In L. Frank Baum’s story, the Wizard of Oz was a regular man behind a curtain, using tricks to create the illusion of power. The audience experienced something magical. The reality was far more mundane.
Your first product should work exactly the same way.
The Wizard of Oz MVP is a product that appears automated and technology-powered from the customer’s perspective, but is actually run by a human behind the scenes. The customer gets the full experience. The founder does the work manually. And nobody needs to know the difference — at least not yet.
This isn’t deception. It’s intelligent sequencing. You’re validating the experience before investing in the technology. Because the most expensive mistake you can make isn’t spending time on manual work — it’s spending months building automation for an experience nobody wants.
Why Manual First Is Smart, Not Lazy
There’s a bias in the startup world toward building technology first. Founders want to automate from day one because automation feels like “real” building. Manual work feels like a shortcut, a compromise, a lesser version.
This bias is expensive and wrong.
Here’s the math: if you build the technology first and the product doesn’t work, you’ve wasted months of development time. If you do it manually first and the product doesn’t work, you’ve wasted a few weeks of labor. The manual approach has 10x lower downside risk.
But the upside is even more compelling. When you deliver a service manually, you learn things that no amount of user research or planning can reveal:
- You discover which parts of the process customers actually care about (and which they don’t notice)
- You find edge cases that would have broken your automation
- You learn the exact language customers use to describe their needs
- You identify where the real value is created (often different from where you expected)
These insights make your eventual technology far better than it would have been if you’d built blind. The manual phase isn’t a workaround — it’s research with revenue.
When I was working with startups at the accelerator, the teams that shipped Wizard of Oz MVPs consistently outperformed teams that built technology-first — in time to market, in customer satisfaction, and in the quality of their eventual automated product. The speed advantage of manual-first is dramatic.
How to Build a Wizard of Oz MVP
The process has five steps.
Step 1: Define the customer experience.
What does the customer see, click, receive, and feel? Map the experience from their perspective, ignoring how it happens behind the scenes.
For example, a customer experience for an AI writing assistant might be:
- Customer submits a topic and some notes
- Within 24 hours, they receive a polished draft article
- They can request revisions
- Final article is delivered within 48 hours
From the customer’s perspective, this could be AI-powered, human-powered, or alien-powered. They don’t care. They care about the inputs (submit topic) and the outputs (receive article). Everything in between is your problem, not theirs.
Step 2: Build the front end.
The part the customer interacts with needs to look legitimate. This doesn’t mean it needs to look beautiful — it needs to look intentional. A clean Typeform, a simple Carrd page, a basic Squarespace site. Something that says “this is a real product” even if the back end is you in your pajamas at your kitchen table.
The front end typically includes:
- A sales page describing the product and price
- A purchase/signup mechanism (Stripe, Gumroad, etc.)
- An input mechanism (form, email, file upload)
- A delivery mechanism (email, shared folder, portal)
All of this can be built in a day using free or cheap tools. No coding required. Building without technical skills is the entire point of this approach.
Step 3: Do the work manually.
When a customer submits a request, you fulfill it. By hand. Every time.
For the writing assistant example: the customer submits a topic, and you personally write the article. Or you use AI tools to help but personally review and edit. The customer receives a quality article. They don’t know (or need to know) that a human did it.
For a data analytics product: the customer connects their data source, and you personally create the analysis in a spreadsheet and format it into a report. The customer receives a professional analysis. The method is invisible.
For a matching service: the customer submits their requirements, and you personally search your database (which might be a spreadsheet) to find matches. The customer receives curated matches.
Step 4: Track everything.
While you’re doing the manual work, document everything:
- How long does each task take?
- Which steps are repetitive and automatable?
- Which steps require human judgment?
- What errors do you make?
- What do customers complain about or ask for?
- What do customers love?
This documentation becomes the specification for your eventual automated system. It’s the most accurate spec you’ll ever write because it’s based on real operations, not assumptions.
Step 5: Automate incrementally.
Once you have 10-20 customers and a clear picture of the workflow, start automating the most time-consuming and repetitive steps. Not everything — just the bottlenecks.
Maybe the first automation is an intake form that automatically creates a task in your project management tool. Then maybe it’s a template system that pre-fills 60% of each deliverable. Then maybe it’s an API integration that pulls data automatically instead of you copying it manually.
Each automation frees up your time to serve more customers. You scale by automating the work, not by hiring more people (though hiring is also valid at some point).
Real Wizard of Oz Examples
Zappos (online shoe store): The founder photographed shoes at local stores, listed them online, and when someone ordered, he bought the shoes at the store and shipped them. No inventory. No warehouse. No supply chain. Just a human running between a computer and a shoe store.
This proved people would buy shoes online. Then they built the infrastructure.
Food on the Table (meal planning app): The founder personally met with the first customer, went through her pantry, checked local grocery store sales, and created a weekly meal plan by hand. Every week. For one customer. Then he did it for two. Then five.
Each manual iteration taught him exactly what the eventual app should do. The manual phase was the product spec.
A startup I advised (HR compliance tool): The founders told employers they had a “platform” that tracked employee compliance requirements. Behind the curtain, one founder manually monitored regulatory websites every morning and sent personalized email alerts. For six months.
Those six months produced: 12 paying customers (proving demand), a detailed understanding of which regulations employers cared about most (proving product direction), and enough operational data to write a software spec that was built in 8 weeks and worked perfectly on launch.
When to Stop Being the Wizard
The Wizard of Oz phase isn’t meant to last forever. Here are the signals that it’s time to automate.
Signal 1: You’re spending more time on operations than on growth.
If you’re working 40 hours a week just to serve existing customers and have zero time for new customer acquisition, the manual approach has hit its ceiling. Automation frees up your time for growth.
Signal 2: Quality is suffering because of volume.
When manual delivery starts producing errors — missed deadlines, inconsistent quality, forgotten requests — it’s time to automate the error-prone steps. Your customers are paying for a reliable experience, and the core outcome must always work.
Signal 3: You’ve served 20+ customers and the patterns are clear.
After 20+ manual deliveries, you know the workflow intimately. You know which steps are identical every time (automate these) and which require judgment (keep these human). This is enough data to build an automation that actually works.
Signal 4: Customers are asking about scale features.
If customers start asking “Can I use this for multiple projects?” or “Can my team access this?” they’re signaling that they see long-term value and want the product to grow with them. That’s the time to invest in technology.
Don’t automate before these signals appear. Premature automation is building infrastructure for a business that might not exist.
The Ethics of the Curtain
Let me address the uncomfortable question: is it ethical to let customers believe there’s technology when there isn’t?
My answer: yes, with conditions.
Condition 1: The customer gets the promised outcome. If your marketing says “receive a data analysis within 24 hours” and they receive a data analysis within 24 hours, the promise is kept. The method of delivery is your operational detail, not a marketing claim.
Condition 2: You don’t claim specific technology you don’t have. If your landing page says “powered by our proprietary AI algorithm,” and there is no algorithm, that’s a lie. But if your landing page says “we analyze your data and deliver actionable insights” and you do that manually, that’s accurate.
Condition 3: Quality matches or exceeds what automation would provide. Manual delivery by a knowledgeable human often produces better results than early-stage automation. If the manual version is superior, the customer is getting more value, not less.
Condition 4: You’re building toward automation. The Wizard of Oz approach is a stage, not a permanent model. If you never intend to automate and plan to forever pretend to have technology, that crosses into dishonesty. The manual phase should have a clear endpoint.
Within these conditions, the Wizard of Oz MVP is not only ethical — it’s the most customer-centric approach to product development. The customer gets a working product faster. The eventual automated product is better because it’s informed by real operations. Everyone wins.
Combining Wizard of Oz with Other Approaches
The Wizard of Oz MVP combines well with several other strategies:
With pre-selling: Pre-sell the experience, then deliver it manually. Pre-selling validates demand. Manual delivery validates the experience. Together, they validate the business.
With the MVE approach: The Wizard of Oz MVP is essentially a specific implementation of the Minimum Viable Experience. The experience is real and complete. The implementation is manual and minimal.
With niche dominance: Manual delivery is only feasible for a small number of customers, which naturally forces niche focus. This is a feature: you deeply understand a specific customer segment before broadening.
Key Takeaways
- The Wizard of Oz MVP looks automated from the outside but is manual behind the curtain. The customer gets the full experience. You do the work by hand.
- Manual first is smarter, not lazier. You learn what to automate, discover edge cases, and validate the experience with 10x lower risk than building technology first.
- Build the front end with no-code tools, do the work manually, track everything, then automate the bottlenecks. Each step informs the next.
- Automate when you’re spending more time on operations than growth, quality is suffering, you’ve served 20+ customers, or customers ask for scale features.
- The approach is ethical as long as the customer gets the promised outcome, you don’t claim specific technology you don’t have, and you’re genuinely building toward automation.
