Our product, Sage, is a financial wellness platform meant to help reduce/eliminate impulse spending behaviors through AI-based intervention and coaching. Our platform relies heavily on behavioral and structural factors that we are not certain about, so in order to move forward with building, we tested the following assumptions:
Assumption 1: Impulse Spenders Want to Stop Impulse Spending.
For this assumption test, I conducted a rapid survey with close friends; because there wasn’t as much time to collect a pool of impulse spenders as there was with our past few studies, I chose to ask people who were comfortable with me about their spending habits so that they wouldn’t feel the barrier to an honest answer that I anticipated others might. I asked this question in a low-stakes, casual manner: if you consider yourself to be an impulse spender, do you actively want to stop/spend less? I let this poll hang amongst viewers for 24 hours, and within this timeframe, here’s what 14 responses showed:
These results were interesting; although we’d considered success to be any sort of positive indication that impulse spenders do have a desire to change their behavior, the results were not as overwhelmingly in favor of the assumption as I’d hoped they’d be; because our entire platform is relying upon this assumption, it is important that we see it as true beyond a shadow of a doubt. In this case, only 57.14% (8/14) of respondents had some sort of desire to stop impulse spending or spend less, and even then, only 35.71% (5/14) solidly said yes.
As far as our product goes, this research informs our branding strategy as well as the product’s voice; because the survey slightly dismantled our assumption and showed that some impulse spenders want to stop impulse spending and some don’t really care to, we can make sure that this product is attractive to impulse spenders of all perspectives by reframing the platform as a financial wellness app/guide rather than a “fix” for a problem.
Assumption #2: People won’t get too frustrated using AI
For this assumption test, we ran ab. Participants were instructed to interact with an AI conversational agent about a recent unplanned purchase and their thoughts/emotions surrounding that decision. The AI asked follow-up questions to help participants reflect on their purchase. After the conversation, participants were asked to rate their frustration level when using the AI on a scale of 1 (1 = Not frustrated at all) to 5 (5 = Extremely frustrated) and to explain why. In order for our assumption to hold, we decided that the average score across participants would need to be less than or equal to 3. 
Three people participated in this assumption test. Overall participants rated low levels of frustration. Two participants provided a rating of 2, indicating slight frustration. Participant 1 explained that the AI repeatedly asked reflective questions without offering new insights, and instead mainly affirmed their emotioned. Participant 2 indicated that the AI sometimes misinterpreted their reasoning but “it’s not that deep.” Participant 3 provided a rating of 1, indicating no frustration at all. He stated that the interaction was basic, and that it simply asked about their experience.
Across all participants, the average frustration score was 1.67. Therefore, our assumption that people would not be too frustrated for reflection was mostly validated. While there was some frustration due to repetitive or misinterpreted questions, participants were still able to complete their task with ease. One key insight from this would be to be more careful about prompting our AI conversational agent so that it knows when to stop asking follow-up questions and is better at understanding user reasoning.
Assumption #3: People can recognize impulsive purchases
To test whether people can accurately recognize impulse purchases, participants were presented with three short purchase scenarios. Two of the scenarios were designed to meet the definition of an impulse purchase, while one represented a clearly planned purchase. The scenarios included: Kathy buying an iPad after thinking about it for two months, Mario buying a coffee at Starbucks after realizing he felt tired while passing by, and John buying three bags of chips at the store after noticing they were on sale. Participants were then asked to classify each scenario as either “Impulse” or “Not Impulse.” The objective of this test was to validate the hypothesis that people can correctly identify impulse purchases. The success metric was defined as at least 70% of participants correctly classifying the scenarios according to the predefined impulse purchase criteria.
Seven people participated in this assumption test. Overall, participants were able to correctly recognize which purchases were impulse purchases and which were not. For Scenario 1, where Kathy had been planning to buy an iPad for two months, all seven participants correctly identified it as not an impulse purchase. For Scenario 3, where John bought three bags of chips after noticing they were on sale, all seven participants correctly identified it as an impulse purchase. Scenario 2, where Mario bought a coffee after realizing he felt tired while passing by Starbucks, was correctly identified as an impulse purchase by six participants, while one participant classified it as not an impulse purchase.
Across all responses, 20 out of 21 classifications were correct, resulting in an accuracy rate of about 95%. This exceeds our success metric of 70% correct classification, meaning the hypothesis was strongly validated. Participants were generally able to recognize the difference between planned purchases and impulse purchases. One key insight from this test is that people clearly recognize purchases that involve prior planning as non-impulsive, while spontaneous purchases driven by situational cues or convenience are widely viewed as impulse purchases. However, these types of purchases can still be interpreted differently depending on a person’s values. Because of this, it may be important to allow users to define what qualifies as an impulse purchase for themselves, since some people may view certain spontaneous purchases as necessities rather than impulsive spending.
