Stay ahead of the curve
Get exclusive insights delivered directly to your inbox. Join thousands of tech leaders.
TL;DR
- In 2004, online payments read an improvement rate of 331% a year against 28% for cash-payment infrastructure (GetFocus backtest). The gap first opened in 1998.
- Cash-handling technology still drew more new patent families than online payments until 2015 in the same data, and the global ATM count peaked in 2017 at 3.28 million (RBR).
- Banks backed both: more than 500 had signed on to Apple Pay by its October 2014 launch (Apple), and ATM numbers peaked three years later.
- The reading is directional and gives no date. Cash was still 52% of euro-area point-of-sale transactions in 2024 (ECB).
Digital payments disruption was readable in the patent record long before it reached the cash machine. By 2004, online payments measured an improvement rate of 331% a year against 28% for cash-payment infrastructure in GetFocus's backtest of payments, and the worldwide ATM count peaked thirteen years later, at 3.28 million in 2017 (RBR, reported by Digital Transactions).
We build GetFocus, which measures how fast technologies improve from the way patents cite each other. The rates in this article come from that public backtest, and every public figure links to its source. For an R&D or strategy leader the payments case answers a practical question: how early can you see a shift like this, and what does the early reading leave open?
What did the improvement rate say about online payments vs cash?
The improvement rate is how much a technology's performance improves per year. It works like an interest rate, so each year's gain builds on the last. GetFocus reads it from how patents build on each other, and the backtest computes every reading using only the patents that were public at that date.
The backtest compares two ways of solving the same problem, moving money from a payer to a payee. Cash-payment infrastructure covers ATMs, point-of-sale equipment and cash handling. Online and mobile payments enter the record in 1995. The readings:
- 1995: online payments 25.7% a year, cash-payment infrastructure 26.1%.
- 1998: online 99.9%, cash infrastructure 25.6%.
- 2004: online 331.2%, cash infrastructure 28.3%.
- 2010: online about 390%, cash infrastructure 29.1%.
- 2015: online 233.3%, cash infrastructure 29.6%.
- 2023: online 261.5%, cash infrastructure 32.9%.
The two start level in 1995. Three years later online payments read about four times the incumbent, and in 2004 more than ten times. The page's chart ends in 2025 with online payments at 190.7% and cash infrastructure at 32%, a gap of about six to one.
As an illustration only: a technology improving at 331% a year more than quadruples its performance every year, and one improving at 28% a year needs almost three years to double. The real rates moved from year to year, as the list shows, and online payments stayed far ahead at every reading from 1998.
Cash infrastructure was a healthy technology throughout, at about 26% to 33% a year, a rate the backtest page calls respectable. An improvement rate means something only next to the rate of the alternative, which is why the comparison stays inside one problem.
What did the industry's own filings show?
Patent volume shows where an industry is putting its effort. In the same backtest, cash-payment infrastructure drew 200 to 570 new patent families a year across 1990 to 2025, and it drew more new families than online payments in every year until 2015. Apple Pay, for example, went live in the United States on 20 October 2014, a year in which the cash side still filed more.
The crossover came in 2015, twenty years after online payments entered the record and eleven years after the 331% reading. By 2025 online payments drew 1,351 new families a year against 571 for cash infrastructure.
An analyst watching volume would have seen a steady incumbent until the mid-2010s. An analyst watching the rate would have seen the gap open by 1998. Our guide to testing a forecasting method on past shifts explains how to run this kind of replay yourself.
Did banks ignore digital payments disruption?
The public record shows banks backing both sides at once. When Apple Pay launched, Apple named American Express, Bank of America, Capital One, Chase, Citi and Wells Fargo as supporters and said more than 500 further banks had signed on (Apple, October 2014). The six named issuers were live on the first day.
The cash network kept growing at the same time. The number of ATMs worldwide peaked in 2017, three years after that launch. RBR, the research firm that tracks the installed base, linked the decline that followed to lower cash use, most sharply in China, where it credited mobile-payment services such as Alipay and WeChat Pay (Digital Transactions, December 2018).
The arguments for continued investment in cash were defensible when they were made. The backtest page lists them:
- Resilience. Cash works in a blackout, needs no account and settles on the spot.
- Trust. Banks expected consumers to hesitate before giving payment details to a website, and later to a phone.
- Regulation. Payments are among the most regulated consumer technologies, and rules were expected to slow newcomers.
- Its own numbers. Cash infrastructure was improving at a healthy rate.
Each argument describes the technologies on the day it is made. The improvement rate describes how fast that picture is changing. We wrote about this pattern in when to move away from a core technology that still pays.
What happened to cash-handling after 2017?
The installed base kept shrinking. The number of ATMs worldwide fell by 2% in 2022 to 3.0 million, and RBR expected the decline to continue until the end of 2028, leaving around 2.9 million machines (RBR press release, June 2023). RBR gave the reasons as bank branch closures, customers preferring digital banking and wider use of cashless payments.
Consumer behaviour moved the same way. In the euro area, cash made up 72% of point-of-sale transactions in 2019, 59% in 2022 and 52% in 2024, according to the European Central Bank's SPACE study. Online payments rose from 7% of day-to-day payments in 2019 to 21% in 2024.
What does Diebold Nixdorf's Chapter 11 show?
The ATM maker Diebold Nixdorf filed for Chapter 11 on 1 June 2023 in the Southern District of Texas (company 10-Q). It restructured over $2.7 billion in funded debt and left the process 71 days later, on 11 August 2023 (Jones Day).
The public record ties the filing to acquisition debt and weak sales. Reuters reported that the company, which sells ATMs and point-of-sale systems in more than 100 countries, had struggled with debt since it bought Germany's Wincor Nixdorf for $1.8 billion in 2016, and that its core business faced flat or declining sales of ATMs to banks and checkout machines to retailers (Reuters, 1 June 2023). A Norton Rose Fulbright review adds that the pandemic made the situation worse.
The useful reading for an R&D leader is about timing. The company took on the $1.8 billion acquisition one year before the global ATM count peaked, and the debt then had to be serviced from sales that Reuters describes as flat or declining.
How does the improvement rate see what filings miss?
GetFocus forecasts the improvement rate from two citation measures, and both were visible throughout the period.
- Cycle time is the number of years between one generation of inventions and the next. Both technologies ran at four to seven years, so the two were level on tempo.
- Knowledge flow is how much later work builds on each invention. Online payments climbed past 3, and cash infrastructure stayed under 1.
Think of two workshops that each release a new design every five years. In the first, each design is a tidier version of the last machine. In the second, each design becomes the starting point for many others. Both look equally busy from the street, and the second is moving much faster. Our technology forecasting software page shows how the same reading is produced for other fields.
What are the limits of this reading?
- It is directional. The improvement rate is an estimate from citation patterns and carries measurement error. Read 331% as "far faster than 28%", and treat the exact figure with care.
- It gives a direction and leaves the date open. The gap was more than ten to one in 2004, and ATM numbers peaked in 2017.
- Cash is still in use. More than half of euro-area point-of-sale transactions were in cash in 2024. In this case the faster-improving rival took share and left cash a large market twenty years on.
- The filing counts are GetFocus platform data. The 2015 crossover comes from the two patent sets the backtest compares: cash-payment infrastructure, and online and mobile payments. Drawing either set differently could move the crossover year.
- One company's bankruptcy has its own causes. Reuters and Norton Rose Fulbright point to acquisition debt, declining sales and the pandemic. The rate describes how fast the technology it sold was improving against its rival.
How do you check your own portfolio for the same pattern?
- Take one technical problem and list every technology competing to solve it, including the ones your team has already set aside.
- Read the improvement rate of each, and look at how the readings have moved over the past ten years.
- Compare that with where your industry's effort goes. A gap between effort and rate is the pattern in this case.
- Before a large commitment to the current technology, such as an acquisition or a new plant, write down the rival's rate next to your own.
Our article on tools that help R&D plan for disruptive technology covers where this reading fits beside scenario work. If you have a decision like this on the table, send it to us and get a free 48-hour Decision Brief: a first-pass radar of the competing options, their improvement rates and the thresholds to watch.
FAQ
What is digital payments disruption?
Digital payments disruption is the move of everyday payments from cash and the machines that handle it to online and mobile systems. In GetFocus's backtest, online payments read an improvement rate of 331% a year in 2004 against 28% for cash-payment infrastructure. The global ATM count peaked in 2017 at 3.28 million, according to RBR.
Has cash been replaced by digital payments?
Cash is still widely used. In the euro area it made up 52% of point-of-sale transactions in 2024, down from 72% in 2019, according to the European Central Bank. Online payments rose from 7% to 21% of day-to-day payments over the same years.
Why did Diebold Nixdorf file for Chapter 11?
Diebold Nixdorf filed on 1 June 2023 with over $2.7 billion in debt. Reuters reported that the company had struggled with debt since buying Wincor Nixdorf for $1.8 billion in 2016, while sales of ATMs to banks and checkout machines to retailers were flat or declining. It left Chapter 11 on 11 August 2023.
Can an improvement rate tell you when a disruption will happen?
It gives a direction and leaves the date open. Online payments read more than ten times cash-payment infrastructure in 2004, and the global ATM count peaked thirteen years later.



