Founded 2017 | HQ San Mateo, CA |52 employees March 2023

Veryfi has built an impressive IDP platform from its humble beginnings as a personal expense app, and the company has a strong track record in spend management and expense reporting use cases. The development team are AI experts who applied some incredibly complex inventions to solve a rather mundane business problem: how to enter data from purchasing documents.


The Company

Veryfi is an intelligent document processing (IDP) software company with a strong track record in spend management and expense reporting use cases. Launched in 2017 out of the legendary Y Combinator start-up incubator in Silicon Valley, Veryfi’s first product was an expense app to assist small business owners with receipts. The mothers of both founders are bookkeepers, providing a close and personal use case to drive the original product development.

The company believes that owners and their accounting helpers spend far too much time reading documents followed by the drudgery of manual data entry into accounting software or spreadsheets. Six years on, the company’s slogan reflects its roots: “Liberating the world from data entry.”

In 2021 Veryfi raised $12.7 million in a Series A round, and the company now has 52 full-time employees. While it positions itself as a general-purpose IDP platform for all unstructured data, spend management is still Verify’s leading business segment. The company is an OEM provider to several of the top spend management software companies.

Veryfi has also diversified its product line by launching data extraction solutions for loyalty marketing programs, landing a top-three food and beverage global giant and a top-five consumer packaged goods brand marketing agency as cornerstone clients. Invoice processing for accounts payable is another use case where the company has made some inroads, disrupting a market long dominated by legacy vendors.

The Technology

Veryfi offers two products: the OCR API Platform and the Lens Mobile Capture Framework. This report focuses on receipt data extraction, which most reflects the company’s core competency.

The OCR API Platform is a cloud-native service that will be invoked during expense reporting or other related financial processes. The API reads and extracts all details from purchase receipts and can understand and extract line-item details. This is done using AI models with no templates. Veryfi is quick to point out that it does not use outsourced humans-in-the-loop (HITL) for data extraction, in contrast to some competitors.

Veryfi’s background in receipts and expense management has provided rich training data for its AI models. This is a distinct competitive advantage, as most IDP AI models struggle to find enough valid training data.

In the process flow visualized by Figure 1, Veryfi accepts documents in several formats via mobile capture, as email attachments, or directly through the API. Underneath the hood, the Veryfi AI engine is using natural language processing (NLP) to enhance optical character recognition (OCR) and intelligent character recognition (ICR). Images are processed using a convolutional neural network (CNN) and data extraction is further enhanced through use of a graph neural network (GNN).

GNNs utilize the interrelations of documents or words to infer document labels and thus can create a connection between the different entities in a given subject, to provide meaning to the data and remove any semantic ambiguity. We are seeing increasing use of graphs in IDP to create a contextualized understanding of the text found in semi-structured or unstructured documents; the most common application to date has been for invoice data extraction, because different invoices from different vendors use a wild variety of data labels for the same thing. After the data is extracted, Veryfi applies data enrichment techniques to classify and enhance it. The end result is a standardized JSON data file fit for consumption by a spend management platform or any number of business applications.

As the vast majority of receipts will be captured by smartphones, every IDP vendor knows it is not enough to provide data extraction tools alone. To overcome the inherent hassles and clumsiness of smartphone document capture, Veryfi provides Veryfi Lens Mobile Capture Framework, with its own advanced mobile capture APIs. Lens is an impressive technology utilizing the latest in computer vision AI. The framework includes state-of-the-art features such as auto document detection (it can select the receipt from a background) and edge detection (to solve the problem of never getting the image proportions right in your phone viewer). The API also has an ingenious long receipt mode that can stitch multiple scans together into one long document.

After we suggested Lens functionality may be no better than Adobe Scan or the new Office 365 Scan app, Veryfi assured us that Lens was far more capable and invited us to test-drive it. So we installed Lens on an Android smartphone and tested a selection of receipts and other purchase-related documents such as invoices and billing statements (see Figure 2). Lens performed very well on a variety of less-than-ideal receipts and also did fine with invoices and bills of all type and size. The OCR extraction is very high quality, and the underlying ICR API also interprets print handwriting into textual data.

Yes, one could try to do this with the Adobe Scan app, which has similar computer vision tools (edge and doc detect) and does a capable full-text OCR on the documents we tested with Veryfi. But of course, Veryfi goes much further than Adobe Scan (or any of the other “free” scan-to-OCR apps popping up everywhere) by virtue of its superior machine learning models for all purchasing-related documents, and its ability to read, extract, and correctly format line-item data.

Figure 1
Veryfi Process Flow Chart
Figure 2
Veryfi Lens Includes Automatic Document Detection and Line-item Extraction

Our Opinion

Veryfi has built an impressive IDP platform from its humble beginnings as a personal expense app. The development team are AI experts who applied some incredibly complex inventions to solve a rather mundane business problem: how to enter data from purchasing documents. With their overwhelming advantage in real-world training data combined with user experience knowledge, Veryfi can more easily and quickly expand the platform into adjacent use cases such as loyalty programs. The company also has inherent technical advantages for invoice processing, but we think they will encounter much stiffer competition in the A/P automation market and will also find themselves selling into a more complex enterprise ecosystem dominated by ERP players and large system integrators.

Advice to Buyers

Data extraction from receipts has always been one of the most frustrating steps for expense management, and for many processes, the single point of failure. Veryfi has cracked the code and offers a near-seamless receipt processing service. Therefore, we think any business in the spend management and expense management market should have Veryfi on their vendor shortlist. Given its proven track record so far, Veryfi should also be under consideration for invoice processing and other semi-structured document processes. With the Series A round, committed investors, and claims of already running at a profit, Veryfi is a low-risk partner for enterprises looking for long-term vendor relationships.


SOAR Analysis

Strengths

  • Experts at mobile capture and data extraction, especially receipts
  • Very strong AI/ML development

Aspirations

  • Free the world from manual data entry

Opportunities

  • Loyalty program market expansion
  • Potential acquisition target for a spend management or FinServ company

Results

  • Impressive blue-chip customer wins
  • Raised $12.7 million and currently profitable

Attribution-NonCommercial-NoDerivatives 4.0 International
CC BY-NC-ND 4.0 license

We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.

Work Intelligence Market Analysis 2024-2029