Traction is one of those startup words that becomes less useful each time it is repeated.
A launch got 40,000 views. A waitlist has 2,000 addresses. Three companies asked for a deck. An investor replied. Which of these is traction?
None is meaningless. None proves much alone.
Traction is repeated movement by the right user, not a large number standing still.
Start with the job, not the metric
Name the user and the recurring job the product is meant to help them complete.
For Launch Ready, a meaningful action is not merely opening the homepage. It is submitting a real public product URL, receiving a useful report, acting on a finding, and returning for another check when the site changes.
Your product will have a different sequence. Write it in verbs a user performs. Avoid “engagement,” “adoption,” and “value” until each has an observable meaning.
Then ask how often the job naturally recurs. Payroll software should not be judged by daily use. A social feed should not need six months to show a return. The cadence belongs to the problem, not to the dashboard template.
Read traction as a ladder
Early evidence becomes stronger as the user gives up more.
Attention
Page views, impressions, likes, and press mentions prove that something caught attention. They say little about the product unless the audience matches the intended user and continues into the journey.
Attention is useful for comparing messages and channels. It is weak evidence of demand.
Intent
A waitlist signup, demo request, reply, or pricing-page visit costs the user a little effort. It suggests the problem or promise matters enough to investigate.
Intent is still cheap. People join waitlists for products they never use and book calls they later cancel.
Activation
The user completes the first action that produces the promised result. This is the earliest point where product behavior begins to replace marketing behavior.
Choose activation carefully. Creating an account is rarely the result. Importing real data, publishing the first page, completing the first scan, or inviting a collaborator may be closer.
If many relevant users express intent and few activate, inspect the product’s first minute before buying more traffic.
Repetition
The same user returns and completes the job again at its natural cadence. Repetition says the first result was not merely curiosity.
YC’s Startup School discussion of retention makes the useful point that a retention curve that eventually holds a stable group tells a different story from one that keeps falling toward zero.
For a very young product, talk to every returning user. Learn what they did, what triggered the return, and what would make the product disappear from their routine.
If accounts appear after launch and then go quiet, use the startup retention diagnosis to separate a tracking error, failed activation, weak result, wrong audience, and missing return loop.
Payment
Payment proves priority under real constraints. It also tests the offer, price, procurement path, and trust required to exchange money.
Revenue can still mislead. One founder-led consulting project does not prove a repeatable software business. A heavily subsidized transaction may grow while destroying value. Record what was sold, how much manual work delivered it, and whether the customer would renew under honest terms.
Expansion and referral
A customer adds usage, data, seats, projects, or spend. A user brings another person without being bribed. These actions suggest the product’s value survives beyond the first transaction.
Do not manufacture referrals before users care. A referral program attached to a leaking product pays people to expose the leak to friends.
Measure cohorts, not the growing pile
Cumulative users almost always rise because yesterday’s accounts remain in the total. That curve can look healthy while every new cohort quietly leaves.
Group users by the week or month they first reached the core result. For each cohort, record how many repeat the job at the next natural interval. Segment by user type and acquisition source when the sample permits it.
The question is not “How many accounts do we have?” It is “Of the right people who reached value, how many found enough value to return?”
Recent a16z analysis of AI products warns that early AI curiosity can produce an initial group of tourists who leave quickly, making later retention behavior especially important. Their retention analysis argues for reading mature cohorts rather than treating the first surge as the long-term curve.
Do not borrow its benchmarks blindly. A weekly workflow tool, an annual tax product, and an occasional emergency service have different honest shapes.
Study power users before averages become useful
With twelve active users, an average can hide the company.
Find the people who return, complete the job deeply, pay, tolerate rough edges, or ask to bring colleagues. Observe their workflow. Which problem do they have that the others do not? Which feature do they ignore? What event makes the product urgent?
a16z calls this narrower state product-user fit: evidence that the product is right for a specific user without yet proving a large market of similar users. The distinction protects a promising small truth from becoming a premature large claim.
Your next task is to find more people who share the conditions that made those users care.
Track the slope without worshipping a number
Paul Graham’s Startup = Growth argues for measuring early progress frequently and prefers revenue, then active users when revenue is not yet available. The durable idea is not that every company must hit one borrowed percentage. It is that a startup needs a measurable direction and a short feedback loop.
Choose one primary weekly measure tied to value:
- retained teams completing the core job;
- paid transactions from returning buyers;
- active projects with a successful outcome;
- revenue from customers still receiving the promised service.
Pair it with guardrails: failure rate, refund or churn, time to result, support burden, and gross margin when relevant. Growth that comes from broken promises, unsafe behavior, or negative economics is not healthy traction.
Remove the founder’s thumb from the scale carefully
Paul Graham’s advice to do things that do not scale explains why founders should recruit and delight early users manually. This work creates traction and teaches what the eventual product must do.
It can also hide which value belongs to the product and which belongs to the founder.
Record the manual service around each account: setup, data cleaning, reminders, custom analysis, support, and exceptions. Do not stop doing it too early. Instead, test which parts users miss when the process becomes slightly more self-serve.
The manual work is not fake traction if the customer receives real value and pays honest terms. It becomes misleading when the company describes bespoke labor as automatic software or assumes it will vanish cheaply at scale.
Do not count fundraising as customer traction
Capital can finance the search for traction. It is not evidence that users receive value.
YC’s Michael Seibel warns in The Real Product Market Fit that founders sometimes point to hiring or funding when the stronger evidence would be growing numbers of happy, loyal, preferably paying customers.
An investor may back the team, market, technology, or possibility before customer pull exists. Say which kind of evidence you have. Precision is more useful than declaring product-market fit early.
Check that the product is not erasing the signal
A broken verification email, production login loop, confusing first screen, or payment mismatch can make demand look weak. Before interpreting low activation as market rejection, complete the core journey from a clean account on the public domain.
Use the common launch problems guide to separate product failure from audience failure. The market cannot pull a product it cannot enter.
Write the weekly traction note
Keep it to one page:
User and recurring job: Natural usage cadence: Primary value event: Primary weekly measure:
New relevant users: Users who reached value: Users who repeated the job: Users who paid, expanded, or referred:
Strongest cohort or segment: Largest observed failure: Manual work per successful user: What changed our belief: One test for next week:
Small numbers are not embarrassing when they describe real behavior. Large numbers are not impressive when nobody returns.
Traction begins when a specific user receives value twice. It becomes compelling when more of the same users arrive, stay, pay, and bring others without the company inventing a new explanation each week.