← Writing

Pivot or Persevere

Published
Reading
19 min
Contents · 24 sections
  1. What a pivot is (and isn’t)
  2. Pivot, don’t jump
  3. The land of the living dead
  4. Why founders wait too long
  5. The pivot-or-persevere meeting
  6. Ten kinds of pivot
  7. Signals: keep going or change?
  8. Signals to persevere
  9. Signals to change something
  10. Three pivots, and what each kept
  11. Instagram: a zoom-in
  12. YouTube: a change of need
  13. Slack: from a game to a tool
  14. Kill metrics: decide before you know
  15. Runway measured in pivots
  16. Worked example: Fieldpad decides
  17. The kill metric, set in January
  18. What happened
  19. The pull
  20. How many tries are left
  21. The decision
  22. The new kill metric
  23. How to pivot without starting from zero
  24. Key ideas

In November 2012, Stewart Butterfield announced that Glitch, the online game his company Tiny Speck had spent years building, would shut down. It had raised money, hired more than 40 people, and still hadn’t found enough players.

The company didn’t close, though. In the shutdown post, Butterfield mentioned that the team had built “unique messaging technology with applications outside of the gaming world.” It was the chat tool they had made for themselves while working on the game.

That tool became Slack. When its preview opened on August 13, 2013, Butterfield later told First Round Review, 8,000 people asked for an invite on the first day, and 15,000 within two weeks.

Stories like this make “pivot” sound like a magic word. Most pivots don’t produce a Slack. The famous ones are famous because they worked. The useful lesson is how the Glitch team decided: they looked at what was actually working, kept it, and dropped the rest.

This post is about making that decision on purpose: when to keep going, when to change, and how to change without starting from zero.


What a pivot is (and isn’t)

The word comes from Eric Ries’s 2011 book The Lean Startup. Chapter 8 is called “Pivot (or Persevere).” In an excerpt Ries published in Entrepreneur that year, he defines a pivot as a “structured course correction designed to test a new fundamental hypothesis.”

Three words in that definition do the work.

Structured. A pivot is a decision you make at a planned moment, using evidence. It isn’t a panic on a bad Tuesday.

Course correction. You keep the team, the code, the customers and, most of all, what you learned. In the book, Ries describes keeping one foot rooted in what you’ve learned.

New hypothesis. A hypothesis is a guess you can test: “property managers will pay $79 a month for this.” A pivot swaps one guess for a different one, which means you need a new test and a new number that tells you if it worked.

Persevere is the other choice: keep the strategy and keep improving it. That’s only stubbornness when the numbers say it isn’t working.

Pivot, don’t jump

In a 2009 essay, republished in 2016, Ries describes two ways startups get stuck. The first is the company that never changes direction: still spending energy, never really moving. The second is the compulsive jumper, who never stays on one direction long enough to find out if anything is there. His advice for both fits in three words: “pivot, don’t jump.”

A jump throws everything away and starts over. A pivot changes one big thing and keeps the rest.


The land of the living dead

The first failure mode has a name. In the same Entrepreneur excerpt, Ries warns that a company that won’t change direction based on feedback “can get stuck in the land of the living dead.” It isn’t growing enough to succeed and it isn’t dying, so it keeps eating money and people’s time.

In the book, he describes these companies as having just enough success to stay alive, but not enough to meet what the founders and investors hoped for.

I think this is the most common outcome for small software companies. There’s no crash, no moment that forces a decision. A few customers sign up, a few leave, and next quarter will be different.

Dying is at least a clear signal. The living dead never get one. That’s why the decision has to be forced by a process instead of waiting for the market to force it.

Why founders wait too long

Ries gives three reasons founders delay a pivot. In my words:

  1. Vanity metrics. A vanity metric is a number that goes up and feels good but doesn’t tell you whether the business works: total signups, page views, a total-customer line that hides churn. If you watch those, you can believe you’re winning for a long time.
  2. A vague hypothesis. If you never said clearly what success looks like, you can never clearly fail. “We’ll see how it goes” can’t be proven wrong.
  3. Fear. Ries writes that founders’ biggest fear isn’t being wrong. It’s being judged wrong before the idea had a real chance. So they keep giving it “one more chance.”

He also notes that most founders who did pivot say they wish they had done it sooner.


The pivot-or-persevere meeting

Ries’s fix is boring on purpose: a regular meeting, booked in advance, whose only job is to decide pivot or persevere.

On timing, he says less than a few weeks between meetings is too often and more than a few months is too rarely, and each startup should find its own pace. Booking it ahead matters because the decision is emotional. If the meeting only happens when someone feels things are going badly, it happens late, or never.

In the book, the product side brings results of its improvements over time, and the business side brings what customers actually said. Here’s the agenda I’d use for a small team:

Step Question Bring
1 What did we say we’d hit, and by when? The kill metric you wrote down last time
2 Did we hit it? One number, not a dashboard
3 Is any part of this pulling much harder than the rest? Cohort table split by segment, feature and channel
4 What did customers say and do? Call notes, cancellation reasons, quotes
5 How many more tests can we afford? Cash, burn, and how long one test takes
6 Decision Persevere, pivot (which kind), or stop
7 New kill metric and date Written down before anyone leaves

Step 7 is the one people skip, and it’s what makes the next meeting work.


Ten kinds of pivot

In the chapter’s “catalog of pivots,” Ries lists ten types. Each changes one part of the business and keeps the others.

Pivot What changes What you keep
Zoom-in One feature becomes the whole product That feature’s users and code
Zoom-out The whole product becomes one feature of something bigger The product, now a part
Customer segment Same problem, different customer The product, mostly
Customer need Same customer, different problem The relationship and the knowledge of them
Platform App becomes a platform for others, or the reverse The core technology
Business architecture High price, few customers becomes low price, many customers, or the reverse The product
Value capture How you make money: who pays, for what The product and users
Engine of growth How you grow: word of mouth, paid ads, or repeat revenue from the same customers The product
Channel How you reach customers: direct sales, app stores, partners The product and the customer
Technology The same solution, built a different (cheaper, better) way The customer and the offer

The point is that “pivot” is not one move. “We’re pivoting” should always be followed by “from what to what, and what stays.” I’d look first at zoom-in, customer segment and customer need, because your data usually points at them directly.

Meeting day: did we hit the kill metric by the date? yes no Persevere keep improving what works Is one part pulling much harder? a feature, a customer group, a need yes no Targeted pivot zoom-in: one feature becomes the product customer segment: same product, new buyer customer need: same buyer, new problem Cash for another full test cycle? yes no Bigger pivot channel, platform, value capture, engine of growth, technology, business architecture Stop wind down cleanly, keep what you learned New hypothesis + new kill metric + new date then back to the next scheduled meeting
The meeting as a decision tree: check the number you set in advance first, look for the part that is pulling before changing everything, and leave every pivot with a new metric and a new date.

Signals: keep going or change?

No single number decides this. But some patterns show up again and again. Here are the ones I trust most.

Signals to persevere

Retention curves that flatten. Follow each month’s group of new customers and check what share is still paying as they age. If the curve drops and then levels off, a core group found something they keep needing. That’s the strongest signal I know, and Cohort Retention: The Chart That Tells the Truth shows how to build the table.

Newer groups doing better than older ones. If January’s customers stick better than last October’s, your changes are working. Keep making them.

Customers who would be upset to lose you. See Product-Market Fit: How to Know You Have It.

Shortening sales cycles. A sales cycle is the time from first contact to paying. If it’s getting shorter as you learn to explain the product, the message is landing.

Signals to change something

Curves that keep falling. A retention curve that slopes down without ever leveling off ends at zero. Better marketing pours more water into a leaking bucket.

Pull from one segment. Split your customers by type. Often one small group retains much better, pays more and buys faster, while the large group you aimed for drifts away. That gap is the most useful signal a struggling startup gets, because it points at a specific pivot.

Pull toward one feature. Most usage lands on one feature you thought was minor. That’s a zoom-in signal. It’s what happened to Instagram, below.

Sales cycles that stay long. If buyers keep needing “one more meeting,” or asking for things the product doesn’t do, your offer may not match their need. Founder-Led Sales covers reading those calls.

Growth that only comes from effort. No referrals, ever. Before you blame the product, reread Distribution Beats Product.


Three pivots, and what each kept

These three are documented by the founders themselves. They’re also survivors, so read them for the mechanism, not as a promise.

Instagram: a zoom-in

Kevin Systrom’s first app, Burbn, was a location check-in web app with plans, points for hanging out with friends, and photo posting, as he described it in a Quora answer TechCrunch reported in November 2010.

After about a year, the team built an iPhone version and found it “cluttered” and “overrun with features.” Location was crowded with rivals like Foursquare and Gowalla. So they cut almost everything. Systrom wrote that they kept only the photo, comment and like features.

That became Instagram. TechCrunch reported about 100,000 users one week after launch, 200,000 the week after, and 300,000 the week after that.

What they kept: the team, the investors, the code for photos, and the part of Burbn they judged most promising. In Ries’s terms, a zoom-in: one feature became the whole product. And at about a year in, cutting most of it was still cheap. The longer you wait, the more you throw away.

YouTube: a change of need

Speaking at SXSW in 2016, YouTube co-founder Steve Chen said the founders first asked what video on the web was actually for, and “We thought dating would be the obvious choice.” They registered the domain on February 14, 2005.

Digital Trends reported that after five days, not one video had been uploaded. Chen described the next step plainly: forget the dating part, and open the site to any video.

What they kept: the video upload and playback technology. What changed was the need it served. Not “find a date,” but “share any video.”

Slack: from a game to a tool

The Glitch team kept their internal chat tool, the people who built it, and the remaining money. TechCrunch’s 2019 history notes that Butterfield offered investors their money back unless they backed the next idea, and a smaller core team moved on to new products.

What they kept: the code, the team, investor trust, and years of using their own tool every day. The customer and market were completely new, but they had been their own first users for years.

None of the three started from a blank page. Each kept something already working. That’s the difference between a pivot and a jump.


Kill metrics: decide before you know

You make this decision while emotionally invested. The fix is to make most of it earlier, before the results come in.

A kill metric is a number and a date, written down in advance, that decides what happens next. Annie Duke, the former professional poker player who wrote Quit, calls these kill criteria. In a 2022 First Round Review interview, she gives a simple template: “If by (date), I have/haven’t (reached a particular state), I’ll quit.”

A state plus a date. For startups, I’d swap “I’ll quit” for “we change one big thing.” Good kill metrics have four parts:

  1. One or two numbers, not ten. If you track ten, one will always look fine.
  2. Numbers that measure behavior, not vanity. Paying customers still active after three months beats total signups.
  3. A date that comes before the money runs out, with enough time left to try something else.
  4. Written down and shared with a co-founder or advisor, so you can’t quietly move it.

Runway measured in pivots

Normally runway means months of cash left: $100,000 in the bank, $10,000 a month spent, ten months. In the book, Ries argues the real measure is how many pivots you have left: how many more times you can make a big change and test it before the money runs out.

So there are two ways to extend runway: more money, or shorter test cycles. Testing in six weeks instead of three months doubles your chances on the same bank balance. Default Alive or Default Dead? walks through the cash side.


Worked example: Fieldpad decides

Fieldpad is a fictional company. All its numbers are made up to show the method, and I checked the arithmetic in Python.

Fieldpad makes an app for writing up building inspections on a phone: photos, checklists, and a finished report in one tap. Two founders launched in January, aimed at small building contractors, at $29 a month.

The kill metric, set in January

Before launch, they wrote down:

By June 30, at least 40% of customers who signed up January to March are still paying.

One number, on the January-to-March groups, because by June 30 each of them is at least three months old.

What happened

On June 30, at their scheduled meeting, they split those signups by customer type. Along the way they had picked up a group they hadn’t planned for: property managers, who run rental buildings and inspect units when tenants move in and out. They paid $79 a month on a larger plan.

Monthly recurring revenue (MRR) is what customers pay you each month on subscriptions.

Contractors Property managers Total
Signed up, Jan-Mar 60 18 78
Still paying on June 30 9 11 20
Share still paying 15% 61% 26%
Price per month $29 $79
MRR from these customers $261 $869 $1,130
New signups per month 20 6 26
Median days, trial to paid 34 9

Against the kill metric: 26% still paying, target 40%. Missed. By their own rule, they don’t persevere.

But the totals hide the most important fact: the two groups behave like two different businesses.

The pull

Property managers are 18 of 78 signups, 23%, but they bring in 77% of this group’s MRR ($869 of $1,130). They stay four times as well, pay 2.7 times as much, and decide in 9 days instead of 34.

A fairer comparison is what each group’s normal flow of signups is worth once the leavers are gone:

  • Contractors: 20 signups a month × 15% still paying × $29 = $87 of lasting MRR per month of signups.
  • Property managers: 6 signups a month × 61% × $79 = about $290 of lasting MRR per month of signups.

The group nobody was even marketing to brings 3.3 times as much lasting revenue from less than a third of the signups. Per signup, a property manager is worth about 11 times a contractor.

That’s the “pull from a segment” signal, and it points at a customer-segment pivot: same product, different customer.

How many tries are left

Counting everyone, including customers who joined April to June, Fieldpad’s MRR is $1,800. Costs are $8,000 a month, so they spend a net $6,200 a month. They have $74,400 in the bank: 12 months of cash.

Each test of a new direction takes them about three months: change the product and the marketing, wait for new customers, then wait long enough to see who stays. So in Ries’s terms they have about four pivots left. If they could run each test in six weeks, they’d have about eight.

Four isn’t many, so spend this one where the evidence is strongest, not on a brand-new idea.

The decision

They pivot to property managers. Here’s what changes and what stays:

Changes Stays
Who they market to and sell to The inspection and report engine, most of the code
Homepage, examples, onboarding checklist Every paying customer, contractors included
The roadmap: features property managers asked for, such as move-in vs move-out comparison What they learned in six months of customer calls
Main price point becomes $79 The two founders and their runway

Contractors keep their accounts. Fieldpad just stops building and marketing for them.

The new kill metric

Before leaving the meeting, they write the next one:

By September 30, at least 8 new paying property managers in each of August and September, and at least 70% of July’s new property managers still paying.

They already get 6 a month without trying, so 8 is a stretch, not a fantasy. Old property managers were at 61% after three months, so 70% asks the new onboarding to do a bit better. If they miss, they’ll still have about three pivots of runway left.


How to pivot without starting from zero

Write down what you’re keeping before what you’re changing. Customers, code, data, content, relationships, what you learned. If the list is short, you may be jumping.

Change one big thing at a time. Fieldpad changed the customer; the price and roadmap followed from that. Change three unrelated things at once and you won’t know which one mattered.

Tell your existing customers honestly what changes for them and what doesn’t. Some will become your references.

Reuse your evidence. Old call notes and cancellation reasons are free research for the new direction. Your First 10 Customers covers finding more of the new type.

Re-check positioning and price. A new customer needs a new answer to “who is it for and why is it better.”

Set the next kill metric before you start, not after the first good week.

Know when it isn’t a pivot. If nothing is pulling and the cash covers less than one more full test, the honest choice may be to stop. Stopping with money left beats a year in the land of the living dead.


Key ideas

  • A pivot is a planned change to one big part of the strategy, keeping the rest. A jump throws everything away. Ries’s advice: pivot, don’t jump.
  • The land of the living dead is the most common failure: not growing, not dying, never forced to decide. A process has to force the decision for you.
  • Hold a pivot-or-persevere meeting on a fixed schedule, every few weeks to a few months, booked in advance.
  • “Pivot” isn’t one move. Ries lists ten types; zoom-in, customer segment and customer need are the usual first candidates.
  • Look for pull before changing everything: one segment, feature or need doing much better than the rest.
  • Set a kill metric in advance: a state and a date, written down and shared, so the decision is made before the emotions arrive.
  • Measure runway in pivots left, and make each test cycle shorter to get more of them.

Sources: Eric Ries, “Pivot or Persevere? The Key to Startup Success”, Entrepreneur, September 2011 (excerpt from The Lean Startup: pivot definition, land of the living dead, why founders delay, meeting cadence); Eric Ries, The Lean Startup, Crown Business, 2011, chapter 8 (catalog of pivots, meeting attendees, runway as pivots remaining); Eric Ries, “Revisiting the Pivot”, Next Big Idea Club, 2016, first published 2009 (the living dead and the jumper, “pivot, don’t jump”); Big Think, “When to Pivot, When to Persevere” (the ten pivot types, cross-checked); Kate Clark, “The Slack origin story”, TechCrunch, May 2019; First Round Review, “From 0 to $1B: Slack’s Founder Shares Their Epic Launch Strategy” (8,000 and 15,000 preview requests); M.G. Siegler, “A Pivotal Pivot”, TechCrunch, November 2010 (Burbn and early Instagram users); Sam Gutelle, “YouTube Was First Conceived As A Dating Site, Says Co-Founder Steve Chen”, Tubefilter, March 2016; Saqib Shah, “YouTube dating: Co-founder says Google’s video platform started life as a hook-up service”, Digital Trends, March 2016; First Round Review, “Grit or Quit? Tactical Advice for Founders Facing Tough Company-Building Decisions”, October 2022 (Annie Duke on kill criteria). Fieldpad and all its numbers are fictional.

Related: Cohort Retention: The Chart That Tells the Truth, the chart that tells you whether to persevere; Validate an Idea in One Week, the same go, pivot or stop decision before you build; and Default Alive or Default Dead?, for counting the runway your pivots come out of.