Before you scale acquisition on a leaky book
Logo Churn for Startups: count lost customers as a pre-written retention tripwire
A slide that prints “logo churn is 8%” without the start count, who counted as a logo, or what you will do if the next whale cancels is not validation. Logo churn asks a smaller, honester question: of the paying accounts you already had at the start of the period, how many left — against a bar you wrote down before you looked? Matching the gate is a decision signal. It is not a forecast that you will succeed.
Last updated: October 22, 2026
Direct answer
What is logo churn — and why write the tripwire first?
Logo churn — also called customer churn — is customers lost during a period divided by customers at the start of that period. It answers “how many accounts left,” not “how many dollars left.” Write the tripwire before you look: “for ICP logos in period N, logo churn must stay ≤X% (or we stop scaling acquisition / change ICP / change product).” Hitting or missing it is a decision signal — not a prediction of success.
Key takeaways
What to remember
- Logo / customer churn counts accounts. For Entrepreneurs / Matrix (Anoushka Vaswani) defines it as customers lost over a period ÷ customers at start, and asks you to track reasons as carefully as the rate. Voluntary churn is a customer who dislikes the product or picks a competitor. Involuntary churn is a customer who goes out of business or is acquired. Dollar retention lives on the NRR page.
- Gross dollar churn (same Matrix piece) is lost revenue from churned customers and from downsells / downgrades. It can diverge from logo churn when you keep cheap logos and lose expensive ones — or the reverse. Net dollar churn subtracts expansion / upsell; it can go negative when expansion exceeds churn plus contraction. Best-in-class net dollar churn is negative — companies that grow even with zero new bookings. Read NRR/GRR depth on the SaaS NRR page. This page does not reproduce Bessemer Scaling GRR/NRR band tables.
- Bessemer Atlas (Gabe Miller-Smith, 29 August 2023): desk mantra that it costs some 5–25× more to acquire a new customer than retain an existing one; $500M ARR with 20% churn loses $100M of ARR, so the first $100M of new bookings only fills the leaky bucket. ICP logo-churn rules of thumb: Good 10–15%, Better <10%, Best <5%. SMB / lower-CAC: Good 15–20%, Better 10–15%, Best 5–10%. Non-ICP early stage: 15–20% normal if the roadmap only solves a couple of use cases; 20–30% means take a closer look. Gabe: if churn is in double digits there is room to improve; at Procore he never had a churn target greater than 9%; 15% is not a panic, but it is org-wide work.
- Always use the start-of-period count. State monthly versus annual. Do not annualize a monthly rate by multiplying by 12 — if you need an annual figure from a stable monthly rate m, compound with 1 − (1 − m)^12. That is arithmetic, not a benchmark. With few logos, show counts (n lost / n start), not only a percent. Matrix’s dummy set uses 30 customers to show how early-stage rates swing. One lost logo of 12 is not a stable rate.
- Write the tripwire before you measure: segment (ICP / SMB / non-ICP) + period + definition (who is a logo; trial→paid; voluntary vs involuntary) + threshold (Gabe’s band or a tighter floor you justify) + named action (stop paid acquisition / re-qualify ICP / cut time-to-value / keep experimenting / do not fill the leaky bucket). A Yibud score is not logo churn and not a success prediction.
Why it matters
Why a blended “churn is fine” slide can still hide a leaky book
Founders treat “people keep paying” as a feeling. Feelings do not name whether the accounts that left were the cheap ones or the whales, or whether next quarter’s paid spend is only filling a hole. Two sittings are useful here as instruments, not as personal odds: Gabe’s logo bands by segment, and Matrix’s three-rate split.
Gabe: neglecting a small renewable book gets worse with scale
Bessemer Atlas (29 August 2023) records Gabe Miller-Smith, then SVP of Global Customer Success at Procore: early-stage teams obsess over new bookings while the renewable book is still small. The piece’s example: $500M ARR with 20% churn loses $100M of ARR — the first $100M of new bookings only “fills the leaky bucket.” The desk mantra he taped up: it costs some 5–25× more to acquire a new customer than retain an existing one. This page does not add a Harvard primary citation beyond that attributed mantra.
Matrix: a kind logo percent can hide ugly dollar losses
Vaswani’s Matrix / For Entrepreneurs piece tracks customer (logo), gross dollar, and net dollar churn together. Their illustrative business keeps $100 logos and loses many $1,000 and $2,000 accounts — a trend you miss if you only watch logo churn. Best-in-class gross dollar churn is typically less than 1% monthly. Best-in-class net dollar churn is negative; Zendesk is named as a portfolio example of expansion after purchase as more agents join the platform. Series A companies may not yet see negative net churn until pricing and packaging mature.
Gabe: blended 5–10% can still hide a segment you should stop selling
The same Bessemer piece: a blended 5–10% can look fine while one geography, vertical, or spend slice is leaking. Sometimes the honest action is to stop selling a segment with no path to profitability. For CS leaders, the piece’s North Star is net retention — one sitting, then go to the SaaS NRR page. This page still owns logo count as the tripwire.
The six terms
Logo churn, voluntary vs involuntary, gross dollar, net dollar, period discipline, small-n honesty
These six fields are the instrument. If any one is fuzzy, the percent is a slide, not a tripwire. This page does not publish a homemade “good monthly logo churn is 2%” table.
Logo / customer churn — accounts lost ÷ accounts at start
Vaswani / Matrix: the number of customers lost over a period divided by customers at the start of that period. Track reasons as carefully as the rate. Gabe’s closed list of common reasons: Price; Lack of adoption; Missing features; Switching to a competitor; Going out of business or merging. Add “other” only if you must. Do not call a logo-churn percentage NRR.
Voluntary vs involuntary — why the account left
Matrix: involuntary churn is a customer who goes out of business or is acquired — often outside your control, especially when you sell to early-stage startups. Voluntary churn is a customer who dislikes the product and cancels, or picks a competitor. Mix them in one unlabeled percent and you cannot name the action.
Gross dollar churn — lost revenue from cancels and downsells
Matrix: total lost revenue from customers churning and from downselling (their example: a $1,000/mo plan downgrades to $100/mo). Logo churn and gross dollar churn diverge when you keep cheap logos and lose expensive ones. Best-in-class gross dollar churn is typically less than 1% monthly. That is a monthly sitting, not a license to print 12%.
Net dollar churn — gross dollar losses minus expansion
Matrix: revenue lost from churn and downsell, minus gains from upsell — the change in monthly recurring revenue excluding new bookings. It can go negative when expansion exceeds churn plus contraction. Best-in-class is negative net dollar churn: the company still grows with zero new bookings. Zendesk is the named portfolio example. Point at the SaaS NRR page for NRR/GRR depth. Do not paste Bessemer Scaling GRR/NRR band tables here.
Period + denominator — start count, and do not multiply by 12
Always divide by the start-of-period count, not the end-of-period count and not an average you invented after the close. State monthly versus annual on the dated page. If you need an annual rate from a stable monthly rate m, compound with 1 − (1 − m)^12. Example arithmetic only: a stable 1% monthly rate compounds to 1 − (0.99)^12 ≈ 11.4%, not 12%. Matrix’s best-in-class gross-dollar ceiling is typically less than 1% monthly — keep that label monthly.
Small-n honesty — counts first, percent second
Matrix’s dummy set uses 30 customers to show that early-stage rates swing until the book grows. They also note they typically spend less time on customer-churn benchmarks internally — use Gabe’s ICP / SMB / non-ICP bands as segment-sensitive vocabulary, not a universal “good logo churn.” One lost logo of 12 is 8.3% and also one account. Write both.
Worked example
One year of logos that look “fine” while dollars do not (illustrative — fictional)
The company, names, and headcounts below are invented so you can see a logo percent sitting in Gabe’s ICP “Good” band while gross dollar churn is ugly. Price points echo Matrix’s published dummy shape ($100 vs $1,000 / $2,000). This is not a study, not an industry benchmark, and not a Yibud report. Copy the method, not the counts.
The promise, written before 1 January closes
Illustrative names: founder Kenji Okada at Stackline, a B2B inventory tool. The tempting trophy is a blended “logo churn is fine” slide. The tripwire is written first, on 2 January, before anyone opens last year’s book: for ICP logos in calendar 2025, logo churn must stay ≤10% (Gabe’s Better band for ICP), counted as logos lost ÷ logos paying on 1 January, or we stop paid acquisition and re-qualify ICP. Report n lost / n start. Bucket reasons with Gabe’s five plus optional other. Dollar NRR is a separate sitting.
Start count — 24 paying logos, not a blended percent
Start of 2025 (fictional): 24 paying logos. Twenty at $100/mo (ARR $1,200 each), three at $1,000/mo (ARR $12,000 each), one at $2,000/mo (ARR $24,000 each). Start ARR: $84,000. Each logo is a paid account that had already converted from trial — the same definition the tripwire named.
What left — 3 / 24 looks “Good”; the dollars do not
Lost during 2025 (fictional): one $100 logo (price), one $1,000 logo (missing features), the $2,000 logo (switched to a competitor). Logo churn: 3 / 24 = 12.5%. That sits in Gabe’s ICP Good band (10–15%) and misses the ≤10% Better floor they wrote before looking. Gross dollar churn on starting ARR: ($1,200 + $12,000 + $24,000) / $84,000 = $37,200 / $84,000 ≈ 44%. Cheap logos stayed. The expensive ones left. Matrix’s warning, in invented clothes.
What the tripwire does — and what must not happen
12.5% is a miss against the ≤10% bar they dated in January, even though a board slide could have printed “Good.” Honor the action: stop scaling paid acquisition; re-qualify ICP; do not fill the leaky bucket. Do not switch the denominator to end-of-year logos after the close so the percent looks kinder. Do not multiply a monthly 1% by 12 and call it the year. Keep next year’s sitting on the same definitions.
Write the tripwire before the first count. If the year misses the bar you wrote, honor the action. If it clears, continue to the next cheapest test — usually the next period on the same definitions, or a kill/pivot contract that now has a count to read. Neither a 3 / 24 nor a Yibud score predicts success.
Published bands
Gabe’s ICP, SMB, and non-ICP logo-churn bands — labeled, not a grade table
The table is what Gabe Miller-Smith published in Bessemer Atlas on 29 August 2023. Use it as vocabulary for a tripwire you write first, not as a law and not as a success certificate. Matrix typically spends less time on customer-churn benchmarks internally — these bands are segment-sensitive.
| Segment | Good | Better | Best |
|---|---|---|---|
| ICP (Gabe; upmarket sitting) | 10–15% | <10% | <5% |
| SMB / lower-CAC markets | 15–20% | 10–15% | 5–10% |
| Non-ICP, early stage | 15–20% normal / expected if the roadmap only solves a couple of use cases | 20–30% → take a closer look, even if non-ICP | Not published as a “best” band |
Source: Bessemer Atlas, “Understanding churn and building an action plan to fix the proverbial leaky bucket” (Atlas Editors; Gabe Miller-Smith; 29 August 2023). Gabe also: “if your churn is in double digits, there’s room to improve”; at Procore he never had a churn target greater than 9%; he is not panicking at 15%, but it signals org-wide work. This page does not invent a universal monthly logo-churn grade.
Five reason buckets — closed list, then patterns
Gabe’s typical causes: Price; Lack of adoption; Missing features; Switching to a competitor; Going out of business or merging. A drop-down beats an open text field if you want to see patterns. Year-one churn often means too much discounting at the outset or time-to-value too long. Years three and beyond often mean weak upsell / value delivery, or a misunderstood segment — including a willingness to downsell.
Procore case — cut first value from 60 days to 30
Gabe at Procore: first value took 60 days; the goal was to cut that to 30. Automating provisioning replaced a 7-day wait to grant access. Role-based certificates replaced generic training. A welcome drip covered the seven most important admin facts. Later, a renewal-prep drip started 120 days out. Those are process changes, not a success certificate.
Segment cuts — a blended 5–10% can mask a leaky slice
Gabe: if you are selling to ICP, blended 5–10% can look good until you cut by geography, vertical, segment, or spend. Sometimes the honest move is to stop selling a segment with no path to profitability. The piece’s North Star for CS leaders is net retention — read that sitting on the SaaS NRR page.
Matrix — track all three rates; 30 customers swing
Vaswani: track customer (logo), gross dollar, and net dollar. Dummy set of 30 customers — early-stage rates fluctuate until the book grows. Best-in-class gross dollar churn typically <1% monthly. Best-in-class net dollar churn is negative; Zendesk named as expansion-after-purchase. Series A may not yet have negative net churn. They typically spend less time on logo-churn benchmarks internally.
Compound, do not multiply — arithmetic, not a band
If you must turn a stable monthly rate m into an annual figure, use 1 − (1 − m)^12. A stable 1% monthly rate compounds to ≈11.4%, not 12%. That 1% here is Matrix’s best-in-class monthly gross-dollar ceiling used as an arithmetic input — not a “good logo churn” law and not permission to invent a 2% monthly logo table.
Scope notes, still the two sources: Bessemer Atlas, 29 August 2023 (Gabe bands, 5–25× mantra, $500M / 20% / $100M, Procore 60→30 / 7-day access / seven admin facts / 120-day drip, 5–10% blended mask). Matrix / For Entrepreneurs (three rates, 30-customer dummy, $100 vs $1,000 / $2,000 divergence, <1% monthly gross dollar, negative net dollar, Zendesk). Do not invent KeyBanc / ChartMogul / Recurly medians. Do not reproduce Bessemer Scaling GRR/NRR tables. Do not treat <5% or negative net dollar churn as proof you will succeed.
How to run it
How to run a logo-churn tripwire without faking scale
This is Yibud’s suggested sequence, not a Bessemer audit and not a Matrix spreadsheet. Honor the tripwire you wrote before the first count. If you cannot name the decision the rate would change, you are building a slide.
- 1
Write the tripwire before you look
Name the segment (ICP / SMB / non-ICP), the period (month / quarter / year), the definition (who counts as a logo; how trial becomes paid; voluntary vs involuntary), the threshold (Gabe’s matching band, or a tighter floor you justify in writing), and the action (stop paid acquisition / re-qualify ICP / cut time-to-value / keep experimenting / do not fill the leaky bucket with more top-of-funnel). Date the page. Do not rewrite the bar after the count.
- 2
Lock start count, lost count, and the clock
Denominator is logos at period start — always. State monthly versus annual. If you only have a quarter, report a quarter. Do not multiply a monthly rate by 12. If you need an annual figure from a stable monthly rate, compound with 1 − (1 − m)^12 and say so.
- 3
Show n lost / n start, then the percent
With tiny n, write 3 / 24 next to 12.5%. Matrix’s 30-customer dummy exists to show the swing. Do not read a “rate” from one lost logo of 12. Split ICP from non-ICP. A pretty blend that hides a leaky vertical is vanity with extra steps.
- 4
Bucket reasons with a closed list
Use Gabe’s five: Price; Lack of adoption; Missing features; Switching to a competitor; Going out of business or merging — plus optional other. Year-one pile-up → discounting or time-to-value. Year-three-plus pile-up → upsell / value / segment, including downsell. Do not leave the why as an open novel in a CRM note.
- 5
Honor the action. If you have no paying logos, use a different instrument.
If the checkpoint arrives and the count misses the bar, do the thing you named. Pre-revenue, there is no honest logo rate: use a smoke test, a willingness-to-pay sitting, interviews, or a usage retention curve if you already have free users. A Yibud score is not logo churn.
What you leave with
The logo-churn tripwire (copy this)
If the sitting produced a percent and no dated action, you did not finish. Fill the brackets. Leave nothing as “churn is fine,” “we kept the cheap logos,” or “we’ll know it when NRR arrives.”
The fill-in contract
Product: [name]. Segment: [ICP / SMB / non-ICP], written before the first count. Period: [month / quarter / year starting ____]. Hypothesis this rate can falsify: [we can scale this channel without a leaky bucket / these ICP logos stay]. Logo definition: [paid account that converted from trial on or before period start — not a free user]. Voluntary vs involuntary: [counted separately / labeled]. Signal: [n lost / n start, then %]. Threshold: [e.g. ≤ Gabe Better for ICP (<10%), or a tighter floor we justify here — not a homemade monthly 2% grade]. Action if we miss: [stop paid acquisition / re-qualify ICP / cut time-to-value / do not fill the leaky bucket]. Action if we clear: [keep going / next period on the same definitions]. Dollar check: [gross dollar churn on the same start base; NRR is a separate sitting]. A Yibud score does not fire this line.
The “≤ Gabe Better” line is Yibud’s suggested template from Gabe’s published ICP band — not a census, not permission to skip counts, and not permission to treat <5% as success. Rewrite the threshold before you measure if you have a reason (non-ICP 15–20% expected; SMB bands sit higher). Do not rewrite it after.
Example tripwire (Yibud’s suggested template, not a law)
If, for ICP logos in calendar year N, logos lost ÷ logos paying on 1 January is above 10% (Gabe Better), we do not 3× paid acquisition; we re-qualify ICP and cut time-to-value. Report n lost / n start. Do not blend ICP and non-ICP. Do not switch the denominator to end-of-year logos after the close. Pair with gross dollar churn on the same start ARR so cheap-logo retention cannot hide a whale cancel.
A 3 / 24 that clears 10% is still one year. A Yibud report score does not fire this tripwire. If you have no paying logos yet, this template is the wrong first instrument — use a smoke test, willingness to pay, interviews, or a usage retention curve instead.
The fork
When to use a logo-churn tripwire — and when not to
Logo churn is the wrong first instrument when nobody has paid you yet. It is the right instrument when paying accounts exist, you can name who counted as a logo at period start, and you are deciding whether to scale acquisition or fix the book.
Use it when you have paying logos and a scale-or-fix decision
You can name the start-of-period book, who is a logo, and a clock. You are about to 3× a channel, tell yourself “logo churn is 8% so we are fine,” or average ICP into non-ICP. That is when a dated ≤X% earns its keep. Optional: run NRR, a usage retention curve, or Quick Ratio as separate sittings — do not average them into a homemade “retention score.”
Do not start here when you have no paying logos
Pre-revenue? There is no honest logo rate. Do not invent a forecast cancel percent. Use a smoke test, a willingness-to-pay sitting, or interviews first. If you have free users but no paid book, a usage retention curve is the closer instrument. Still writing Stop / Pivot / Continue? Those are the decision contracts. This page supplies one count instrument those contracts can later read. A Yibud score is not logo churn.
Not the same page
Logo churn vs NRR, retention curves, Quick Ratio, LTV:CAC, vanity metrics, Yibud scores, and decision contracts
Several Yibud pages sit next to this one. Mixing them up turns dollar retention, a usage flatten, a same-period inflow÷outflow, a lifetime ratio, or a waitlist into “logo churn is 8%, so scale.”
NRR — dollar retention of a starting ARR base, not this account count
That page asks whether customers live at period start paid you more or less after churn, downsell, and expansion (NRR% / GRR%). This page asks how many accounts left. Ten $2K logos leaving can print a kinder logo-churn number than two $50K logos leaving. A 110% NRR does not tell you 3 / 24 left. A 3 / 24 does not tell you the starting ARR base retained 110%. Gabe’s North Star for CS leaders is net retention — read it there. This page does not reproduce Bessemer Scaling GRR/NRR band tables.
SaaS NRR →Retention curve — cohort usage over time, not paid cancels
That page plots, for users who started together, the share still doing the core action at day/week/month N. This page is paid / account cancellation count. A flatten at 35% weekly logs does not tell you 3 / 24 logos canceled. A 12.5% logo rate does not tell you the usage curve flattened.
Retention curves for startups →Quick Ratio — same-period inflow ÷ outflow, not a starting-count cancel rate
That page asks whether this period’s new + expansion outran churned + contraction (a multiple). This page is logos lost ÷ logos at start. A 4× Quick Ratio does not tell you 3 / 24 left. A 12.5% logo rate does not tell you this month’s dollar bucket is not leaking.
SaaS Quick Ratio →LTV:CAC and CAC payback — unit economics that churn feeds
Those pages ask whether a cohort’s contribution cleared a multiple of CAC, or how many months of contribution recover CAC. Logo churn is an input to lifespan, not the ratio and not the months. A 3.2 LTV:CAC does not tell you 3 / 24 left. A 12.5% logo rate does not tell you paid CAC pays back.
Vanity metrics — hygiene class, not this count tripwire
That page is the sorting hat: hits, cumulative sign-ups, and waitlist size feel like progress and do not tell you what to do next. A blended logo-churn percent with no start count, no segment, and no tripwire is vanity in a nicer font. Cumulative registered users are not n lost / n start.
Vanity metrics for startups →Yibud scores — not logo churn, and not a success prediction
The validation-score and score-calculator pages explain Startup MRI numbers from a deterministic rule engine on a one-line idea plus five questions. Those scores point at the weakest assumption to test next. They are not a logo rate, not NRR, not a flatten, and not a forecast. Do not read a mid-band score as “logo churn is fine,” and do not read <5% as a Yibud GO.
Kill and pivot — contracts that can read this count
Those pages write Stop, or rewrite one dimension, before hope edits the bar. This page can supply the count instrument those contracts later read: a dated n lost / n start against a bar you wrote first. An 8% slide with end-of-period logos in the denominator cannot fire a kill line honestly.
Where Yibud fits
Neither a low logo rate nor a Yibud score predicts success
Yibud is a free, no-signup startup idea validator. Scores come from a deterministic rule engine, not from a language model guessing success. Optional AI text, when it is used, only polishes prose. It does not invent the number. A Startup MRI report can name a weak demand or distribution assumption, then a Day-7 Continue / Refine / Re-test / Stop signal. The score is a flashlight on the weakest assumption to test next — not logo churn, not NRR, not a flatten, not a forecast. Logo-churn work sits next to that assumption: write the tripwire, do not hope an 8% slide means the book is healthy. If you already have a report, start with the named assumption, not the overall score. Then count the logos yourself. The method on this page stands alone if you never open the analyzer.
Analyze my idea →Common mistakes
What usually wastes a logo-churn week
Annualizing a monthly rate by multiplying by 12
A stable monthly rate m becomes 1 − (1 − m)^12, not m × 12. Matrix’s best-in-class gross-dollar ceiling is typically less than 1% monthly — keep the monthly label, or compound it and say you compounded. A 1% × 12 = 12% slide is arithmetic you made up.
Using the end-of-period count as the denominator
Matrix and this page use customers at start. An end-of-period or “average logos this month” denominator moves after the fact. If you change the denominator after the report, you changed the sitting.
Blending ICP and non-ICP into one company percent
Gabe publishes separate bands: ICP Good 10–15% / Better <10% / Best <5%; SMB Good 15–20%; non-ICP 15–20% expected, 20–30% look closer. A blended 8% can hide a 25% non-ICP row. Split the book the way you wrote the tripwire.
Celebrating low logo churn while whale accounts leave
Matrix’s dummy keeps $100 logos and loses $1,000 / $2,000 accounts. Pair logo churn with gross dollar churn on the same start base. A kind logo percent next to a 44% dollar hole is not a win.
Scaling acquisition on a leaky bucket
Gabe’s $500M / 20% / $100M example: the first $100M of new bookings only fills the hole. The 5–25× desk mantra is why you write stop-paid-acquisition before you raise the ad budget. Do not “fix churn” by buying more logos.
Treating <5% or negative net dollar churn as proof you will succeed
Gabe’s Best ICP band is <5%. Matrix’s best-in-class net dollar churn is negative. Clearing a bar you wrote means that period cleared a gate. It is not NRR, not a flatten, and not a Yibud GO. Neither a <5% logo sitting nor a Yibud report score predicts success.
Sources
Where these ideas come from
- Bessemer Atlas, Gabe Miller-Smith, “Understanding churn and building an action plan to fix the proverbial leaky bucket” (29 August 2023) — Used for the 5–25× desk mantra (attributed in the piece; no extra Harvard primary citation); $500M ARR / 20% / $100M leaky-bucket example; ICP Good 10–15% / Better <10% / Best <5%; SMB Good 15–20% / Better 10–15% / Best 5–10%; non-ICP 15–20% expected, 20–30% look closer; “double digits… room to improve”; Procore target never greater than 9%; not panicking at 15%; five reason buckets; year-one vs years three+; Procore first value 60→30 days, 7-day access, seven admin facts, 120-day renewal drip; blended 5–10% can mask a slice; North Star = net retention for CS leaders.
- For Entrepreneurs / Matrix Partners, Anoushka Vaswani, “Demystifying Churn: Measuring and Benchmarking this Metric” — Used for tracking customer (logo), gross dollar, and net dollar together; 30-customer dummy and early-stage swing; $100 logos kept vs $1,000 / $2,000 lost; voluntary vs involuntary; best-in-class gross dollar churn typically <1% monthly; best-in-class net dollar churn is negative; Zendesk expansion-after-purchase; Series A may not yet have negative net churn; they typically spend less time on customer-churn benchmarks internally.
In one paragraph
Summary you can quote
Logo churn (customer churn) = customers lost during a period ÷ customers at the start of that period. It counts accounts, not dollars. Write the tripwire before you look: for ICP logos in period N, stay ≤X%, or stop scaling / change ICP / change product. Gabe (Bessemer Atlas, 29 August 2023): 5–25× acquire-vs-retain mantra; $500M ARR at 20% churn loses $100M; ICP 10–15% / <10% / <5%; SMB 15–20% / 10–15% / 5–10%; non-ICP 15–20% expected, 20–30% look closer; Procore target never >9%; 60→30 day first value, 7-day access, seven admin facts, 120-day drip. Matrix: track logo, gross dollar, and net dollar; 30-customer dummy; $100 vs $1,000 / $2,000 divergence; <1% monthly gross dollar typical best-in-class; negative net dollar churn is best-in-class (Zendesk). Show n lost / n start. Do not multiply monthly × 12. Matching the gate is a decision signal. Neither a low logo rate nor a Yibud score predicts success.
FAQ
Questions founders actually ask
What is logo churn for a startup?
Customers lost during a period divided by customers at the start of that period — also called customer churn. It answers how many accounts left, not how many dollars left. Write the ≤X% tripwire before you look. It is not NRR, not a usage retention curve, and not a success score.
What is a good logo churn rate?
This page does not publish a universal monthly grade. Gabe (29 August 2023): ICP Good 10–15%, Better <10%, Best <5%; SMB Good 15–20%, Better 10–15%, Best 5–10%; non-ICP 15–20% normal if you only solve a couple of use cases, 20–30% take a closer look. He never had a Procore target greater than 9%; double-digit churn has room to improve; 15% is not a panic. Matrix typically spends less time on logo-churn benchmarks and instead watches gross and net dollar churn (best-in-class gross dollar typically <1% monthly; best-in-class net dollar is negative). Write your own tripwire.
How is logo churn different from NRR, a retention curve, and Quick Ratio?
NRR/GRR is dollar retention of a starting ARR base. A retention curve is whether a sign-up cohort still does the core action over time. Quick Ratio is same-period (new+expansion)÷(churn+contraction). This page is logos lost ÷ logos at start. They answer different decisions. Do not average them.
Can I read logo churn from 12 customers?
Not as a percent-only slide. Show counts (n lost / n start). Matrix’s dummy uses 30 customers to show the swing. One lost logo of 12 is not a stable rate. Do not scale paid acquisition on one row of 12. If you have no paying logos yet, use a smoke test, willingness to pay, interviews, or a usage retention curve first.
Is a Yibud score the same as logo churn? Does <5% predict success?
No and no. Yibud scores come from a deterministic rule engine on your idea inputs. They point at the weakest assumption to test next. They are not a logo rate, not NRR, not a flatten, and not a forecast. Neither a <5% logo sitting nor a Yibud report score predicts success. You do not need an account to use this page.
Write the logo-count tripwire before you scale the funnel
Yibud scores weak assumptions with a deterministic engine. You write segment, period, definition, threshold, and action while you are still honest — then a year of n lost / n start is a tripwire, not an 8% slide.
Generate a Yibud report →