.webp)
Every transaction, every single tap with a credit card creates a stream of 0 and 1, and the ability to extract meaningful insights from that data is a key differentiator for banks and fintech companies. Data enrichment, especially in the payment sector, involves enhancing transactional data with additional information, offering a deeper understanding of customer behavior, reducing fraud, and ultimately improving the overall payment ecosystem.
When it comes to data enrichment services, three fundamental qualities need to be balanced to achieve the best results. In this article, we explore the crucial aspects of Coverage, Accuracy and Information richness, emphasizing the interconnected nature of these qualities for a comprehensive understanding.
Coverage
Coverage, measured as the percentage of transaction volume a solution provides results for, is a critical parameter. In the fintech environment it also follows the Pareto Principle 80% of consequences come from 20% of causes), where a handful of major merchants drive a significant portion of transactions. Striking a balance is key -reaching an optimal 80-90% coverage ensures that resources are not wasted on marginal increases, maintaining efficiency while upholding other essential qualities. Note that coverage is measured for each data point separately.
Example:
Out of 1,000,000 transactions...
- 85%of transactions have a clean merchant name (e.g., original merchant name"HFB ECO IKEA 069 OLDENBUR" --> clean merchant name"IKEA").
- 75%of transactions have a merchant logo (512x512 px, optimized for circular cutand visually checked by humans to ensure high quality).
- 95%of transactions have a correct merchant category.
- 70%of transactions have been enriched with location details, including a fulladdress and GPS coordinates of the shop where the transaction occurred (not the headquarters).
- 60%of transactions have been enriched with the merchant's URL.
Accuracy
Accuracy, or the percentage of enriched transactions done correctly, is the hallmark of data quality, since even a small error can lead to significant consequences, such as reputation loss for the entire company[LH1] . While it might be tempting to boost coverage at the expense of accuracy, the potential fallout from inaccurate enrichment, such as increased customer disputes or flawed analytical models, emphasizes the need for a meticulous approach. A company will lose valuable data when focusing on coverage at the expense of accuracy, which will have a negative impact on all banking features. In short – 75% Coverage with 99.9% Accuracy always beats 99.9% Coverage with 4% false data.
Example:
Merchant name – Is the correct merchant assigned?
Location
Is the detected location the exact address, wider area, or address of headquarters? Location verification ensures that enriched locations correspond to the actual shop where the transaction occurred. This is crucial as some providers might display the address of the merchant's headquarters or, in uncertain cases, provide a very general location, such as GPS coordinates of the Municipal District of Prague 8.
Examples:
Brand logo
What is the quality of the enriched logo and is it correct?
Category
How accurate is the categorization?
Information richness
The level of detail provided in the enriched information often dances on the boundary with accuracy. This aspect is subjective, dependent on specific use-cases and sensitivity to the level of detail required. Whether it's transaction recognition behind payment gates, GPS locations, or logo quality, a higher level of detail typically leads to less confusion and, consequently, better results. It's a delicate balance that should be evaluated based on individual needs and use-cases.
Know the difference
Logo Coverage: Addressing the logo coverage requires a nuanced understanding of the situation. Small businesses, often without logos, impact the feasibility of achieving 70-80% logo enrichment. In such cases, placeholder pictograms may be utilized, but it's crucial not to directly compare them to instances where logos are present. Additionally, recognizing online transactions to the level of payment gateways versus understanding the end recipient is crucial for proper coverage. For instance, transactions may be recognized at the gateway level (e.g., PayPal), but the true end recipient may differ.
GPS Location: Similarly, the differentiation between the GPS location of the shop, the merchant's headquarters, or the middle of a city street is an important detail. Recognizing the specific nuances, especially in online transactions, safeguards against artificial boosts in overall claim coverages. For instance, transactions conducted online may provide the GPS location of the company headquarters, but this doesn't necessarily reflect the location of the actual purchase. Understanding these subtleties is critical for delivering precise and reliable data.
POC as a cornerstone
In decision-making processes, relying solely on claimed coverages can be misleading. Running a Proof of Concept (PoC) becomes a perfect safeguard, where a sample of transactions is enriched within the same deadline. This approach allows for a direct and representative comparison based on evidence, avoiding simple claims and facilitating a holistic evaluation of solution metrics. POC is a unique topic we will explore in detail in the future articles.
Conclusion
Understanding the delicate interplay between coverage, accuracy, and detailing is what makes a good business tick. Striking the right balance ensures that data enrichment services not only meet but exceed expectations. As TapiX continues to enrich data for prominent companies, the emphasis on achieving this trifecta of qualities remains central to our commitment to delivering the highest value.
.webp)
