MVP vs MLP vs MVA: Minimum Viable, Lovable, or Awesome?
Three acronyms that get used interchangeably, inconsistently, and often wrongly. MVP (Minimum Viable Product), MLP (Minimum Lovable Product), and MVA (Minimum Viable Awesome, sometimes Minimum Awesome Product). Every agency blog has a comparison table. Most get the history wrong and all skip the question that actually matters: in 2026, when building is nearly free, what’s the smallest plausibly shippable product worth committing a team to?
The short answer is that “minimum” is itself a suspect framing now. When the cost of building more was real money, minimising it was smart. When it’s nearly free, the optimisation target shifts — from how little can we ship? to what’s magnificent enough that a pragmatist customer will actually switch to us? And that reframes MVP, MLP, and MVA as three different answers to three different questions, not three points on a spectrum.
This article is a practical comparison, a correction of the common attribution errors, and — more importantly — an argument about which of the three to reach for in different circumstances once you understand what each was actually designed for.
MVP (Minimum Viable Product) was coined by Frank Robinson in 2001 and popularised by Eric Ries in The Lean Startup (2011). It is a learning vehicle — the smallest version of a product that can test the core hypothesis. MLP (Minimum Lovable Product), coined by Brian de Haaff of Aha! in 2013, adds an emotional bar — the smallest product users will love, not merely tolerate. MVA (Minimum Viable Awesome) argues that in an AI-era market, “minimum” is the wrong ambition — the correct target is magnificent in at least one dimension (your “crown jewel”), even if other dimensions remain rough. The three are answers to different questions: MVP asks will anyone use it?; MLP asks will anyone love it?; MVA asks will anyone switch for it?.
TL;DR: I was working with a client whose absolute beating-heart feature — the report that closed their deals and retained their accounts, the thing that if you took it away the entire product would cease to have a reason to exist — had fallen behind their competitors. The sales team was losing deals on it. Everybody knew it was the problem. So the team set about building a new version. They called it an MVP.
What they built was a Minimum Product. Not a viable one. A technically-functional one that handled the happy path, looked like a developer had designed it (because a developer had designed it, at 4pm on a Friday), and did nothing to reclaim the ground they’d lost. The team had been so deeply conditioned into doing the bare minimum that they applied the same approach to the single most important feature in their entire product. There’s a whole separate article on the culture of adequacy that this story sits in; the short version is that the same team, given explicit permission to build something magnificent, took customer satisfaction on that feature from 3/10 to 8/10 in twelve months. Same people. Same skills. Different ambition bar.
This is why I don’t find the MVP framing useful any more for anything that isn’t purely a learning experiment. It has been diluted into “the minimum we can get away with” and it produces minimum products optimised for being shippable, not for being chosen. The 2026 target for a launched product is almost always MVA — magnificent in one dimension, rough elsewhere — and the only defensible place for an MVP in the original Ries sense is as a learning artefact, which is almost always better named a Riskiest Assumption Test .
A Bit of History (Getting the Attributions Right)
Most comparison articles credit Eric Ries with inventing the MVP. He didn’t. Frank Robinson, CEO of SyncDev, coined “Minimum Viable Product” in 2001. Ries popularised the term in The Lean Startup (2011), and the popularisation is genuinely important — Ries gave the MVP its build-measure-learn loop and its economic rationale — but the invention was ten years earlier.
MLP (Minimum Lovable Product) was coined by Brian de Haaff, co-founder of Aha!, in 2013, and expanded in his 2016 book Lovability. De Haaff’s argument was that MVPs got watered down into ugly minimum products that nobody loved and therefore nobody became a word-of-mouth advocate for. Aim for minimum lovable instead: what’s the smallest product with enough design polish and emotional resonance that users will actually recommend it?
MVA / MAP (Minimum Viable Awesome / Minimum Awesome Product) has no clean canonical attribution. Carlos Beneyto’s 2018 Medium piece “The MVP Is Dead, Long Life to the MAP” is the most-cited source, but he doesn’t claim invention. The term has been circulating loosely since around 2017. It’s genuinely fuzzy — some articles use “MVA” to mean Minimum Viable Audience or Minimum Viable Architecture. For this article I’ll use MVA to mean “minimum awesome product” — small in scope, magnificent in at least one dimension, rough elsewhere.
A related attribution worth correcting: Henrik Kniberg’s skateboard→bike→motorbike→car illustration (from his 2016 “Making Sense of MVP” Crisp blog post) is widely cited as illustrating the MLP idea. It doesn’t. Kniberg was illustrating iterative MVP delivery — you give the user a skateboard so they have transport now, then upgrade to a bike, then to a car, always shipping something usable. It’s a rebuttal to the car-delivered-in-pieces anti-pattern, not an argument about lovability. The illustration gets mis-cited constantly.
MVP, MLP, and MVA Compared
| Dimension | MVP | MLP | MVA |
|---|---|---|---|
| Primary goal | Learn (will anyone use it?) | Attach (will anyone love it?) | Differentiate (will anyone switch?) |
| Economic premise | Building is expensive; minimise to maximise learning | Polish matters; love creates word-of-mouth | Build cost is low; minimum-everywhere is now undifferentiated |
| Success metric | Validated learning, retention cohort direction | NPS, organic advocacy, emotional retention | Narrow but decisive competitive advantage in chosen dimension |
| Breadth | Narrow | Narrow | Narrow |
| Polish | Low — rough on purpose | High across the whole experience | Magnificent in one “crown jewel” dimension; rough elsewhere |
| Best suited to | Pre-PMF testing, cheap-to-build hypotheses | Consumer products where brand and delight matter | B2B products in crowded, AI-commoditised categories |
| Risk | Users may dismiss it before learning anything | Misallocating polish budget across the whole product | Pragmatist buyers may care about the roughness elsewhere |
| Canonical example | Dropbox 2007 explainer video | Instagram (focused filter experience) | Stripe’s developer experience; Linear’s UI |
Each one answers a different question for a different moment in the lifecycle. None is universally right.
When to Reach for MVP
An MVP is right when the binding risk is will anyone use this at all?. You don’t know if the problem is real enough. You haven’t done problem-solution fit work yet, or you’ve done some and still aren’t sure. The cost of being wrong is wasted build, and the cost of being right is a validated hypothesis and the chance to do the polishing afterwards.
Classic examples — Dropbox’s 2007 explainer video, the Airbnb air-mattress rental, Zappos’s hand-photographed shoes — are all MVPs in Frank Robinson’s original sense: evidence-generating activities, not polished products. They are roughly the same as the Riskiest Assumption Test approach, and in 2026 I’d argue the RAT vocabulary is sharper than the MVP vocabulary for this use case. An MVP in the Lean Startup sense has been diluted by years of misuse into “a rough first product”; a RAT is explicitly an experiment.
When to Reach for MLP
An MLP is right when you know the problem is real and the question shifts to will users stick around and tell their friends?. You’ve done your Mom Test interviews, the problem is validated, a prototype would solve it — but a prototype users merely tolerate produces no word-of-mouth, no organic growth, and no virality. For consumer products in particular, the emotional bar is higher than the functional bar, and MLP is the lens that keeps the team focused on that.
De Haaff’s argument in Lovability is worth taking seriously for consumer products: the difference between 10% retention and 40% retention is usually not more features, it’s emotional investment in the first few sessions. An MLP concentrates polish in the moments that matter — the onboarding flow, the core loop, the first magic moment — at the expense of the long tail of features.
When to Reach for MVA — The Crown Jewels Argument
Here’s where the 2026 reframe matters most. The classic MVP/MLP framings assume your product is being compared to alternatives that don’t exist yet or are themselves rough. That’s mostly no longer true. Your pragmatist prospect has a dozen AI-built alternatives — each one a competent MVP — in their inbox this week. Being a thirteenth competent MVP buys you nothing.
What buys you something is being magnificent in at least one dimension — a crown jewel capability that’s visibly, palpably better than the alternatives. Stripe’s developer experience. Linear’s keyboard navigation. Notion’s block editor. Superhuman’s speed. Each one of these products is rough in dimensions other than their crown jewel, and each one succeeded because the crown jewel was so decisive that prospects didn’t mind the roughness elsewhere.
This is the crown jewels thesis applied at the early stage, and it pairs with the culture of adequacy article — “your customers don’t want minimum, they want magnificent”. MVA says: pick the single dimension where you can be magnificent, concentrate your engineering and design capacity there, accept roughness everywhere else, and ship. The decisive differentiator is what converts pragmatists; minimum-everywhere doesn’t.
In practice MVA looks like:
- One feature or capability that is better than any alternative in the market
- The rest of the product intentionally rough but functional enough not to block usage
- Design investment disproportionately concentrated on the crown-jewel workflow
- Marketing built around that one dimension, not around breadth
This is also the answer to the ship-it-and-move-on failure mode — the pattern where a team ships a feature at minimum viability and immediately moves to the next one, producing a product that’s “minimum of everything, magnificent of nothing”.
The 2026 Reframe: “Minimum” Was a Pre-AI Optimisation
The MVP was designed for an era where building was the dominant cost. Minimising scope was rational because minimising scope minimised build cost. The whole point of the MVP as Ries described it was to stretch limited budget across the maximum number of learning loops.
In 2026 that economic equation has inverted. Build cost is low enough that “minimise scope” is no longer the dominant optimisation. What you’re minimising against is the pragmatist buyer’s attention — and their attention is being flooded with equally minimum products. Minimum-everywhere, in 2026, is a commodity. Magnificent-in-one-dimension is a differentiator.
This is the through-line of the whole early-stage validation cluster . AI collapsed build cost; sell cost (attention, trust, distribution) is unchanged. MVP optimises against the thing that no longer matters; MVA optimises against the thing that does. MLP sits in between, optimising against an emotional-engagement bar that matters for consumer products but often not for B2B.
The practical consequence: for most 2026 product bets, MVA is the right target after initial assumption testing has cleared the biggest risks. Ship magnificent in one dimension. Accept roughness elsewhere. Use the roughness as the roadmap for quarters two and three. Don’t ship a rounded product that’s merely competent everywhere and magnificent nowhere — that’s a recipe for the culture of adequacy failure mode.
Cagan’s Critique of MVP
Marty Cagan has been a consistent critic of the way MVP is used in practice. His version — articulated in Transformed and elsewhere — is that MVP’s original learning-vehicle purpose got corrupted into ship ugly early versions and iterate, which has caused enormous damage to product quality expectations. Cagan prefers a sharper distinction: prototypes (discovery artefacts, not shipped to customers) and products (shipped with enough quality that customers can actually adopt and stay).
This aligns with the MVA framing. A Cagan-aligned empowered team doesn’t ship a minimum product and iterate on customer complaints; it runs discovery experiments (prototypes, RATs ) before shipping, then ships something good enough that adoption happens on the first try. The culture-of-adequacy failure mode is what happens when the MVP language pushes teams to ship too early.
Products, Not Companies: MVP Choices Per Product
Every new product a mature company launches forces the same MVP-vs-MLP-vs-MVA decision. The company’s existing product being high-quality tells you nothing about the appropriate launch shape for the new product. A Series C SaaS company launching a new product line may run an MVP (if the problem isn’t validated), an MLP (if it’s a consumer-like extension), or an MVA (if it’s entering a crowded B2B category) — the choice is per-product, not per-company.
Worse, mature companies often default to the MLP or MVA bar for new products because that’s what their existing products are held to, and overspend on polish for a new product whose problem isn’t yet validated. That’s an expensive mistake — the right sequence is still RAT → MVP (for learning) → MVA (for launch) even inside mature companies. The lifecycle is per-product, not per-parent-company’s reputation.
The Side-of-Desk Anti-Pattern
An MVA built side-of-desk is not an MVA. Magnificent-in-one-dimension requires concentrated engineering and design effort. A team that’s splitting its attention with Run work simply cannot produce a crown-jewel capability — they’ll produce an average capability in the one dimension and still-average capabilities everywhere else, which is the worst of both worlds.
The fix, as this blog has argued repeatedly: a dedicated minimum viable team — two engineers and a product person — with a proper business case , protected capacity, and an explicit remit to concentrate on the crown jewel. The measurement is outcomes, not features (see outcome-based roadmaps ), and the protection mechanism is WIP limits and resisting priority whiplash . A dedicated team can ship a crown jewel. A side-of-desk team will ship a beige product.
The PE / NED Diagnostic: Reading an MVP Plan from the Board Seat
When a portfolio CEO presents a product plan labelled “MVP”, the board question is always the same: what specifically is this a minimum of?. The honest answers sort into three bins:
- “It’s a minimum of learning” — this is an MVP in Robinson’s / Ries’s original sense, ideally a RAT . Fund it cheaply and quickly, kill it quickly if the learning says kill. This is a pre-PSF activity, not a launch.
- “It’s a minimum of features, polished across the board” — this is an MLP. Suitable for consumer products pre-launch. Ask hard questions about whether the emotional hook is sharp enough; polish alone doesn’t differentiate in 2026.
- “It’s a minimum viable magnificent thing in one dimension” — this is an MVA. Fund it well. This is the likely winner in crowded B2B categories. Ask: which is the crown-jewel dimension, and how do we know that’s the one that matters to the pragmatist buyer?
If the CEO says “MVP” and means “we’re launching a broad but rough product”, the board question is: against what AI-built competitor is this going to win, and on what dimension?. If the answer is vague, the product will be undifferentiated and the investment is doomed. Kill it or redirect it to MVA.
How RoadmapOne Helps
RoadmapOne lets you set an objective for the team as “ship a crown jewel in dimension X by end of Q3” rather than a feature list (see objectives to key results and OKRs for product teams ). The grid makes visible whether that team is actually concentrated on the crown jewel or whether their capacity is being diluted across Run work. Tagged as a Transform objective (see Run / Grow / Transform ), the analytics tell the board what percentage of engineering capacity is on crown-jewel work — often much less than the plan claims.
Frequently Asked Questions
What is the difference between MVP, MLP, and MVA?
MVP (Minimum Viable Product, Frank Robinson 2001) is a learning vehicle — the smallest version of a product that tests the core hypothesis. MLP (Minimum Lovable Product, Brian de Haaff 2013) adds an emotional bar — the smallest product users will love, not merely tolerate — and is especially relevant for consumer products. MVA (Minimum Viable Awesome) argues that in crowded 2026 markets, minimum is the wrong target — the correct target is magnificent in at least one dimension (the product’s crown jewel ), even if other dimensions remain rough.
Who invented the Minimum Viable Product?
Frank Robinson of SyncDev coined the term in 2001. Eric Ries popularised it in The Lean Startup (2011) and gave it the build-measure-learn loop, but he did not invent it. Most comparison articles mis-attribute the term to Ries. The Ries contribution is the method — how to use an MVP — not the name.
What is a Minimum Lovable Product?
Brian de Haaff of Aha! coined MLP in 2013 and expanded it in his 2016 book Lovability. The central argument is that MVPs got diluted into “minimum everything”, which produced products nobody loved enough to recommend. An MLP is the smallest product with enough design polish and emotional resonance that users will actively advocate for it. Instagram’s focused filter experience is the canonical example — narrow scope, high polish, decisive emotional hook.
Is MVP dead in the AI era?
The MVP-as-learning-vehicle concept is alive and important, but in 2026 it overlaps heavily with the Riskiest Assumption Test — which is the sharper vocabulary for the same idea. The MVP-as-rough-first-product concept (the corrupted version) is increasingly bankrupt, because in a market full of AI-built rough first products, another rough first product doesn’t differentiate. For launch shape, MVA is usually the better target. For pre-launch learning, RAT is the sharper name.
When should I build an MVP vs an MLP vs an MVA?
MVP / RAT for pre-problem-solution fit learning — the question is will anyone want this?. MLP for consumer products launching into markets where emotional engagement drives retention and word-of-mouth — the question is will users love it enough to tell their friends?. MVA for B2B or crowded categories where you need a decisive differentiator against a sea of AI-built competitors — the question is will pragmatists switch to us for this specific capability?. The sequence for most 2026 product bets is RAT → (more RATs) → MVA, skipping the traditional MVP launch shape entirely.
Do mature companies still need MVPs?
Yes, per-product. A mature company launching a new product line still faces the same choices. Importantly, the existing product’s quality reputation doesn’t transfer — if the new product ships at MVP-level quality, customers will compare it to the parent company’s flagship, not to the start-ups the MVP concept assumed. That’s why most mature companies should target MVA or MLP for new-product launches even when they’d be happy launching a pure MVP inside a pre-seed start-up.
Conclusion
MVP, MLP, and MVA answer different questions for different moments, and conflating them — as most comparison articles do — leads to poor launch decisions. Frank Robinson’s MVP was a learning vehicle. Brian de Haaff’s MLP added emotional polish. MVA argues that in 2026’s crowded, AI-commoditised markets, being minimum in every dimension is a commodity, and the only defensible target is magnificent in at least one — your crown jewel .
The practical rule for most 2026 product bets: do your learning through RATs and Mom Test interviews , not through a rough public launch. When you’re ready to ship, ship an MVA — narrow scope, one magnificent dimension, rough elsewhere — rather than a beige MVP. Reserve MLP for consumer products where emotional engagement drives retention. And always, always, allocate a dedicated team (two engineers, one product person) to do the work — side-of-desk will produce a beige MVP every time, no matter what your plan said at the start.
Stop shipping minimum of everything. Start shipping magnificent of something. That’s the only reliable differentiator left.
Baxter image prompt (photorealistic, 4:3): Baxter the wirehaired dachshund as a master jeweller in a worn leather apron, seated at a precision workbench under a single adjustable lamp, setting a single flawless emerald into a gold ring band. On the bench beside him, three rough pieces of metal labelled “good enough” pushed to one side. His loupe pushed up onto his forehead. Close-focus on the emerald, the rest of the workshop blurred into warm bokeh. Craftsman-like, no showiness — quiet pride in one thing done magnificently.