When Should You Pivot or Persevere? A Practical Evidence Test

The difficult part of a pivot decision is that both choices can sound rational.

Persevering can be disciplined because the evidence is still incomplete. It can also become a way of defending a thesis that repeated experiments are weakening.

Pivoting can be adaptive because the business has learned something important. It can also become a reaction to one disappointing customer, a slow month, or founder fatigue.

A useful pivot decision therefore starts with evidence, not mood.

Pivot when repeated, well-designed tests have weakened a critical business-model assumption and the evidence points toward a credible alternative. Persevere when the core assumptions are still gaining support and the next experiment can reasonably increase that confidence.

A pivot is more than changing a feature, rewriting a homepage, or trying another advertising channel.

The Lean Startup methodology places the decision inside a Build-Measure-Learn loop: startups turn assumptions into experiments, measure customer response, and use learning to decide whether to pivot or persevere. Source: https://theleanstartup.com/principles

Steve Blank similarly describes a pivot as a substantive change to one or more parts of the business model, which can include the customer segment, value proposition, channel, revenue model, or other business-model components. Source: https://steveblank.com/2014/01/14/whats-a-pivot/

That distinction matters because founders can mistake an iteration for a pivot.

Changing the onboarding flow is usually an iteration.

Changing the customer you are building for may be a pivot.

Changing a price slightly may be an iteration.

Changing from a subscription model to a transaction model may be a business-model pivot.

Changing the color of the website is not a pivot.

Changing the problem the company exists to solve is much closer to one.

Write down the original hypothesis before interpreting the new evidence.

For example:

Independent consultants will pay $500 a month for a system that turns their expertise into recurring content.

Then define what would count as evidence.

• Who must show interest?
• What behavior would demonstrate real demand?
• How quickly should that behavior appear?
• What would make the hypothesis weaker?
• What result would make you change the model?
• Which assumptions are critical and which are replaceable?

Without this baseline, the pivot conversation becomes a story about the past.

The founder remembers the effort invested. The salesperson remembers positive conversations. The product team remembers users who liked a feature. Finance remembers the cash already spent.

None of those observations alone tells you whether the underlying business model is becoming more credible.

Separate observations from interpretations.

Suppose twenty prospects were interviewed.

Observation:

• 14 understood the problem immediately.
• 9 said the problem mattered.
• 4 tried the product.
• 1 paid.
• 0 renewed.

Interpretation:

People love the idea but we need better marketing.

That interpretation may be right. It may also be wrong.

The data could indicate:

• the problem is real but the solution is weak;
• the solution works but the buyer is wrong;
• the buyer is interested but unwilling to pay;
• the sales process is targeting the wrong moment;
• the product creates value but not enough value to justify the price;
• the initial demand was curiosity rather than a recurring need.

A pivot decision should be made at the level of the failed assumption, not the emotional impression.

A surprising result is not automatically evidence that the strategy is wrong.

Sometimes the experiment is wrong.

Ask:

Did the test actually expose the assumption we think it tested?

A weak experiment can produce a misleading negative signal.

Examples include:

• interviewing people who are not the target customer;
• showing a concept without asking for meaningful commitment;
• testing a premium offer with an audience that cannot buy it;
• measuring clicks when the real hypothesis concerns willingness to pay;
• changing several variables at once;
• stopping after too little exposure to the market;
• using a sales channel that the target buyer does not normally use.

Before pivoting, identify whether the evidence says:

the hypothesis failed

or

the test failed to test the hypothesis.

Those are different problems.

One bad result should usually trigger investigation.

Repeated failure of a critical assumption is different.

A simple decision log makes the pattern visible:

Do not require every experiment to succeed.

Require the experiments to make the model more informative.

Recent entrepreneurship research supports a more nuanced picture than “good founders pivot” or “good founders persevere.”

A 2023 Journal of Business Venturing study combined qualitative evidence with quantitative data from 80 startups engaged in business-model experimentation. The researchers found that founders can resist changing their value proposition even when they receive negative feedback. They also found that entrepreneurial experience, startup mentoring, and team size may help teams become more receptive to changing business-model assumptions.

Source: https://doi.org/10.1016/j.jbusvent.2023.106314

A separate 2025 study examined 230 teams in the NSF I-Corps program. In that specific deep-tech population, 95% changed at least one business-model component during the program and 66% made changes involving four or more components. Value propositions and customer segments were among the most frequently changed components.

This is evidence about a specific population of early-stage deep-tech teams, not a universal rate for startups.

Source: https://doi.org/10.1002/jsc.2619

Research examining five ventures and 22 pivot-or-persevere episodes also suggests that the two choices are not simply “change” versus “do nothing.” The study describes perseverance as continuing and extending the existing trajectory, while pivoting involves repositioning actions around a revised future.

Source: https://doi.org/10.1016/j.jbusvent.2021.106126

The practical implication is simple:

Perseverance should still contain learning. A pivot should still contain continuity.

You are rarely choosing between standing still and starting again. You are deciding which assumptions deserve another test.

Persevere when the evidence is still strengthening the central thesis.

That does not mean everything is working.

A business can have:

• poor conversion but strong problem evidence;
• weak acquisition but strong retention;
• slow sales but high willingness to pay;
• a good product but an immature channel;
• a promising customer segment but an inefficient sales process.

In these situations, changing the entire business model may destroy useful learning.

Instead, isolate the constraint.

For example:

Problem validated + product value validated + channel weak

is different from:

Problem weak + willingness to pay weak + retention weak.

The first may call for a channel experiment.

The second may justify testing a different value proposition, customer, or problem.

A pivot becomes more defensible when four conditions appear together.

Not a single bad week. Not one difficult conversation. A pattern that survives reasonable testing.

If a secondary feature is unpopular but customers still buy and retain, that may not justify a pivot.

If the target customer consistently does not value the core problem, the implication is much larger.

A pivot should not simply be an escape from the current strategy.

There should be a new hypothesis worth testing.

For example:

The current product is not valuable enough for small agencies, but the same underlying capability may solve a more expensive workflow for larger professional-services teams.

That is testable.

“Let's try something completely different” is not.

A good pivot does not throw away everything.

It may preserve:

• technology;
• customer relationships;
• distribution;
• domain knowledge;
• data;
• brand credibility;
• operational capability;
• a validated problem;
• a working part of the business model.

The point is not to preserve the old plan.

The point is to preserve what the evidence says is still valuable.

A founder can become tired of a strategy before the strategy has actually failed.

Ask:

Would I make the same decision if I had invested nothing in the current business?

This is not a mathematical test. It exposes sunk-cost thinking.

Then ask the opposite:

Would I still want to pivot if I had to personally execute the new direction for the next twelve months?

That exposes novelty bias.

A new direction can feel attractive simply because it is new.

The 2023 research on business-model experimentation is relevant here because it found resistance to changing value propositions despite negative feedback, while suggesting that broader perspectives can make change easier.

Sometimes the useful next step is to bring in a mentor, experienced operator, co-founder, or teammate who can challenge the assumptions without owning the conclusion.

Customer conversations are evidence.

They are not automatic instructions.

One customer can reveal an important problem. One customer can also be an edge case.

Steve Blank has warned against treating every new customer insight as a reason to change direction immediately. His definition emphasizes a substantive change to the business model rather than a reaction to every new piece of feedback.

Source: https://steveblank.com/2012/08/27/vision-versus-hallucination-founders-and-pivots/

The better sequence is:

signal → investigate → test → compare → decide

not:

feedback → panic → pivot

Past investment explains why the decision feels expensive.

It does not tell you what the next dollar should fund.

Separate:

What have we already spent?

from:

What would we fund today if we had today's information?

The first is history.

The second is strategy.

The previous six months are already gone. The decision concerns the next six months.

If the evidence supports considering a pivot, do not immediately rebuild the company.

Design a 90-day test.

Write:

• what is changing;
• what is staying;
• why the old assumption is weaker;
• what evidence supports the new direction;
• what would disprove the new direction.

Keep the change narrow enough to learn from.

Prefer behavior over compliments.

Depending on the business, measure:

• paid pilots;
• repeat usage;
• retention;
• qualified sales conversations;
• conversion;
• willingness to pay;
• referrals;
• usage of the specific new capability;
• time-to-value.

Do not replace meaningful behavior with vanity metrics because the new direction needs a quick win.

Ask:

• Is the new customer segment responding differently?
• Is the problem more urgent?
• Is the sales cycle changing?
• Is willingness to pay improving?
• Is retention improving?
• Are acquisition economics becoming more credible?
• Are we learning faster?

The objective is not to prove the pivot correct.

The objective is to discover whether the new hypothesis deserves continued investment.

Choose one of four states:

Persevere: the evidence is improving.

Iterate: the core thesis still looks credible, but one part needs adjustment.

Pivot: a critical assumption has weakened and the new direction has enough evidence to justify another serious test.

Stop: there is no credible direction that fits the remaining resources and responsibilities.

“Wait and see” should not become a permanent fifth state.

Before changing direction, write one page.

What are we currently trying to prove?

What has increased confidence?

What has reduced confidence?

Which assumption is no longer credible enough?

What are we going to believe instead?

Which assets, relationships, capabilities, or insights remain valuable?

Which customer, problem, product, channel, revenue model, or growth assumption changes?

What will the change consume in:

• cash;
• time;
• team capacity;
• customer commitments;
• technical work;
• opportunity cost?

What is the smallest serious experiment that could increase or decrease confidence in the new direction?

When will you review the evidence again?

This prevents the pivot from becoming a permanent mood.

A pivot can fail for a different reason than the original strategy.

Watch for:

• stronger interest but weak payment;
• more users but worse retention;
• a larger market with a much harder sales cycle;
• enthusiastic early adopters who are not representative;
• founder excitement without customer behavior;
• a technically interesting product with no urgent buyer;
• revenue that depends on one exceptional customer;
• increasing complexity without increasing value.

The new strategy must eventually produce evidence that the old one could not.

Otherwise the company has not pivoted.

It has only moved the uncertainty somewhere else.

The useful question is not:

“Should I be persistent?”

Nor is it:

“Should I be adaptable?”

Those are virtues, not decision criteria.

Ask instead:

1. What did we believe would happen?
2. What actually happened?
3. Which assumption explains the gap?
4. Was the experiment strong enough to learn from?
5. Has the critical assumption failed repeatedly?
6. What evidence still supports the current direction?
7. What alternative direction has real evidence behind it?
8. What would we preserve from what we have already learned?
9. What is the smallest serious test of the alternative?
10. When will we make the next decision?

If you cannot answer those questions, you may not have a pivot decision yet.

You may simply have discomfort.

Discomfort is a reason to investigate.

Evidence is a reason to change direction.

This article synthesizes entrepreneurship research and established customer-development methodology. Research populations are described where they are not representative of startups generally. Founder examples are not used as universal proof.

Key sources:

• Burnell, D., Stevenson, R., & Fisher, G. (2023), *Early-stage business model experimentation and pivoting*, Journal of Business Venturing. https://doi.org/10.1016/j.jbusvent.2023.106314
• Bellavitis, C., Tran, M. H., & Wiklund, J. (2025), *Strategic Pivoting in Deep Tech: An Investigation of NSF I-Corps Teams*, Strategic Change. https://doi.org/10.1002/jsc.2619
• Berends, H., van Burg, E., & Garud, R. (2021), *Pivoting or persevering with venture ideas: Recalibrating temporal commitments*, Journal of Business Venturing. https://doi.org/10.1016/j.jbusvent.2021.106126
• The Lean Startup, *Methodology*. https://theleanstartup.com/principles
• Steve Blank, *What's a Pivot?*. https://steveblank.com/2014/01/14/whats-a-pivot/

SOE TEAM

If you are working through the same decision, these related Knowledge Objects take the next question further:

• Which Customer Requests Should You Actually Build? - distinguish useful customer signal from a one-off request before changing the roadmap.

• When One Customer Carries Too Much of Your Business - examine when one customer can distort strategic decisions.
• If Customers Cannot Easily Leave, Your Subscription Problem Is Bigger Than Churn - consider whether retention signals real value or artificial friction.

Related reading: \2 · \2