Improve innovation with Amazon’s Working Backwards Process
Andy Jassy, Amazon’s CEO, published his first shareholder letter in April, and it was dedicated to how Amazon has created an innovation-based company. What I think is most notable I about the letter is Andy outlined how innovation is a long-term strategy that needs to be intentionally managed by leaders. He even gave 7 specific tips for how to conduct innovation, based on how Amazon has found success.
Since leaving Amazon last year, and driving innovation and expansion projects for Amazon globally, I now train and consult with companies on how to be more innovative, and that’s the same message I try to give: its possible to understand how innovation occurs, and improve it in any company.
Most often I am asked about how to implement Amazon’s Working Backwards process, which is the most public and most significant individual process Amazon has created that enables innovation. Although I help companies with Working Backwards, I also tell them that it takes more than reading about it to be successful, and in fact, most often companies fail when they try to implement it on its own.
The reason for the failure is that that Working Backwards is only one part of how an organization increases customer-focused innovation. The 7 tips Andy shares in his letter are an example of how else you need to build innovation into how a company operates to take full advantage of Working Backwards. For example, if you have a highly risk-averse, consensus-driven organization, switching to a highly iterative, fast, customer-focused process is likely to fail without change management to make the organization more risk-accepting. This article focuses on Working Backwards, and some recommendations to improve it. If your current product development processes are not producing enough success for you, you may benefit from using these recommendations, and use them as part of your overall innovation approach.
The Working Backwards concept itself is simple: every major initiative, investment, or customer-impacting decision first starts with writing a fictional press release before work begins. This ensures that all ideas start with a clear and concisely defined customer benefit, and “work backwards” to what solutions or changes need to occur to support them. This process is the foundation of Amazon’s culture, and is likely the number one reason that Amazon has been so continuously successful in growth and innovation.
What most people don’t know—including many Amazon employees—is that “Working Backwards” as a concept didn’t start with Amazon or Jeff Bezos. In fact, it was Steve Jobs, co-founder of Apple, who back in 1997 outlined the concept in a speech at the Worldwide Developers Conference, saying “You've got to start with the customer experience and work back toward the technology, not the other way around." Some may even argue Peter Drucker was responsible for the idea, when even further back in 1986 he outlined “An enterprise’s purpose begins on the outside with the customer.”
It wasn’t until 2004 when Amazon built their mechanism of the press release and FAQ and titled it, “Working Backwards.” The earliest publication of the process I have found to date is from Werner Vogels, Amazon CTO, back in 2006, where he described it like this: “The product definition process works backwards in the following way: we start by writing the documents we'll need at launch (the press release and the FAQ) and then work towards documents that are closer to the implementation.”
It’s notable that these leaders—possibly the biggest innovators and most successful business leaders of the modern age—agree on this one foundational concept. Amazon believes so strongly in it, it publicizes the approach and teaches it to customers and partners, and employee often consider it the #1 best process they learned at Amazon.
What’s surprising to me, is that with this concept available, and Amazon even providing a great model to follow, most companies ignore it and continue to fail at how they create and execute change for customers. In fact, on average across all businesses, almost every innovation project will fail: in 2019 Deloitte found 96% of innovation projects fail by not providing any return on investment. Start-ups are no better, with 90% expected to fail within 10 years, almost solely due to failure in executing innovation on behalf of the customer.
If you look at Amazon (or Apple) they certainly don’t have 90%+ failure rates. I believe their success stems almost fully from their core belief that customer benefit needs to drive everything they do. Even so, in my almost 9 years leading global technology innovation projects at Amazon, I saw continual failure. I estimate that around 50% of projects fail within Amazon—they fail outright, fail due to goal, organizational, or priority changes, or may launch but provide little or no benefit. Amazon doesn’t necessarily see this as a problem, however, as they understand some amount of failure is required to create innovation. Jeff has recognized this many times, and in 2019 said: “We need big failures if we’re going to move the needle—billion-dollar scale failures.”
Amazon has proven that its innovation processes are strong enough it can overcome its failures: it has delivered a sustained 36% average revenue growth every year for the last 20 years! I credit Amazon’s creation of an innovation-focused organizational culture as enabling them to better deliver on innovation than anyone else—even a 50% failure rate is many times better than 96% failure. Their success is due to their “Working Backwards” process, and any company can begin to incorporate it and improve their own results and growth.
It’s a remarkably simple process: take the time to write a one-page fictional press release before starting work on any new idea. Follow it up with a list of FAQs, which distill the customer problem and proposed solution in more detail, and explicitly raise uncertainties or conflicts for leadership feedback. This allows leadership and all stakeholders to be aligned as to the goal of the work, how it will be accomplished, and resolve any conflicts before work has begun! Readers can find a Working Backwards template with tips on writing it on my website here.

Of course, just writing a fictional press release doesn’t solve all innovation problems and cause all projects to be fully successful. It just helps reduce the risk of the #1 cause of innovation failure: poor customer definition and identification of customer benefit. Once the organization believes it has a compelling customer benefit that is feasible to deliver, and stakeholders are aligned to the idea, it still needs to validate and execute the project well. But starting with the press release is one way to get companies working in what is most likely to be a good direction.
Amazon’s Working Backwards process is not without fault, and as I mentioned many projects still fail. The process is great in that it focuses on the customer, but its biggest opportunity is that it doesn’t include standardized ways to identify or validate that the customer benefit is the right or best benefit that can be provided, or even that customers want it.
It’s an age-old problem: you get a bunch of like-minded people together, and they may agree on what seems like a good idea. Without validating it with customers however, they may spend a lot of time and effort on building something that doesn’t gain adoption in the marketplace (e.g., the Fire Phone). Amazon’s Working Backwards process is also often slow, and waterfall-like, in that it often requires months of discussion and refinement until work is approved to begin. There are clear ways to address both these issues, by incorporating other technology, business, and innovation best practices.
I refer to these improvements as “Working Backwards, v2”.
Amazon’s “Working Backwards” process (what I refer to as “v1”) was created in 2004 and has essentially not changed in the last 18 years. Internally, the primary improvements have been to standardize on questions for the Frequently Asked Questions (FAQ) part of the document that follows the main press release. For example, Amazon Alexa may add required questions that its’ teams need to answer, such as potential impact to customer privacy.
Beyond just changes to the FAQ, I think there are many other ways in which the overall Working Backwards process can be improved to be even more successful and faster. Since leaving Amazon in 2021, I have consulted with companies on how to incorporate Working Backwards, and am in process of documenting, testing, and consulting with companies on how to do it better. I’ve also partnered with Territory Global, a leading Portland-based innovation consultancy, to incorporate aspects of human-centered design and storytelling into the process. I’m happy to share a summary of the biggest improvements here, including Jobs to be Done Theory, popularized by Clayton Christensen, and iterative validation through assumption and hypothesis testing.
The Jobs to be Done Theory (JtbD) reverses the typical approach of product development to focus on what the customer is trying to achieve with a product, instead of what the product itself should do. Think of it this way: there is a “job” that the customer is “hiring” the product to accomplish, and if you want to create a successful product, you need to be very clear not only who that customer is, but what the “job” is they are hiring your product for.
For example, few people purchase a car just to have it sit in their garage. Most often, people purchase (hire) a car to improve something in their life, typically, to improve ease of transportation (the job). If you take a traditional “Product”-centric view of a car purchase, you may consider adding unique product features such as color, speed, seating capacity, or other aspects. You may compare your product to other cars on the market, and make product-differentiating changes to try to retain or attract customers.
Instead, the “Jobs”-centric view expands the opportunity for innovation, by giving you an unconstrained scope to innovate around: improving the ease of transportation. This is what Uber did: it took a jobs view to see how it could better help customers access transportation.
Taking the “Jobs” view can provide a myriad of ways to create new innovation, all of which are better for long-term growth and competitiveness than focusing on product differentiation. In the Uber example, it created a new, innovative business model to get customers quick, easy access to vehicles when they need them, and has now expanded in many different ways for different vehicle types and services. If they set out to just create a new vehicle, they would not have found these new markets.
I believe that incorporating this concept into Working Backwards is the best way to improve it, because it is not only established with almost 20 years of testing and business theory, it reinforces working backwards by providing a consistent structure on how to do it, and the job the customer is trying to achieve. It makes that more effective by expressly defining who the customer is and identifying the “job.” Only after the “job” is defined is a solution proposed to address it.
Consider today’s technology start-ups, or R&D functions within large companies. Many of them focus on interesting new technologies, and hope to identify a potential market or customer at some point in the future. That’s exactly the wrong approach for successful innovation—instead, start with the “job” and then figure out the way for technology to support it.
I think Web3 is a case-in-point. To date, I have yet to see any compelling definition of a “Job” to be done for it. There is a lot of innovation in the space for sure—plenty of new companies, technologies, and ideas being generated, but yet no one has identified the customer job it will serve. As a result, the idea is still nebulous for most people since there isn’t a clear product offering people want to pay for. As my associate Parker Lee from Territory Global likes to say, “It’s an answer looking for a question.”
Similarly, how about the Metaverse? Same thing. The Metaverse concept has been around for ages—remember Avatars on the Xbox 360? Again, no one has yet identified the “Job” for the Metaverse. As a result, it’s only a potential new way to give people exactly what they have today, and for that you have to make it useful enough for them to want to switch. Without a compelling job it will be an uphill battle.
So how do you define a job to be done? That requires some thoughtful work, with a team of stakeholders. There are several books and articles on the subject you can/should read (I suggest starting with this Harvard Business Review article as an overview, and Jim Kalbach’s book Jobs to be Done Playbook.) That said, here’s a quick summary to get you started.
First, when you brainstorm the “Job”, keep it to 3-5 words, in the format of Verb + Objective + Clarifiers (if needed). Continuing the Uber example, a potential job could be: “Transport Myself Inexpensively”. Of course, it can have many other forms, such as “Share Vehicle Cost”. Create a list of potential Jobs with stakeholders, and try to agree on the top 2-3 potential Jobs. Very likely, you’ll find a lot of differing opinions on what you are trying to achieve—that’s a good thing! It helps address those up front rather than 6 months down the road after your product is built.
Second, once you have potential “Jobs” defined, complete the Job Canvas to help identify additional information about your Job and Customer. This step is critical: you will evaluate the Job from all sides: how the customer discovers they even need your product, to the steps necessary to execute it. You may find sub-jobs, such as “Select Transportation Method”, which provide opportunity for innovation in the future (e.g., bicycles, scooters, taxis).
Third, evaluate the emotional and social reasons why someone may want to complete the job (also in the Job Canvas). Is the customer looking to feel a certain way by making the purchase—such as more financially secure, or adventurous? Would they like others to perceive them a particular way for their purchase—do they want to be seen as a cutting-edge technologist to their peers?
Fourth, review the Job Canvas to identify and prioritize the tasks or sub-jobs that are the most beneficial for customers, and then those that are the most feasible for you to improve. This becomes the base of your long-term product roadmap.
Fifth, now write your Working Backwards document, using all the thought and information you have learned, and why you believe the job and solution you propose are the best ones to pursue. Provide your Job Canvas in the FAQ portion of the document for feedback and alignment with others.
Finally, execute it! As you execute, make sure you take an iterative approach. Start with identifying the assumptions you have made and rank them, with the most critical at the top. Then, establish how you can test the assumption, based on a specific hypothesis. Your project should seek to test and validate your hypotheses as quickly as possible, and when issues are found, quickly incorporate learnings and adapt your plan, or pivot to a new plan altogether. Eric Ries’ Lean Startup methodology of Build > Measure > Learn is a great methodology to follow, but a topic for a future post. Here is a spreadsheet template for documenting assumptions and testing plans.
If you are interested in learning how to use Working Backwards and other innovation concepts at your company, or have suggestions, questions, or comments, let me know! You may follow me on LinkedIn, sign up for one of my public or private classes on improving innovation, or contact me directly at trent@d1innovation.com.