Thanks to Data Science, Analyze + Quote Payments In Under A Minute With Statement Scanner
You might remember that we put out a survey last week to bolster some of our observations in working with agents and ISOs on a data science project over the past few years. The results indeed validated what we’ve been hearing:
- People want an easy way to accurately analyze statements in the field
- People want an easy way to close the analysis with a quote
A lot of you survey respondents are presumably Global Payment agents because the third biggest concern was getting screwed by the very processor you “partnered” with — i.e. the processor raises rates and you, as the account holder, only hear about when the merchant calls you to complain about the bill. Not a great look for your reputation, we know.
At any rate, on September 1, 2020 we will publicly launch Statement Scanner, a statement analysis and quoting tool for ISOs and agents (or really anyone that touches payments processing). Up until now the tool has been in private beta with super ISOs whose logos will be made public in the coming months.
By virtue of recognizing uploaded statements, Statement Scanner can serve as your default CRM, automatically organizing your statements by MID. The top 25 statement contributors will get this CRM for free: all you have to do is sign up and drop your statements into the portal. And no, we’re not sharing your MIDs to sell processing into your existing customers. You don’t trust us by now?
Let’s do a quick rundown of your most pressing questions.
What Does Statement Scanner Do?
Merchants who accept credit and debit cards are given monthly statements — like receipts — from their credit card processors. These statements are intentionally obfuscated and it’s gotten so bad that even the payment processors who must review competitive statements to make an accurate proposal to a merchant cannot analyze them properly.
Statement Scanner instantly analyzes payment statements, avoiding the long delay that’s traditionally associated with statement analysis where sales personnel wait more than 24 hours for a result. Furthermore, Statement Scanner allows users to build a proposal right on the spot, greatly increasing sales velocity while eliminating quoting errors.
Who Is Statement Scanner For?
Statement Scanner works for ISOs and acquirers that want to expedite their statement analysis with a cost-effective, scalable software solution, freeing up personnel to be more productive. Statement Scanner also helps ISOs and acquirers set guardrails for quoting while managing multiple Schedule As.
Statement Scanner delivers independent agents the same benefits it confers to ISOs and acquirers with the added benefit of efficiently managing merchant portfolios: uploaded statements are automatically organized by MID with unlimited storage, acting as a CRM.
How Does Statement Scanner Work?
Statement Scanner uses a convolutional neural network (CNN) trained on tens of thousands of statements to accurately identify inaccurate and suspicious fields. Statement Scanner was built as a progressive web application, enabling the tool on any device, from desktop to mobile. The user need only upload a saved file or take a picture to begin analyzing and quoting a prospective account.
What Does Statement Scanner Cost?
$10 per statement. Today that fee covers both analysis and the quoting.
What Are Statement Scanner’s Limitations?
You know we’re all about transparency so here are the limitations of the tool:
1. Image Quality
We build Statement Scanner on top of Tesseract, Google’s open source image recognition software. Image recognition has improved markedly over the past decade but it’s still not perfect. That means low quality images still have a tough time being picked up. Technically you’d assess this by DPI minimums, but it just means that crappy pictures won’t fly.
Here’s the good news: the speed and accuracy of Statement Scanner has far outweighed the concerns of image quality in our pilots, and our pilot partners have justified reworking their statement processes to use the tool. In other words, our partners have told their agents something to the effect of “if you don’t submit a good enough image to be picked up by Statement Scanner then you can analyze your statement yourself.”
It’s time to grow up: no more pictures of statements with your hand covering half the text, mmkay?
2. Statement Coverage
The world of statements is a bit of you-don’t-know-what-you-don’t-know, and nobody truly knows when they’ve achieved 99% market coverage of statement types. Because we’re data scientists we hold ourselves to really high standards. While other people might accept a tight coupling in their solution, we want to make sure we’ve seen enough statement samples to undeniably mark the particular statement format as solved.
That said, we’ve focused our tool on interchange plus statements in the US, which represent somewhere around 75% of all statement types. Flat rate and tiered statement types will be next, as they’ll be built on a growing repository of tens of thousands of statements which will make the estimations better than what all but perhaps the four largest processors can estimate, and even then we question how well the processors have analyzed their own data.
3. Line Item Margin
As anyone familiar with statements will tell you, there’s no law dictating what processors must call each card type on a statement. A Visa Rewards Card with a known rate could show up on a statement as Bob’s Armpit with arbitrarily assigned rates and we wouldn’t know what that line item really was. Our approach has been to flag line items that we don’t recognize, but we make no attestation to what they really are.
Longer term this is a solvable problem with data dictionaries and some other machine learning techniques, but it might not be financially justified. Again, since we’re all about transparency, here’s a quick rundown on that logic.
There are probably 1 million statements quoted annually. Assuming every quoted statement used Statement Scanner at $10 per statement, that’s a total annual revenue potential of $10M. Because Statement Scanner is a transaction business, the valuation multiple of the business is likely to be more inline with payments at ~4x revenue, not ~8x revenue for SaaS. Therefore, at 100% market penetration, Statement Scanner would be worth $40M.
But we all know 100% penetration is impossible. An acceptable outcome would achieve half of that, or $20M in enterprise value, and that’s still assuming a lot of things go right.
Now let’s say it costs $500K to build the line item analysis and we treat that $500K investment as any rational investor and demand a 25% IRR on the money. That means we would need to triple the value of our money over five years. If we put in $500K that must net us $1.5M. In a perfect world without hiccups, competition, or other risks, we would need to own no less than 7.5% of Statement Scanner at an exit value of $20M. But because the world isn’t perfect, we’d need to add in a buffer: let’s 3x our ownership which gets us to 25% (which is coincidentally the amount of dilution to be expected if you raised a seed round).
Now you’re writing a check of $500K at a valuation of $2M. That leaves 75% of the company for founders who, mind you, are not being paid a salary on a meager $500K since that all goes into COGS for data processing. Thus the dilution is too much for too small a pie, and the market outcome is not large enough to even entice the $500K to begin with (investors look for businesses with TAMs of $10B or more, not a measly $20M.)
Therefore the people who can perform line item analysis fall into one of the following categories:
- Financially illiterate with zero value placed on their time. We can’t imagine it would be a pleasurable experience to do business with someone who doesn’t behave rationally.
- Intellectually dishonest with many shortcuts taken. Accuracy will be very low, leading to a plethora of risk in the quoting process.
- UNICEF. Tongue in cheek for a charitable endeavor because you’re just creating a massive tax write off.