Bannerbear: 8 Months at $0, 7 Failures, 2 Pivots — the Complete Record of Reaching $10K MRR

Few indie-hacker records capture failure at this resolution. Jon Yongfook, founder of the image-generation API Bannerbear, has published on his own blog the MRR and tactics of every phase: eight months at $0, a first product that plateaued at $400, a pivot that churned existing customers. Success stories tend to congeal into survivorship bias. This case preserves, intact, an itemized list of “what didn’t work.” That is where the value lies.

The Full MRR Progression (Self-Published)

$10K was a waypoint. According to Starter Story’s tracking, $6K came on October 23, 2020, and $10K arrived exactly two years after the challenge began. Six months later came $20K, late 2021, $25K. 2022, $40K→$45K (by then a full-time remote team of 7), 2023, $50K. And as of May 2024, a reported $51.3K MRR (about $616K in annual revenue, roughly 92 million yen). On the $1M ARR milestone, Yongfook wrote: “Bannerbear got to $1M ARR on Rails 6 and jQuery. You don’t need to be on the cutting edge.”

What the Business Is

Bannerbear is a SaaS that auto-generates social images, e-commerce banners, OG images and more via API. It integrates with Zapier, Airtable, and WordPress to automate image production in marketing workflows. Pricing runs in three tiers: Automate at $49/month, Scale at $149/month, Enterprise at $299/month. Yongfook had spent years in e-commerce operations hand-making visual assets every single day. The pain of that repetition later became his “north-star problem.” To go independent he set aside two years of savings, and burned half of it, nearly a full year’s worth, in the first eight months.

Three Stalls, and How He Escaped Each

Stall 1: Built 7 products with no revenue model (2019). The public challenge of building 12 startups in 12 months drew attention, but MRR stayed at $0 for eight months. His verdict: “none of them gave anyone a reason to pay”, and later, more harshly, that “churning out unrelated products was a fruitless waste of time.” The ability to build and the design of something worth paying for are separate skills.

Stall 2: Picked a problem that didn’t hurt (the $400 wall). Previewmojo (automated OG-image generation) reached the first milestone of $1,000 ARR, and stopped there. Useful, but “not a hair-on-fire problem.” OG images were only a fragment of the deeper problem: automating marketing images in general. It is a case of executing the “niche down” maxim so literally that the market itself vanished. Around the same time he wrote an article titled “don’t charge $9/month for SaaS”, so the quagmire of underpricing was running in parallel too.

Stall 3: The rebrand only grew traffic ($472). Refreshing the look and brand lifted site visits but barely moved revenue. The turning point was a redesign of the product itself, from standalone tool to API. Early customers who had come for OG images churned. A correct pivot carries the pain of losing existing customers, but Starter Story records that the churn trough was quickly refilled by accelerating new signups.

There is one more retreat that rarely surfaces: withdrawing from the Shopify App Store. Traction never came. Beyond API constraints, customers inside the store expected prices in the “$5–10/month” range, a mismatch with what Bannerbear delivers. Every channel carries its own price norms, and no amount of product-side effort will move them. Choosing where you sell is itself your pricing strategy.

Operations During the Growth Phase — “7 Days Code, 7 Days Marketing”

The working method during the $488→$6,109 stretch was simple and explicit.

• Alternate development and marketing week by week, pinning 50% of resources to marketing. Prevent imbalance with a system, not willpower
• Write large volumes of content and API documentation. “The more documentation I wrote, the more conversions went up”
• Publish progress in a weekly newsletter, launch on Product Hunt, post on Indie Hackers, and ship free tools and demos to generate traffic
• Use only technologies he knows well, eliminating the learning cost of anything new

The unexpected lift came from the Zapier integration, which produced organic word-of-mouth inflow from the no-code community.

The final leap ($6K→$10K) came not from features but from positioning. From customer conversations and months of usage-pattern analysis, he sorted customers into two Jobs to Be Done, “those who want to automate via Zapier” and “those who want to mass-generate by calling the API directly”, and reorganized pricing plans, tutorials, and site structure around the words “Automate” and “Scale.” The fact that today’s plan names are still Automate/Scale is the residue of that late-2020 reorganization. Articulating not what the product is, but whose job it does, is what broke the $10K wall.

Lessons and Analysis

“Niche down the target, not the problem”, the biggest discovery in this case. Previewmojo narrowed the problem so far the market disappeared. Bannerbear grew by selling a broad capability, “automated image generation”, to a narrowed target: “businesses that want to automate repetitive marketing work.” It is the mirror image of the lesson in MENTA blossoming after narrowing from “advice on anything” to “programming only”.

The eight $0 months were the sample size, not waste. The 7 failures, run as a public challenge, generated attention and followers that became Bannerbear’s early customers and distribution base. It is the same law of attempts as Levels’s “4 out of 70+” and Irie’s “the 30th product,” compressed here into a single year. Even so, his own retrospective is harsh on the “7 unrelated products,” drawing a sharp line between them and the post-September-2019 iterations “within the single domain of image generation.” Given the same number of attempts, only the ones held inside a fixed domain accumulated as learning.

He corrected the builder’s overbuilding instinct by force of calendar. “7 days code, 7 days marketing” is an excellent mechanism for securing marketing time through structure rather than willpower. Since the single biggest reason indie projects fail is “building without selling,” this is the most portable practice in the whole story.

Documentation is the strongest salesperson an API SaaS has. “The more I wrote, the more I sold” shows that for developer products, content can substitute for sales headcount. As an asset that gets picked up by both search and AI search, its effectiveness has only increased today.

If You Were to Apply This Yourself

The readily reproducible parts: “7 days code, 7 days marketing,” designing the reason to pay first, repositioning via Jobs to Be Done, and building fast on boring technology, all operations executable at individual scale. The preconditions are just as clear: a runway of two years’ savings, the audience already earned through the public challenge, and the depth of the English-speaking developer-API market that powered early acceleration. Also, the post-$10K trajectory ($20K→$51.3K) rests on secondary sources (Starter Story) and cannot be verified first-hand like the founder’s own blog. His closing words describe exactly how to use this record, “find out what works for you, and do more of it.”

Related Cases

• MENTA — the domestic case that blossomed through specialization
• PhotoAI — the probability theory of “just ship more”
• A 50,000-yen-a-month indie app — “what worked and what didn’t” at small scale