wanting to understand “why” a product was suggested, not simply to see accurate suggestions – explaining why transparency outweighed accuracy in the model. One high-trust participant put it directly: she trusted a retailer more once it told her a suggestion was based on a previous purchase, even when the suggestion itself was only moderately relevant.
By contrast, low-trust participants described feeling “watched” rather than helped when recommendations felt too accurate without any accompanying explanation, which helps make sense of why raw accuracy failed to predict trust on its own in the survey data.
Slide 10
RQ1 is answered directly: transparency and usability are the strongest predictors of trust, while personalisation accuracy alone is insufficient. RQ2 explains why – participants trust systems they can interrogate, not merely systems that guess correctly.
This integration is the study’s core contribution: accuracy-focused personalisation strategies may under-deliver on trust unless paired with visible explanation, a finding with direct implications for how UK retailers design recommender systems.
Practically, the findings suggest that a simple, low-cost explanation label – such as “because you bought X” – may do more for trust, and by extension conversion, than further investment in recommendation-engine accuracy alone. Theoretically, the study extends existing trust models by showing that transparency and accuracy are not interchangeable and can pull in different directions.
Speaker notes: Keep this section anchored to the two research questions – examiners listen for a direct answer to each, not a general summary of everything found.
Slide 11
Anticipated question: “Why not use a longitudinal design?” Response: time and scope constraints of a Master’s dissertation; proposed as future work in the concluding chapter.
Speaker notes: Do not undersell the sample-size or causality objection – acknowledge it directly, then pivot immediately to how the mixed-methods design mitigates it.
Slide 12
This study explains, rather than simply measures, why UK consumers trust or distrust algorithmic recommendations – with transparency, not accuracy, as the decisive factor.
Thank you. I welcome the panel’s questions.
Slide 13
Key sources consulted in preparing this presentation:
Need a Model Presentation Written to Your Exact Brief?
Our 350+ UK-qualified writers deliver referenced model documents from £15 per 250 words, with free plagiarism and AI-detection reports.
You May Also Like
All work is written by human writers. 100% AI free, guaranteed.
100% money back guarantee if you find plagiarism in our work.
COMPANY DETAILS
