Optimizing
rewards
for
sustainable
purchasing
@
Role:
Behavioral
Science
MSc
Consultant
Duration:
6 months
Skills:
Quant
methods,
Qual
methods,
behavioral
diagnosis,
triangulation
Tools:
R
Studio,
Kumu,
Teams,
Excel
Languages:
R
Role:
Behavioral
Science
MSc
Consultant
Duration:
6 months
Skills:
Quant
methods,
Qual
methods,
behavioral
diagnosis,
triangulation
Tools:
R
Studio,
Kumu,
Teams,
Excel
Languages:
R
While completing my MSc Behavior Change at University College London, I consulted for Reewild, a Mastercard Start Path startup, to evaluate and generate research-backed recommendations for their Planet Points sustainability rewards scheme pilot.
The remit was broad with only two requirements:
1. Tell us whether it's working
2. Tell us how to make it better
Reewild (now Vela Rewards) created Planet Points to reward people for purchasing lower-impact food and drink options. Users receive points per pound (£) spent depending on the eco-rating (A-E) of their purchased items.
The pilot ran for 6 months from January to June 2025 with 900 participating users and 23 participating locations across University College London and the surrounding Bloomsbury borough.
Reewild's
rationale
for
Planet
Points
is
to
target
the
sustainability
intention-behavior
gap:
Many
people
care
about
the
environment
to
some
degree;
however,
this
rarely
translates
into
consistent
action.
Pilot
evaluation:
Reewild
initiated
the
pilot
to:
1. Understand the degree and nuances of how rewards may influence sustainable purchasing
2. Identify and leverage optimization opportunities
Participants purchasing A and B-rated food and drink options at participating locations when hungry or thirsty.
To what degree and across what food types and locations does the scheme influence performance of the target behavior?
Which behavior change techniques are already operationalized in the scheme's digital and in-person experiences?
What barriers and enablers influence performance of the target behavior?
How can the scheme's effect on target behavior performance be improved?
POS quantitative analysis
Point-of-sale data from 187,000+ transactions were analyzed through a difference-in-difference econometric model to quantify the impact of Planet Points on purchasing behavior.
Behavior change technique (BCT) audit
185 unique components of the UI, messaging, and in-person experience were audited against the BCT Taxonomy (BCCTv1), identifying 10 integral and 14 subsidiary BCTs.
Participant interviews
15 45-minute interviews were conducted. Interview schedules were generated based on the components of the Theoretical Domains Framework. Transcript data was analyzed using the COM-B model, mapping behavioral barriers and enablers across: capability, opportunity, and motivation.
Triangulation
Findings from the quantitative analysis, BCT audit, and participant interviews were synthesized using the Behavior Change Wheel framework.
This triangulation allowed behavioral outcomes, intervention components, and participant experiences to be evaluated together, generating a clear diagnosis of how and why the intervention influenced behavior.
The combined evidence was then used to identify a series of informed recommendations to optimize the Planet Points scheme.
Difference-in-difference modelling
To gain a baseline understanding of the causal effect of joining Planet points on target behavior performance, I analyzed 200,000 transactions using difference-in-difference (DiD) econometric modelling. DiD tracks individuals' behavior before and after joining Planet Points as well as temporal trends, avoiding concerns of self-selection bias and supporting true causal inference.
DiD would also allow me to estimate the causal effect in different contexts, such as across food types and locations.
Sample:
900 users; 28,000 non-users; 200,000 transactions
Behavior change technique audit
To generate decision-ready recommendations, I first needed to understand the strategies already operationalized by the scheme. This would allow me to situate my recommendations around which strategies could be improved, added, or cut with reference to the identified barriers and enablers of the target behavior.
After collating 185 unique components of the UI, UX, messaging, and in-person experience, I conducted deductive thematic analysis against the Behavior Change Technique (BCT) Taxonomy (BCCTv1) to identify integral and subsidairy operationlized BCTs.
Sample:
185 touch points against 93 distinct research-backed BCTs
Participant interviews
To determine which BCTs are relevant to the target behavior, I first needed to identify its barriers and enablers. To do so, I conducted 15 45-minute interviews with a range of users. Interview schedules were generated based on the components of the Theoretical Domains Framework (an extension of the COM-B model). I then conducted inductive then deductive thematic analysis using the COM-B model.
Sample:
28 participants recruited & screened, 15 interviewed
POS quantitative analysis
Point-of-sale data from 187,000+ transactions were analyzed through a difference-in-difference econometric model to quantify the impact of Planet Points on purchasing behavior.
Behavior change technique (BCT) audit
185 unique components of the UI, messaging, and in-person experience were audited against the BCT Taxonomy (BCCTv1), identifying 10 integral and 14 subsidiary BCTs.
Participant interviews
15 45-minute interviews were conducted. Interview schedules were generated based on the components of the Theoretical Domains Framework. Transcript data was analyzed using the COM-B model, mapping behavioral barriers and enablers across: capability, opportunity, and motivation.
Triangulation
Findings from the quantitative analysis, BCT audit, and participant interviews were synthesized using the Behavior Change Wheel framework.
This triangulation allowed behavioral outcomes, intervention components, and participant experiences to be evaluated together, generating a clear diagnosis of how and why the intervention influenced behavior.
The combined evidence was then used to identify a series of informed recommendations to optimize the Planet Points scheme.
The methodologies and frameworks were specifically selected to support triangulation. The BCCTv1 provides literature-supported links between each BCT and one more component of the COM-B Model. By deductively coding participant interview transcripts against the COM-B model, I determined the relevance of each operationlized BCTs to the barriers and enablers of target behavior performance. I was also able to identify which BCTs should be added to the scheme. Finally, by bringing in the DiD model results, I could prioritize certain recommendations against the estimated magnitude of their effect.
To what degree and across what food types and locations does the scheme influence performance of the target behavior?
From difference-in-difference modelling
17 pp
increase
in
target
behavior
performance
for
hot
meals
Planet Points increased the likelihood of choosing low-carbon hot meals by 17 percentage points, showing stronger behavioral effects in high-emission food categories.
p < 0.05
From difference-in-difference modelling
11 pp
increase
in
target
behavior
performance
with
on-menu
eco-labels
When eco-labels were displayed on menu cards, sustainable purchases increased by 11 percentage points (though not significant at the 5% level), suggesting visible prompts may influence decisions.
p < 0.1
To support Reewild's external communications and investor relations, I also ran t-tests on emission reductions, basket value, and transaction frequency.
From t-tests analysis
16.8%
reduction
in
emissions
amongst
users
p < 0.05
From t-tests analysis
9.8%
more
transactions
per
month
amongst
users
p < 0.05
From t-tests analysis
5.5%
higher
basket
value
per
purchase
amongst
users
p < 0.1
Which behavior change techniques are operationalized in the scheme's digital and in-person experiences?
Integral
Feedback on outcome(s) of behavior
Instruction on how to perform the behavior
Information about consequences
Prompts/ cues
Material incentive (behavior)
Incentive (outcome)
Reward (outcome)
Conserving mental resources
Adding objects to the environment
Subsidiary
Goal setting
Discrepancy between current behaviour and goal
Feedback on behaviour
Social support
Anticipated regret
Social comparison
Credible source
Non-specific reward
Future punishment
Remove reward
Verbal persuasion about capability
Imaginary reward
Vicarious consequences
What barriers and enablers influence performance of the target behavior?
After thematic analysis, I mapped each theme in a behavioral systems map. This allowed me to see how each theme was connected and identify key chains, loops, and parent behaviors.
Combined with a deep understanding of the product's touch points gained via the behavioral strategy audit, the behavioral systems map and thematic analysis, I uncovered 8 targetable levers for furthering Reewild's positive impact on sustainable purchasing.
Value Framing
Users associated sustainability with health and ethics, suggesting wider value framing could strengthen motivation.
“For me, it's more of a health choice rather than a sustainable choice.”
Rating tangibility
Eco-ratings were understood at a high-level but left user wondering what the actual tangible impact of there decisions were.
“It would be really nice to be able to see what the impact actually is”
Recall friction
Participants frequently forgot to scan their Planet Points card at checkout, reducing engagement with the intervention.
“I think it’s easy to forget and maybe there could be more advertising for it.”
Reward schedule
Users often could not accumulate enough points quickly enough to redeem rewards, weakening motivational impact.
“it took so long <that I> gave up on trying to redeem again”
Personal impact knowledge gaps
Many users underestimated the impact that an individual’s food and drink choices can have on the environment.
“I don't know how much of an impact I as an individual can make”
Social proof
Users indicated stronger engagement with peer participation, suggesting social norms could reinforce adoption.
“When you see people using it you think... I'm generating more of an impact”
Reward Preference
Points reinforced sustainable behavior when users had a particular reward goal but failed to do so in lieu of a desired reward.
“I feel actually is motivates me to use more of this app”
Food emissions knowledge gaps
Effects were stronger in high-impact food categories where differences between options were more intuitive.
“I know that meat production is terrible and that I should not eat it.”
How can the scheme's effect on target behavior performance be improved?
After identifying which behavior change techniques (BCTs) were operationlized in the digital and in-person experiences as well as which factors were most influential to users' sustainable purchasing under the reward scheme, I generated clear decision-ready optimization recommendations.
Adapt existing BCT: Information about social and environmental consequences
Frame sustainable choices within broader value narratives (e.g., health, ethics) rather than just environmental outcomes.
Adapt existing BCT: Information about social and environmental consequences
Translate impact ratings into relatable environmental equivalents (e.g., “choosing A over C saves as much CO₂ as 10 plastic bottles”).
Adapt existing BCT: Reward (outcome)
Calibrate the reward schedule so users can earn their first reward quickly, experience the ah-ha moment, and start engaging more deeply.
Adapt existing BCT: Conserving mental resources
Prioritize engagement in high-emission categories where differences between choices are most salient and have higher environmental impacts.
Adapt existing BCT: Information about social and environmental consequences
Provide clearer educational cues explaining how individual purchasing decisions contribute to environmental impact.
New BCT: Environmental restructuring
Remove the need to scan at checkout and instead leverage institutional partnerships (Mastercard) to link users’ payment cards.
Adapt BCT: Material incentive (behavior)
Situate the reward pool within users' existing purchases (discounts, free items, etc.) to maximise the number of users pursuing a meaningful reward goal.
New BCT: Information about others’ behavior
Introduce social proof signals from other users, highlighting participation and the collective environmental impact of the user base.
What barriers and enablers influence engagement?
Value Framing
Users associated sustainability with health and ethics, suggesting wider value framing could strengthen motivation.
“For me, it's more of a health choice rather than a sustainable choice.”
Rating tangibility
Eco-ratings were understood at a high-level but left user wondering what the actual tangible impact of there decisions were.
“It would be really nice to be able to see what the impact actually is”
Reward schedule
Users often could not accumulate enough points quickly enough to redeem rewards, weakening motivational impact.
“it took so long <that I> gave up on trying to redeem again”
Category context
Effects were stronger in high-impact food categories where differences between options were more intuitive.
“I know that meat production is terrible and that I should not eat it.”
Knowledge gaps
Many users underestimated the impact that an individual’s food and drink choices can have on the environment.
“I don't know how much of an impact I as an individual can make”
Recall friction
Participants frequently forgot to scan their Planet Points card at checkout, reducing engagement with the intervention.
“I think it’s easy to forget and maybe there could be more advertising for it.”
Reward Motive
Points reinforced sustainable behavior when users had a particular reward goal but failed to do so in lieu of a desired reward.
“I feel actually is motivates me to use more of this app”
Social proof
Users indicated stronger engagement with peer participation, suggesting social norms could reinforce adoption.
“When you see people using it you think... I'm generating more of an impact”
How can the intervention’s efficacy be improved?
Frame sustainable choices within broader value narratives (e.g., health, ethics) rather than environmental outcomes alone.
Adapt BCT: Information about social and environmental consequences
Translate impact ratings into relatable environmental equivalents (e.g., “choosing A over C saves as much CO₂ as 10 plastic bottles”).
Adapt BCT: Information about social and environmental consequences
Calibrate the reward schedule so users can earn their first reward quickly, experience the ah-ha moment, and start engaging more deeply.
Adapt BCT: Reward (outcome)
Prioritize engagement in high-emission categories where differences between choices are most salient and have higher environmental impacts.
Adapt BCT: Conserving mental resources
Provide clearer educational cues explaining how individual purchasing decisions contribute to environmental impact.
Adapt BCT: Information about social and environmental consequences
Remove the need to scan at checkout and instead leverage institutional partnerships to link users’ payment cards.
New BCT: Environmental restructuring
Calibrate reward thresholds and expand the reward pool to maximize the number of users pursuing a meaningful reward goal.
Adapt BCT: Material incentive (behavior)
Introduce social proof signals from other users, highlighting participation and the collective environmental impact of the user base.
New BCT: Information about others’ behavior
"Your data analysis and recommendations from the UCL pilot have genuinely been invaluable.
We’ve used it extensively in joint external communications with Mastercard, and was also central to testing the new emissions avoidance framework developed by the World Resources Institute and the World Business Council for Sustainable Development.
Your work has also played an important role in our sales efforts. The evidence base from the pilot has helped underpin several upcoming partnerships, including deals with a large global food conglomerate, a major UK retailer, and a new collaboration with a world-renowned university."
Kit Nicholl - COO of Vela Rewards (Prev. Reewild)