What it means
The term comes from the startup world, where it describes turning on a fixed foot: one thing stays anchored while everything else moves. A company might keep its technology but sell it into a completely different industry, or keep its audience but replace the product entirely.
What separates a pivot from a restart is that something already learned or built is deliberately carried forward. Pivots usually follow evidence that the current plan is not working.
Typical signals include flat usage after launch, customers who sign up but never renew, an acquisition cost that refuses to fall, or a small subset of users getting far more value than everyone else. That last signal is the most common trigger, because it points to a market the business found by accident.
There are recognisable types worth naming. A customer segment pivot keeps the product and changes who it is sold to, a business model pivot changes how revenue is earned, for instance moving from one-off licences to subscription, a zoom-in pivot turns a single popular feature into the whole product, and a platform pivot moves from selling an application to selling the infrastructure underneath it.
Naming the type forces a team to be precise about what is changing and what is staying. The practical constraint on any pivot is cash.
A change of direction resets the clock on product development and sales, so the real question is whether the remaining runway covers the time needed to prove the new direction works. Boards typically want a specific hypothesis, a cheap test, a deadline and an agreed measure of success before releasing the money.
Pivoting too often is as dangerous as pivoting too late. Each change discards accumulated learning, confuses customers and staff, and spends cash on work that then gets abandoned, which is why serial pivoting is a recognised warning sign for investors.
The discipline is to decide in advance what evidence would justify changing course, and then act on it once that evidence appears.
In practice
Real-world examples
Example 1
A conference organiser whose events were cancelled builds a paid online community instead, after noticing that recordings of its expert panels were being rewatched far more than the live sessions. The audience and the speaker relationships carry over, but the revenue model changes from ticket sales to subscriptions.
Example 2
A furniture manufacturer built an internal scheduling tool to manage its own workshop, then discovered that three competitors wanted to buy it. Over two years the software becomes a separate product line with better margins than the furniture business that created it.
Example 3
A consumer budgeting app finds that most of its active accounts belong to small business owners tracking business spending. It rebuilds the product around business needs, raises the price from $6 a month to $29 a month, and loses two thirds of its users while tripling revenue.
Think of it
Imagine you are driving to a holiday destination and hit a massive road closure. Instead of sitting in traffic until your fuel runs out, you check the map, turn around, and take a different route that gets you to the same destination safely.
Formula
Calculation
There is no standard formula for a pivot, but the arithmetic that decides whether one is affordable is the cash runway calculation:
Runway in months = Cash in the bank / Monthly net cash burn
Suppose a company holds $1,800,000 in cash and burns $150,000 a month.
Runway = $1,800,000 / $150,000 = 12 months.
The team estimates the pivot needs nine months to produce a first meaningful revenue signal, which leaves 12 - 9 = 3 months of margin for error, an uncomfortably thin cushion.
By cutting monthly burn to $120,000 through slower hiring, the runway becomes $1,800,000 / $120,000 = 15 months, giving 15 - 9 = 6 months of cushion instead of three. It is usually that extra cushion, rather than the appeal of the new idea, that persuades a board to approve a change of direction.Case study
Seen in the real world
Kettleford Labs is an illustrative software company used here to show how a pivot decision gets made. It had spent two years selling a scheduling tool to independent physiotherapy clinics, reaching $420,000 of annual recurring revenue but with churn running at 4% a month, which meant it was replacing most of its customer base every two years.
Digging into the numbers, the team found that eleven customers, all veterinary practices rather than physiotherapists, churned at under 1% a month and used a reporting feature the core market barely touched. The founders framed a hypothesis, ran a three-month test selling only to veterinary practices, and set a clear threshold in advance: forty new customers and monthly churn under 1.5%.
In this fictional case the test produced fifty-two customers and 1.1% churn, so the board approved the pivot and the physiotherapy product was placed into maintenance. The point of the illustration is not the outcome but the structure, since the company defined its evidence, its deadline and its success measure before spending the money rather than after.
Watch out
Common mistakes
- Calling any change of plan a pivot, when a genuine pivot keeps a deliberate anchor such as the technology, the customer base or the channel while changing the rest.
- Pivoting on opinion or fatigue rather than on evidence, which usually means the same underlying problem reappears in the new direction.
- Starting a pivot without checking the cash runway, so the business runs out of money halfway through proving the new direction works.
Questions
People also ask
How do you know it is time to pivot?
When repeated, well-executed attempts at the current plan produce no improvement in the numbers that matter, and a smaller adjacent group of customers is clearly getting far more value than the target market.
Does a pivot mean the previous work was wasted?
Not usually, because the technology, brand, team knowledge and customer relationships typically carry across, and the failed attempt is what produced the insight behind the new direction.
How many pivots are too many?
There is no fixed number, but each one should be justified by fresh evidence and given enough time to be tested properly, since repeated changes with no learning between them signal that the underlying problem has not been identified.