What Is Traction? How Startups Prove Demand

Ask a founder how things are going and you'll hear a story. Ask an investor what they want to see and you'll hear a number. Traction is the bridge between the two: quantitative evidence that customers want what you've built, measured in a way someone else can check.

The word gets used loosely, so this article pins it down. It covers a working definition, the difference between real traction and vanity metrics, which signals matter for different business models, how traction relates to product-market fit, how investors read it by stage, the Bullseye framework for finding channels, and the common ways founders overstate what they've got.

A working definition

Traction is measurable progress that shows demand is real and growing. That's a plain-language definition rather than a quote from a single authority, and it has three parts worth separating.

• Measurable. A number, not a feeling. "Customers love us" isn't traction. "38 of 50 trial accounts were still active after eight weeks" is.
• Evidence of demand. The number has to reflect people choosing your product, not you pushing it onto them. Signups from a press hit and revenue from a founder's own friends both look like demand without being it.
• Growing or holding. One good week is an event. A pattern across weeks or cohorts is traction.

What counts depends on what you sell. A consumer app measures weekly active users. A marketplace measures completed transactions. An enterprise software company might measure signed pilots. We'll go through those below.

Traction versus vanity metrics

The sharpest distinction in this topic comes from Eric Ries. Writing about vanity metrics versus actionable metrics, he describes vanity metrics as numbers that might make you feel good but don't offer clear guidance for what to do. His example is total website hits: you have 10,000, and now what? The number doesn't tell you what drove those visitors or what to do next.

Ries's alternative is to track groups of customers over time. He suggests a weekly report showing what percentage of registered customers move through lifecycle events like signup, trial use and purchase. If those rates hold steady week to week, nothing significant changed. If they jump or drop, you've got something to investigate.

A useful test for any metric: would the number change a decision you're about to make? If the answer is no, it's decoration.

The a16z team makes a similar point in its guide to 16 startup metrics, where it calls downloads "really just a vanity metric." The same piece says investors want to see engagement, ideally expressed as cohort retention on the metrics that matter for that business, such as daily or monthly active users.

Growth rate as the headline number

If there's one number to know, Paul Graham's answer is the growth rate. In Startup = Growth, he writes that if there's one number every founder should always know, it's the company's growth rate, and that it's the measure of a startup.

He also gives rough benchmarks for startups going through Y Combinator: a good growth rate is 5-7% a week, 10% a week is exceptional, and if you can only manage 1%, it's a sign you haven't figured out what you're doing yet. Those figures describe early-stage startups in that specific setting. They aren't a law for every business.

The reason they matter is compounding. These are our own calculations, not figures from Graham's essay:

A startup growing 1% a week takes about 16 months to double its base. One growing 7% a week multiplies it more than thirty-fold in twelve months. Graham makes the same point from the other direction in Do Things That Don't Scale: if you have 100 users, you need 10 more next week to grow 10% a week, and sustained at that pace you'd reach about 14,000 users in a year.

Two cautions. First, percentage growth is easy when the base is tiny, so a jump from 3 users to 6 isn't the same as 3,000 to 6,000. Second, growth in a vanity metric is still growth in a vanity metric. Measure the rate on something that reflects real use or real money.

Traction by business model

There's no universal traction number. The right signal depends on how the business makes money and where the risk sits. The table below is a practical guide, not a standard.

SaaS and recurring revenue

For software sold on subscription, the central question is whether customers stay. Revenue figures matter, and the standard ones are explained in ARR and MRR. But a growing MRR number can hide a leaky bucket if new customers are replacing ones who leave. Pair it with churn rate and with cohort curves that show whether any group of customers settles into steady use.

Marketplaces

A marketplace has to attract two sides at once, which makes traction harder to read. The a16z guide says investors often start with gross merchandise value, revenue and bookings because they indicate the size of the business, and only then look into growth. For an early marketplace, a better question is how many transactions happened without you personally matching the buyer and seller.

Consumer products

Consumer traction is mostly about habit. Daily or weekly use that persists, and people telling other people without being asked, matter more than any download count. If paid ads supply the entire user base, you've bought attention, not shown pull.

B2B enterprise pilots and letters of intent

Enterprise deals are slow and few, so teams reach for softer evidence: pilots, letters of intent, "we're very interested" emails. These have value, and they're also the easiest to overstate. A reasonable ranking, from weakest to strongest, is a verbal expression of interest, a non-binding letter of intent, an unpaid pilot, a paid pilot, and a signed contract that renews. That ordering is judgment rather than a published standard, but it follows one rule: the more it costs the customer, the more it tells you.

Traction and product-market fit

People often use the two words as if they meant the same thing. They don't.

Traction is the evidence. Product-market fit is the condition that evidence points to. You can have traction without fit: a strong sales push can produce revenue from customers who then churn. And you can have a form of fit before you have impressive traction, if a small group of users clearly can't live without the product but you haven't yet reached many of them.

A practical way to hold both ideas:

1. Early signals, such as the Sean Ellis test and flattening retention in a small cohort, suggest fit may be forming.
2. Traction shows it's reproducible: new customers keep arriving through channels you can repeat, and they keep staying.
3. Scaling spends money to amplify something both of the first two have shown works.

Traction without retention is a warning, not a milestone. Fit without any channel to reach customers is a promising product with no business around it yet.

Key Facts

• Traction is measurable evidence that demand for a product is real, as opposed to opinion, interest or effort.
• Eric Ries defines vanity metrics as numbers that make you feel good but offer no clear guidance for action, and recommends tracking customer cohorts over time instead (source).
• Paul Graham calls growth rate the one number every founder should know, and says 5-7% a week is a good rate during Y Combinator (source).
• a16z says investors often look at GMV, revenue and bookings first for size, then at growth, and that downloads are a vanity metric (source).
• The Bullseye framework from Gabriel Weinberg and Justin Mares uses five steps: brainstorm, rank, prioritize, test and focus (source).
• The right traction signal depends on the business model: retention for SaaS, completed transactions for marketplaces, habit for consumer apps, paid pilots for enterprise.

How investors read traction by stage

Investors don't ask for the same evidence at every stage, because the available evidence changes. The pattern below is a general description of how stages tend to work, not a rule from any one firm.

The a16z guide frames the order well. Investors usually look at size first (revenue, bookings, volume), then at growth, and then at engagement expressed as cohort retention. It compares this to a pediatrician's checkup: check weight and height first, compare against earlier estimates, then go deeper.

Two practical consequences follow. At the earliest stages, a handful of customers who'd be hurt by losing the product is stronger evidence than a large but shallow signup list. And as you move toward later rounds, a single good number stops persuading. Investors want the trend, the cohorts and the explanation. For how the stages map to funding rounds, see Series A, B and C funding.

Finding channels that work: the Bullseye framework

Getting traction usually means finding a way to reach customers repeatedly. Gabriel Weinberg and Justin Mares wrote a book on this, Traction: A Startup Guide to Getting Customers, which describes 19 traction channels and a method for choosing among them called Bullseye. Brian Balfour's write-up of the Bullseye framework summarizes it as five steps: brainstorm, rank, prioritize, test, and focus on what works, then repeat.

Here's how that plays out.

1. Brainstorm. Walk through every channel in the book and write down at least one specific idea for how it could work for you, even the ones that seem unlikely. The point is to avoid defaulting to whatever channel you already know.
2. Rank. Sort the ideas into three groups by promise. Balfour's article describes these as the inner circle (most promising), a potential group, and a long-shot group.
3. Prioritize. Pick roughly three channels for the inner circle. The article quotes the authors on why it should be more than one: they don't want you to waste time testing channels one after another.
4. Test. Run cheap, small tests on those channels in parallel. The goal is to learn which one produces customers at a cost that makes sense, not to build a full program.
5. Focus. Put most of your effort into the channel that works, and repeat the process when it saturates.

Balfour's article gives one example: Noah Kagan used the method at Mint, testing blogs, PR and search marketing before focusing on blogger sponsorships and guest posting, which the article says brought in 40,000 initial customers.

The framework has a limit worth stating. It helps you choose where to spend effort, but it assumes there's a product worth marketing. If the retention curve keeps sliding to zero, finding a better channel just fills the bucket faster. Check that first, using validated learning and the minimum viable product loop.

One more point from Graham applies to the earliest channel of all. In Do Things That Don't Scale he says the most common unscalable thing founders have to do at the start is recruit users manually. Your first traction often comes from you personally, and that's normal. The question is whether the manual work produces something you can turn into a repeatable channel later.

Common ways founders overstate traction

Most overstated traction isn't dishonest. It's a number chosen because it looks good. These are the patterns to watch for in your own pitch and in anyone else's.

• Cumulative totals. "10,000 users" sounds big. Users who were active last month is the number that matters. Totals only go up and hide churn.
• Counting the wrong people. Friends, family, employees and investors' portfolio companies are real customers in the ledger but not evidence of a market.
• Unpaid interest as demand. Waitlists, "very interested" emails and non-binding letters of intent cost the customer nothing, so they predict little.
• Paid acquisition passed off as organic. Growth that stops the day ad spend stops tells you about your budget, not your product.
• Cherry-picked windows. The best week, the best month, or a launch spike, shown without the weeks around it.
• Percent growth from a tiny base. "Up 300%" is easy at 5 users. Say the base alongside the percentage.
• Blended numbers. An average across segments can hide one group that loves you and one that doesn't. Segment before you celebrate or despair.
• Revenue without renewal. First-year revenue from a heavily discounted or founder-sold contract isn't the same as customers who renew unprompted. See growth versus profitability for why the quality of growth matters as much as the speed.

If a number would look different with a different date range, a different denominator or one customer removed, you haven't found the honest version yet.

A short checklist

Use these as questions, not scores.

• What's the one number that best reflects real use or real money in our business, and what's its weekly or monthly growth rate?
• Does that number come from customers staying, or from new ones replacing leavers?
• How much of our growth would continue if we stopped paying for acquisition?
• Which customers cost us the least effort to win, and can we find more like them?
• Have we tested at least three channels cheaply, or have we assumed the one we know best?
• Could an outsider check our number from raw data?

Related reading

• Product-market fit explained
• The Sean Ellis test
• Minimum viable product
• Validated learning
• Founder-led sales
• Growth vs profitability