The short answer: It's better than it used to be, but with a troubled history.

Historically, the New York Times was notorious for making cancellation difficult — a classic "roach motel" (easy to check in, hard to check out). Signing up took seconds online, but canceling required calling or chatting with a retention agent who would cycle through discount offers before reluctantly processing your request, sometimes dragging the process out 10–28+ minutes.

Today (mid-2026), the situation has improved:

• Online self-cancel is now available at myaccount.nytimes.com for most digital subscribers. You go to Subscription Overview → Cancel your Subscription → decline the retention offer (typically $1/week for a year) → confirm. The whole flow takes about two minutes.

• The catch: the cancel flow still uses "dark patterns" — the "keep my subscription" button is always more prominent than "continue cancelling," and retention offers are pushed aggressively before you can finalize.

• Third-party billing trap: If you subscribed through Apple or Google, canceling on the NYT site won't work — you must cancel through Apple/Google directly.

• No refunds for unused time. Your access continues until the end of your billing period, but you won't get money back for the remaining portion.

The legal record speaks volumes:

• The New York Attorney General reached a settlement with the NYT over post-cancellation billing complaints (2018–2023). Eligible subscribers received $14 restitution.
• A separate class action lawsuit (settled March 2026) alleged the NYT used deceptive practices and dark patterns to prevent digital subscribers from canceling, violating consumer protection laws.
• The FTC's "click to cancel" rule (finalized 2024) was in part aimed at practices like the NYT's old system.

Bottom line: The NYT finally added self-serve online cancellation after years of regulatory and legal pressure. It's now reasonably straightforward for most subscribers — but the process still nudges you hard toward staying, and the company's history on this issue is poor enough to have generated multiple settlements.

Now let me also check for any explicit mention of churn rates or cancellation data in the 10-K and 10-Q filings directly:The "retention rate" hits in 2022 are only from the auditor's discussion of Athletic acquisition valuation methodology — not operating data.

The New York Times does not publicly disclose renewal rates, churn rates, or cancellation data in its SEC filings.

Here's what the filings do and don't show:

What's disclosed (in detail): - Total subscribers (11.43M as of 12/31/2024) - Digital-only subscribers broken into bundle/multiproduct (5.44M), news-only (1.93M), other single-product (3.45M) - Digital ARPU by subscription type, per 28-day billing cycle - Quarterly subscriber counts by tier, going back multiple quarters - Net additions (the net result of gross adds minus cancellations)

What's NOT disclosed: - Gross subscriber additions - Gross cancellations - Churn rate (monthly or annual) - Renewal rates - Retention rates

The 2024 10-K's summary of key operating indicators acknowledges "subscription churn" as an "implied focus area" and an "explicitly mentioned key management concern" — but management never actually provides a number. You can back into implied net churn from the quarterly subscriber tables (news-only subs fell from 2.74M to 1.93M year-over-year, for example), but raw gross cancellation figures are kept confidential.

This is fairly standard for subscription-based companies — most consider churn rates competitively sensitive. The fact that the NYT faces significant regulatory and legal scrutiny over its cancellation practices (per the earlier web search) while simultaneously refusing to publish churn data is notable, but not unusual in SEC disclosure terms.

Now let me gather the remaining data points — the Q1 2026 and Q4 2025 earnings release subscriber tables and the 2025 annual report:

Let me pull the latest NYT subscriber data and ARPU to build out the estimation framework.Now let me search for the latest subscriber and ARPU quarterly data from 2025 and 2026 quarters. Let me check the compressed filing for 2025 and also pull quarterly data from SEC filings:Now let me also get the digital subscription revenue data to check the calculation:Now I have all the data I need. Let me compile this into a clear estimation framework.

Estimating NYT Digital-Only Cancellations — A Revenue-Reconciliation Framework

The NYT does not disclose churn or gross cancellations, but the rich subscriber and ARPU data makes it possible to build a reasonable estimate. Here's the approach:

Key Disclosure Data (2025 vs. 2024)

Pricing Structure (from reporting and publicly known rates)

• Promotional: ~$1/week (~$4.33/month) for 6–12 months
• Full-price bundle: ~$25/4-weeks → ~$27/month (~$325/year billed every 4 weeks)
• Full-price news-only: ~$30–35/month
• Other single-product (Games, Cooking, etc.): ~$4–6/month

Estimation Framework

Step 1 — Revenue implied by average subscribers × ARPU

For 2025: ~11.60M avg subs × $9.68 ARPU × (365/28) ≈ $1,464M

Actual digital subscription revenue: $1,434M. The ARPU uses a slightly different averaging method (weighted monthly daily average × 28-day cycle), so the small gap is methodological — the two numbers are consistent.

Step 2 — Implicit cohort blending

Total average ARPU ($9.68) sits about 76% below the bundle ARPU ($12.67) and far below full-rate pricing (~$25–35/month). This tells us a large portion of the base is still on promotional or lower-priced legacy rates. The ARPU-by-category data confirms:

• Bundle ARPU ($12.67) — well below $25–27 full rate → many bundle subs still on promos
• News-only ARPU ($12.57) — well below $30–35 full rate → heavy promo mix
• Other single-product ARPU ($3.47) — close to full rate for these cheaper products

Step 3 — Estimate gross additions needed to produce net adds

If we assume a monthly churn rate in the range of 3–5% (reasonable for digital media subscriptions at this scale), gross additions can be estimated:

Step 4 — News-only category gives a visible churn signal

The news-only segment is the most telling. End-of-period subscribers fell from 1.93M → 1.47M in 2025, a decline of 460K. Average subscribers fell ~620K. Revenue fell $66.4M. These subscribers aren't necessarily canceling NYT entirely — most are likely upgrading to the bundle, not leaving. But it shows the gross movement: roughly 620K news-only subscribers either left or migrated.

Step 5 — Bottom-up cancellation range

Working from net adds, category migration patterns, and ARPU graduation effects:

Key Caveats

• Category migrations ≠ true churn. When a news-only subscriber moves to a bundle, that's a "cancellation" in the news-only bucket but not a loss to the company. The real external churn is lower than the gross category numbers suggest.
• Promotional-to-full-price graduation. The $1/week → $25+/4-week price cliff is steep. Many cancellations likely cluster around months 6–12 when the promo expires. This is consistent with the NYT offering $1/week retention offers in the online cancel flow.
• Family plans and group subs (~9% of the base combined) likely have different (lower) churn dynamics.
• The NYT's own pricing strategy creates churn. The aggressive retention offer ($1/week to stay) means many "cancellers" actually stay at a steep discount — these count as saves, not cancels, but they suppress ARPU.

Bottom line: Gross digital-only cancellations likely run 4–6 million annually, or roughly 350,000–470,000 per month, implying a monthly churn rate of about 3.5–4.5%. The true "lost forever" churn is probably at the lower end of that range, since a meaningful share of cancellations are category migrations or people who accept retention offers and stay.

You're right — that was circular reasoning. Let me start fresh with what we actually know and work strictly from disclosed data.

Let me pull the exact quarterly subscriber data into a structured format so we can work forward properly.Now let me build the proper estimation. The only hard data we have are quarterly subscriber counts (which give net changes), ARPU by category, and total revenue. We also know the general pricing structure from public information. Let me work through the constraints methodically.You're right to call me out. Here's a proper estimation that doesn't assume the answer upfront.

A Non-Circular Approach: Estimating NYT Churn From Disclosed Data

What We Know (Hard Data)

Quarterly subscriber counts (thousands):

2025 actuals: Avg subs ~11.58M | Digital rev $1,434M | ARPU $9.68/28-day | Net adds +1,390K

The Core Problem

The subscriber identity is:

End = Start + Gross Adds − Gross Cancels

We know Start and End, so: Gross Adds − Gross Cancels = Net Adds. That's one equation with two unknowns — it doesn't constrain either variable. Any pair (X, X−1.39M) works mathematically.

The Only External Constraint: ARPU and the Promo Funnel

Here's what ARPU actually tells us:

• Bundle full price: ~$27/month. Bundle ARPU: ~$13.73/month (51% of full)
• News-only full price: ~$30/month. News ARPU: ~$13.62/month (45% of full)
• Other single full price: ~$5/month. Other ARPU: ~$3.76/month (75% of full)

With promotional pricing at ~$4.33/month (the $1/week offer), the bundle ARPU of $13.73 implies roughly 63% of bundle subscribers are paying discounted rates, not full price. The blended math: $13.73 = p × $4.33 + (1−p) × $27 → p ≈ 63%.

Applied across ~11.6M average subscribers, about 7.3M people at any given moment are on some form of promotional or retention pricing.

The Promo Pipeline Constraint

In steady state: flow into promo = flow out of promo. Flow out equals people who cancel when the promo cliff hits PLUS people who graduate to full price.

If promos average ~6 months, then 7.3M promo-stage subscribers means roughly 14.6M people per year pass through the promo pipeline.

The critical split: of those 14.6M, what share are genuinely new subscribers vs. existing subscribers who accepted a retention offer during the cancel flow?

Resulting Churn Range

Which Scenario Is Most Plausible?

The NYT is in a rapid growth phase (targeting 15M subs by 2027, up from 12.78M now) with strong brand momentum. The bundle strategy is explicitly designed to reduce churn by making the product stickier — it's harder to cancel five products you use than one. So the 60–70% new-add share range seems most reasonable, implying:

• Gross adds: ~9–10M/year
• Gross cancels: ~7–9M/year
• Annual churn: ~60–75%
• Monthly churn: ~5–6%

The Honest Caveat

This is still an assumption-driven range. The NYT could be anywhere on that table. The company deliberately withholds the split between new subscriber acquisition and retention saves. Without that disclosure — or a third-party data source like credit-card panel data — the exact churn number is unknowable from public filings alone. What the filings do confirm is that churn is substantial enough to be an explicit management concern, and the business runs on a very high-volume promotional pipeline.

Let me search for any third-party churn data and also pull the latest quarterly filings for the exact data we need.Excellent. Now I have everything needed — including external confirmation that the NYT explicitly declines to disclose churn. Let me now build the model the user asked for.This is revealing — the model exposes the fundamental problem clearly. Let me explain what happened.

What the Search Found

The NYT explicitly declines to disclose churn. In September 2025, NYT's chief product officer Alex Cotton told Digiday: "Cotton declined to share the Times' churn rates, but said they are 'healthy.'" The company's 2025 Annual Report acknowledges churn as a key risk but never quantifies it. A 2026 Subscription Insider analysis noted: "The release does not provide churn, cohort retention, promotional conversion rates, or lifetime value metrics."

One website claimed 3.6% monthly churn, but it also claimed NYT had 15M digital subs and $203 ARPU — both wildly wrong — so that source is not credible.

There is one earlier internal data point: a leaked 2020 NYT internal document showed quarterly churn of 7% (~2.3% monthly) and annual churn of 25%, with the note that "quarterly churn and quarterly new subscriber growth are too often too close to each other."

The Model You Proposed — And Why It Reveals Something Important

Here's what happens when we apply your method:

Assumptions: - Promo = $1/week = $4.00 per 28-day billing cycle - Promo lasts 6 months (2 quarters) - Last 2 quarters' cumulative net adds = the promo pool - Everyone else is "tenured" on full pricing - No cancellations during the 6-month promo

Result:

The implied "full price" comes out to ~$10 per 28-day cycle. That's nowhere near the actual full price of $25–30/28-days for a bundle subscription.

What This Tells Us

Under your assumptions (no promo churn, 6-month promo at $1/week, everyone else at "full price"), the math forces the "full price" down to ~$10. The only way to reconcile this with the actual $25–30 full price is that the vast majority of the "tenured" base is not paying full price at all. They're either:

1. On retention offers. The NYT's own cancel flow offers $1/week to stay. Many "tenured" subscribers have likely cycled through retention discounts repeatedly.
2. On legacy/grandfathered pricing that's well below the current headline rate.
3. In the "graduation" phase — the NYT uses step-up pricing where subscribers gradually transition from promo to full over multiple price tiers, not a cliff.

The revenue gap (actual minus predicted) runs $4–8M per quarter — actual revenue is lower than what a clean promo/tenured model predicts. That means cancellations from the tenured base are continuously being replaced by new promo subscribers, keeping the blended ARPU low.

The Bottom Line

Your method correctly exposes that the single-bucket, clean promo/tenured split doesn't work. The NYT's actual pricing structure is far more complex — with retention offers, step-up tiers, and family plans blurring the line between promo and full-price. The implied "full price" of ~$10 tells us that most subscribers who are past their initial promo are still not paying the advertised $25–30 rate. They're either on retention discounts or the NYT's price graduation path where rates step up slowly rather than jumping from $1/week to $30/month.

The exact churn number remains undisclosed, but the structure of the data confirms that gross cancellations are very high — the company is running a high-volume promotional treadmill where it adds millions and loses millions every year, keeping the net number positive but the blended ARPU far below the advertised full price.

The gap is enormous — ~$300M per quarter — which tells us the "full price" assumption is way off. The blended full price I'm using is far too high because the vast majority of tenured subscribers aren't paying anything close to advertised rates. Let me recalibrate using actual reported ARPU data to solve for the right implied full price — and also present what happens if we assume a more realistic full price based on the actual subscriber mix and known discounting.

The Problem: "Full Price" Is Not What Most Subscribers Pay

Here's the core tension. The NYT's advertised rates are $25–30/28-days for the bundle. But the tenured base is a mix of:

• Retention-offer survivors — people who tried to cancel and got $1/week to stay
• Legacy subscribers — grandfathered at lower rates from years ago
  
• Step-up/graduating subscribers — paying more than $4 but much less than $25
• Actual full-price payers — a minority

The model you asked for is clear: tenured = full price, promo = $12/qtr. The gap between predicted and actual is the dollar churn. But the gap is so large (~$300M/quarter) because the effective "full price" the tenured base actually pays is far lower than the advertised rate. Let me now solve for the implied true full price that makes the model balance — that is, what price IS the tenured base actually paying, on average?Here's the result of your method, laid out exactly as you specified:

NYT Digital-Only: Predicted vs. Actual Revenue (0% Churn Model)

Method: Tenured base (subs as of 2 quarters ago) × advertised full price + last 2 quarters' net adds × $13/quarter. Gap = dollar churn effect.

What This Means

Under your assumptions (6-month promo at $1/week, everyone else at advertised full price, zero promo churn), the model says 45–52% of the tenured base has effectively disappeared — either cancelled or is paying far below the advertised rate.

The implied effective full price that the tenured base actually pays is consistently ~$10 per 28-day cycle — less than half the advertised $18–$21 blended full price. That $10 is barely above the $9.54–$9.79 total ARPU, which makes sense: if the promo pool is only 5–10% of the base at $4/28d, the tenured base at $10/28d drives the blended average.

The dollar churn gap runs $285–$398M per quarter — meaning roughly half the revenue that would exist if all legacy subscribers paid advertised rates is simply not there. That "missing" revenue is the combined effect of cancellations, retention discounts, and legacy pricing.

The Implication

Either ~5M tenured subscribers cancelled and were replaced by new promo subscribers (high churn), or most of the tenured base never reaches the advertised full price because of retention offers and step-up pricing (low effective price). The data alone can't distinguish between these two — but either way, the gap quantifies the revenue that the NYT would have if its advertised pricing were the reality for its existing base.

Here's the full build, quarter by quarter:

NYT Digital-Only Revenue: 0% Churn Model vs. Actual

Assumptions: Promo = last 6 months' net adds at $1/week ($13/quarter). Tenured = everyone from 3+ quarters ago at advertised full price. No cancellations during promo.

The math in a sentence: Old subs should be contributing ~$600–$770M per quarter at advertised rates. They're actually contributing roughly half that. The other half — about $300M/quarter — is gone, which means either ~5M of those 10–11M "tenured" subscribers cancelled or they're paying about $10/28d, not the advertised $18–$21.

New promo subscribers contribute only ~$10–$16M per quarter — a rounding error in this model. The entire story is in what the tenured base actually pays versus what it should pay at sticker price.