What should an early-stage startup focus on first: product, sales or marketing?

Is it product first, sales second, marketing third?

No. At an early-stage startup, product, sales and marketing are not three departments in a queue but one learning loop around one buyer’s problem. What changes over time is the weight: selling by hand starts in week one, and marketing gets a real share only after the same kind of buyer keeps paying and using.

Marc Andreessen writes that “The only thing that matters is getting to product/market fit” and defines it as “being in a good market with a product that can satisfy that market.” Fit is a property of product and market together, so you cannot reach it by building alone. Paul Graham puts the consequence bluntly: “at least one founder (usually the CEO) will have to spend a lot of time on sales and marketing.”

So ask a different question: which piece of evidence am I missing right now? Each function produces a different kind.

Function	What it means before fit	The evidence it produces	What it is not yet
Product	The smallest version that delivers one outcome for one buyer	Whether people who use the core reach the outcome and come back	A roadmap built from every request
Sales	The founder asking real buyers to commit time, data or payment	Whether strangers pay, and why the wrong-fit ones say no	A team, a quota or a sequence tool
Marketing	Positioning: who it is for, who it is not for, and what changes for them, in buyers’ words	Whether a stranger understands the offer without you in the room	Campaigns, ads or a content calendar

Each strengthens the other two. A sales call shows which product gap matters; a product change gives the next call something new to test; the words that closed the deal become the page. Treating them as separate phases breaks that circuit.

Illustration: product, selling and message travel as one loop between the founder and the buyer, not as three steps in a row.

How early should a founder start selling?

A founder should start selling before the product is finished: talk about the problem first, then ask for a commitment as soon as one buyer can use a narrow version. Waiting for a complete product delays the only evidence that tells you what complete means.

Graham calls manual recruiting “The most common unscalable thing founders have to do at the start” and adds: “You can’t wait for users to come to you. You have to go out and get them.” He treats a big launch or a big partner, used instead of that effort, as “ipso facto suspect.” Euclid Ventures goes further: “the best founders sell well in advance of product.”

Selling this early is mostly qualification, and qualification works in both directions. You are checking whether this buyer has the problem badly enough to change something, and the buyer is checking whether you understand it. Say plainly what exists and what does not. Ask before you send a deck or book a demo. When the buyer needs something you will not build, say no; a customer who does not fit pulls the product toward a market you did not choose and teaches you the wrong lesson. Outbound designed for mutual fit covers how to reach out on that basis.

Bessemer quotes a revenue leader: “Founders should be closing deals at the early stage, and at least one deal with a total skeptic.” A skeptic who buys tells you more than ten friends who say it looks great.

What evidence tells you to shift weight from product to sales to marketing?

Shift weight when a specific piece of evidence appears, not when a calendar date arrives. The shift is gradual: the founder keeps selling throughout, and each new stage adds work rather than replacing it.

Stage	Where most founder time goes	Evidence to move on	Hold off on
1. Problem	Conversations with one narrow buyer about how they handle the problem today; a first usable slice	Several buyers describe the same problem and workaround unprompted, and some agree to try it on real work	Launches, ads, a brand project, hires
2. First customers	Selling and onboarding by hand; fixing whatever blocks the outcome	People who did not know you before pay and use the core; recent users would be very disappointed to lose it	Paid channels, a second segment
3. Repeat in one segment	The same pitch closing similar buyers; writing the path down; a page in buyers’ words	Similar buyers close for the same reason and keep using the product without chasing	A sales leader, broad campaigns
4. Repeatable beyond the founder	A first hire who does the work, one channel with a stop rule	Someone else follows the written path and gets comparable results	Many channels at once

Use these published markers as directions, not laws:

1. The disappointment test. First Round Review describes Sean Ellis asking users “how would you feel if you could no longer use the product?” After benchmarking nearly a hundred startups, Ellis found that companies with strong traction almost always passed 40% “very disappointed.” Superhuman started at 22% and climbed by narrowing to the people who loved it most.
2. Who the first customers are. Bessemer’s playbook assigns founder-led sales to the stage of landing the first 10 to 20 customers via cold outreach, and asks that “a significant percentage” of them come from cold outreach rather than existing relationships.
3. Repeatability before handover. Mercury advises founders to “Lead sales yourself until you’ve found repeatability.”
4. Levels, not a switch. Euclid describes five levels from pre-revenue to scaling and says founder-led selling “should be the core activity” at the first one. Even at the later levels, the founder is “still selling.”

A stall is a product problem when buyers in your segment use the core and still miss the outcome, or when fewer than 40% of recent users would be very disappointed to lose it. It is a go-to-market problem when users reach the outcome and would miss the product, but you cannot find or convert more buyers like them. Calls that end in compliments and never in a next step with a date, data or payment point first at the pitch or the segment, not the build. For the full diagnosis, read how to tell a GTM problem from a product problem before you move to the next row.

How should a founder split the week between product, sales and marketing?

No published study measures the right split of a founder’s week, so do not start from a fixed number of days. Write the split you plan into the record below, and at every two-week review move time by what the record and the stage table show:

What the review shows	Move time	Toward
Conversations held stay below the number planned	From product to selling	More conversations with the same narrow buyer
Buyers recognise the problem but do not pay or use the core	From selling to product	The smallest version that delivers the one outcome
Users reach the outcome, but a stranger does not understand the offer without you	From product to positioning	One page that says who it is for, in buyers’ words
Similar buyers close for the same reason and keep using without chasing	To positioning and marketing	Page, proof, one public note per week from a real objection
Someone else follows the written path with comparable results	From selling to coaching	The first hire, key deals, one channel with an owner and a stop rule

Two rules hold in every row: selling never drops to zero, and until similar buyers repeat, positioning stays a short task of writing down the buyer, the problem and the words they use.

Two founders can split these lines between them, as long as both hear buyers directly. A technical founder who never joins a call loses the information that should steer the build.

Before each two-week block, fill in a short record. It forces the question of which evidence is missing:

Buyer we are learning from (role, company type, trigger):
Who we will refuse, even if they want to buy:
The outcome they should reach, and how we will see it:
Evidence we have (paying strangers, repeat use, disappointment answers):
Evidence we are missing:
The one change this block (product, pitch or segment):
Founder days on product / selling / positioning:
Conversations planned and held:
What would make us stop or narrow:
Review date:

What does marketing mean at pre-seed and seed?

At pre-seed and seed, marketing mostly means positioning: deciding who the product is for, who it is not for, and saying what changes for them in words buyers already use. Campaigns come later, because marketing amplifies what is already true in the product and in the customer’s result; it cannot create that truth.

That makes early marketing narrow and cheap in time: one sentence a stranger can repeat, a page that says it, and notes that answer the objections you hear on calls. A clear homepage message for a seed startup is often the highest-value marketing task in the first months, because every buyer you talk to will check it.

Loud marketing before fit fails in a specific way. Startup Genome lists “Overcompensating missing product/market fit with marketing and press” as one way startups scale the customer dimension too early. Lenny Rachitsky, reviewing how fast-growing B2B companies found their first customers, concludes that “Getting press is rarely the way to get started.”

There is an honest counterview. ChartMogul notes that at pre-seed and seed “many companies choose to start with outbound sales motions,” but argues that content, social and “even small paid ad tests, can generate valuable data on your ideal customer profile (ICP).” That holds when the test is small and you read it as research. It does not hold when spend replaces conversations. Before you give marketing more weight, check what to validate before scaling marketing at seed.

How do you tell real demand from a false product-market fit signal?

Real demand is a stranger in your target segment paying and using the product for the outcome you promised; most other early signals are weaker than they look. The most common false signal is a buyer who bought because they trust you, not because they need the product.

Three sources warn about the same trap. Bessemer: “Relying on familiar connections can create false signals of product-market fit since these buyers already trust you.” Mercury quotes a founder on the first hundred customers: “you cannot count your friends in that.” Euclid names the pattern a “product-market fit mirage”, where a few initial customers lean in but do not represent broader demand. The trap is hard to avoid because the network is where first buyers often come from: Lenny’s review puts “Tap your personal network” first among the strategies that worked, and quotes Workday’s former CEO Dave Duffield on its earliest customers, midsize-company CEOs who knew the founders: “They bought the friendship, not the software.” Start there if that is where your buyers are, and test those buyers harder on whether they would pay for the product itself.

Signal	Why it misleads	What counts instead
Friends, former colleagues, investors’ portfolio	They buy the relationship	A comparable buyer who did not know you
Compliments after a demo	Costs the buyer nothing	A next step with a date, data or payment
A free pilot with no owner	Nobody is accountable for the outcome	A pilot with a named user, a success event and an end date
A deep discount to close	Tests price sensitivity, not need	A buyer who pays a price you can sustain
Waitlist sign-ups	Interest without commitment	Sign-ups who agree to a call and then use the product
Many feature requests	Often from people who would not miss the product	Requests from users who would be very disappointed without it

None of the weak signals is worthless. Friends are a good first place to learn setup and onboarding. Just keep them in a separate column so they do not inflate the count.

When should you hire your first salesperson or marketer?

Hire your first salesperson or marketer for a gap you can point at, after the founder has a path someone else can follow; never hire to discover whether the product sells. The first hire should do the work, not manage people who do not exist yet.

Bessemer is direct: “Don’t hire a VP of Sales too early,” because that skill set suits scaling an existing operation, not building one from scratch. Startup Genome counts “hiring managers (VPs, product managers, etc.) instead of doers” as premature scaling, and found the teams of prematurely scaling startups 3 times bigger than consistent startups at the same stage. When a sales leader does come, Euclid advises hiring “the right person for the next 18 months.”

The sequence of hires, and the tests that say a seat is ready, is worked out in running GTM before hiring a full team.

How does the answer change for product-led, sales-led and other models?

The product, sales and marketing loop stays the same across business models, but the place where you learn from buyers moves. In a product-led model the product itself does much of the selling; in a sales-led model the founder’s calls do; in a marketplace you sell to two sides at once.

Model	Where the founder learns	Early marketing looks like	Watch out for
Sales-led B2B	Calls with a narrow buyer group	Positioning, one page, notes from objections	Hiring sellers before the pitch repeats
Product-led B2B	Activation and usage data, plus calls with active users	Onboarding copy and the first screen	Reading sign-ups as demand
Marketplace	Both sides, in one small area	Filling the thin side by hand	Growing one side the other cannot serve
Long technical cycle	Design partners and pilots	Clear explanations for evaluators	Pilots without a success event

Sources support the first three rows. ChartMogul notes that in self-serve or product-led companies “there may not be any sales reps at all.” Bessemer writes that “a seed-stage SaaS or AI startup with a PLG sales motion may not need a traditional sales team at all,” and that “Marketplaces often start with a small, well-balanced supply and demand group before expanding.” The fourth row is my own reasoning; I found no source that measures it.

Which sequencing mistakes hurt early startups most?

The costliest sequencing mistake is scaling something before the evidence for it exists: spend before fit, managers before a repeatable pitch, features before a proven problem. Running out of money is usually where that story ends, not where it starts.

CB Insights analysed 431 VC-backed startups that shut down since 2023. Running out of capital tops the list at 70%, “but it’s almost always the final cause of death, not the root problem.” Poor product-market fit appears in 43%, and many companies cited more than one reason. Startup Genome, with an older dataset from 2011 of over 3,200 high-growth tech startups, found premature scaling in 70% of them and reported that no startup that scaled prematurely passed the 100,000 user mark. It counts five dimensions that can run ahead of reality: customer, product, team, financials and business model.

Beyond the hiring and false-signal traps above, watch for:

1. Raising marketing volume before the people who already use the product would miss it.
2. Building features to avoid another uncomfortable sales call.
3. Taking every deal, including buyers who need a different product.
4. Running three channels at once, so none of them produces a readable result.

What can AI take over in the first months, and what not?

AI can take over research, preparation and record keeping around early selling; it cannot hold the conversation in which a buyer tells you what the problem costs them. Keep the judgement about fit, price and promises with a person.

Bessemer warns against using AI for generic mass email and says “The real power of AI lies in sales research and personalized outreach.” In practice: account research before a call, a follow-up draft, call notes summarised into your CRM, and a weekly view of repeating objections. Check summaries against what the buyer said, approve every message a buyer reads, and spend the freed time on more conversations, not more volume.

When is this advice wrong for you?

Putting product and founder-led selling ahead of marketing fits a B2B startup with a first version and a handful of users. The order needs adjusting when your situation differs in one of these ways:

1. You already have strong pull. If buyers arrive faster than you can serve them and keep using the product, your bottleneck is capacity, not evidence.
2. Nobody can use anything yet. With no usable slice at all, the week is mostly problem conversations; selling a commitment comes once one buyer can try something real.
3. Regulated or long-cycle buyers. Evidence arrives in months, not weeks. Set milestones such as a signed pilot with a success event rather than a paid close.
4. The buyer is a consumer. Marketing tests may give you evidence earlier than calls can, as long as you measure return use and not only sign-ups.

For the full first plan, from buyer filter to stop rule, see building a first go-to-market strategy at seed.

How I’d approach this with a founder

With a founder deciding between product, sales and marketing, I would first ask which evidence is missing, put the week behind that gap, and keep marketing quiet until buyers confirm the message.

I am Piet Baudoin and I work under the name Poldermarketing, remote, in Dutch and English. I work with post-seed startups that are building their go-to-market: the first customer, the second, and then a repeatable process that brings the product to the customer who is genuinely happy with it. Next to positioning and messaging and findability in Google and AI answers, I build and run AI automations for clients with care: lead follow-up, enrichment and qualification, CRM setup, reporting, content distribution and connections between tools. I can set up a CRM and read out Search Console queries; Google Ads and Meta Ads are relatively new to me. The conversations with buyers stay with the founder; I take the preparation, the records and the page around them.

If you want an outside read first, the free growth scan shows how your site explains the offer to a stranger.