Transaction declines are a pain for online merchants because, depending of merchants, they eat between 5% to 10% of transactions and are a source of bad customer experience. They are hard to analyze and to explain, and this is the reason why we have created Telescope. Today, we are releasing a major update to Telescope with a completely new way of explaining failed transactions: Telescope Diagnosis.
The trouble with failed payments
If you have read our article about the secret life of a transaction, you know that failed payments can be hard to analyze, and even harder to recover. This is why we introduced a product called Telescope in January of last year. We designed Telescope to improve the view online merchants have of their payment flow, and to give them a deeper understanding of what is happening in their payment infrastructure (what we affectionately call “the black box”).
How Telescope works
In the first version of Telescope, we simply began by feeding machine learning models a lot of transaction data. The trained models would in turn dynamically match transaction patterns and generate recommendations customized for all possible payment flows. This approach works well and gives very relevant recommendations for most merchants; unfortunately, because of the nature of the algorithm, we were effectively recognizing patterns only for problems we had trained the model for.
This means that when Telescope did not have a recommendation readily available for a specific class of transactions, we could ignore some glaring payment processing problems. Moreover, this approach can only find the best possible provider for a given transaction profile: it really shines when routing transactions in real-time, but it is not optimal for offline analysis. We felt we could go further and build instead a system that could dynamically find new issues ?
The very first version of Telescope, released in January 2017
Our new approach: Telescope Diagnosis
In that objective, we decided to rethink our approach for failure matching to produce recommendations that would be way more flexible. We needed to build a solution that could work on new decline patterns our models had never seen before. After several iterations… Say Hi to Telescope Diagnosis.
At the core of Diagnosis is a profile detection algorithm that is ran over groups of similar transactions with low authorization rates. Telescope Diagnosis prioritizes parameters that tend to lower authorization rates, such as issuing banks, currencies, or even the relative time of the transaction.
After multiple levels of aggregations, we end up with distinct groups of transactions that use common parameters, and that present worse performance than expected.
The results
As mentioned previously, Telescope Diagnosis is a tool specialized in recognizing bad authorization patterns. We use additional models to generate custom recommendations for every issue uncovered by the algorithm. This allowed us to upgrade three key features for merchants:
- Give a clear and transparent view of failed payments by listing every problem, its impact, and its source;
- Benchmark the performance of merchants with the performance they should be having;
- Offer actionable, prioritized insights to increase authorization rates.
We are very happy with the results the algorithm surfaces, and we discovered many new patterns. In addition of creating this new feature, we have updated the Telescope dashboard with a brand new UI to show off the level of details Diagnosis offers. This method works well for smart-routing and also gives merchants that use ProcessOut unprecedented depth to explore their failed transactions.
Telescope Diagnosis is a new step towards ProcessOut’s vision to blow the lid off the payment black box. Thanks to new algorithms, the main Telescope product offers improved results for every type of merchant, and brings payment performance to new levels for all customers.
If Telescope Diagnosis sounds like a tool you’d like to try or you need, register on ProcessOut today and / or contact us!