Unit Economics for Pre-Seed: When LTV/CAC Lies and How to Tell

The 3:1 LTV/CAC rule was popularised by venture investors looking at mature SaaS portfolios where cohorts had years of retention data and channels had stabilised. Pre-seed founders inherit the rule without inheriting any of those preconditions. The result is a metric used as a green light when it has no resolving power, and as a red light when it reflects sample noise rather than a real economic signal. The four lies below explain where the math fails and which numbers replace it before you have the data to compute LTV honestly.

1. The textbook formula and what it pretends about

The standard pre-seed founder encounters the formula in this shape:

LTV = (ARPU × Gross Margin) / Monthly Churn Rate

LTV / CAC ≥ 3

This compresses several non-trivial assumptions into a single number. Monthly churn is a single stable rate. Gross margin is constant across cohorts. CAC is a single number across channels. Future revenue is worth the same as today's revenue. The cohort you observed yesterday will behave like the cohort you acquire tomorrow. Every assumption is provisional pre-seed. Stacked together, they generate enough error to invalidate the ratio entirely. The honest version adds a discount rate, segments by cohort age, and reports a confidence band rather than a point estimate ^[2]. Pre-seed, the data to compute even the honest version doesn't exist yet.

2. Lie #1: retention curves you don't have yet

The denominator of the LTV formula, monthly churn rate, implies a retention curve that has stabilised. In real cohorts, churn is front-loaded: the first 90 days see disproportionate drop-off, then retention flattens into a long tail. Fader and Hardie's BG/NBD models explicitly need observation windows long enough to disentangle the early-dropout rate from the steady-state rate ^[1]. Below 12 months of cohort data, the steady-state rate is essentially unidentified.

Pre-seed founders typically extrapolate from 3 to 6 months of data. If aggregate monthly churn looks like 3% in month four, the implied lifetime is 33 months. If the real cohort decomposition is 8% in the first quarter and 1.5% thereafter, customers who survive the early period have a 67-month expected lifetime. Same data, two answers, both technically defensible. Plug either into LTV/CAC and you over-invest in acquisition by ~2x or starve a viable business.

The practical correction at pre-seed is to refuse to publish a single LTV. Report a range bounded by aggregate-churn LTV (optimistic) and last-30-days churn LTV (pessimistic). When the range spans 3x CAC to 1x CAC, the metric is not decision-useful and you should stop using it ^[1]. The customer lifetime value calculator exposes both numbers so you can see the spread before committing to either.

3. Lie #2: CAC ignores blended-channel reality

Pre-seed CAC is almost always a blended number: one part founder-network warm intros at zero marginal cost, one part SEO content already written, one part paid ads on a small budget, one part referrals. Each channel has different unit economics, different ceilings, different scaling behaviour. The blended CAC tells you what the average customer cost looked like when most volume came from the cheapest channel.

The trap is that cheap channels do not scale linearly. Founder warm intros run out at 30 to 80 customers depending on network size. Referrals scale with the customer base, not with effort. SEO content compounds slowly. The only channel that scales on demand is paid acquisition, and paid CAC is usually 3 to 10x the blended number. So a 4:1 LTV/CAC computed against blended CAC becomes a 1.5:1 ratio against the marginal CAC of the channel you would actually scale into.

Compute CAC by channel and use the marginal-channel CAC for any decision about scaling spend. Blended CAC is fine for a board update; it is wrong for a budget allocation. The CAC calculator handles per-channel inputs so you can see which segment of the blended figure is actually load-bearing. For ad-driven channels specifically, pair it with the ROAS calculator to confirm payback before scaling.

4. Lie #3: pre-seed cohorts are too small to be stable

Churn rate is a sample statistic with a confidence interval. The width depends on the number of customers in the cohort and the number of observed events. Pre-seed founders quote churn rates from cohorts of 20 to 50 customers and treat them as point estimates. They are not.

A useful approximation. With N customers in a cohort and observed monthly churn rate p, the standard error on the rate is roughly sqrt(p × (1 − p) / N). For N = 47, p = 5%, the standard error is 3.2%. A 95% confidence interval runs roughly from −1% to +11%. The true churn could be anywhere in that band, which means the implied lifetime ranges from 9 months to "infinity, no churn observed yet." This is the situation every pre-seed founder is in.

The minimum cohort size for a stable churn estimate depends on the rate. Rule of thumb: 20 churn events to start getting a useful estimate, 50+ events for the confidence interval to narrow enough that decisions hang on it. At 5% monthly churn, that's 400+ customer-months of observation, meaning ~150 paying customers tracked for 3 months, or ~50 customers tracked for 8 months. Below that threshold the LTV calculation is sample noise dressed as analysis.

This is a counsel of using the right metric. Below the cohort-size threshold, anchor on payback period (sum actual cash) rather than LTV (extrapolate a curve you can't yet see). More on that in section 7.

5. Lie #4: the discount rate matters more than founders think

The textbook LTV formula treats every future dollar as equivalent to a dollar today. For a 33-month implied lifetime, this overstates LTV by 20 to 35% at typical bootstrapped-business discount rates of 15 to 25% annually ^[2]. The longer the implied lifetime, the larger the haircut.

The discounted LTV formula:

LTV = Σ [(Margin per period) × (1 − churn)^t] / (1 + r)^t

For pre-seed founders, the right discount rate is the higher of (a) cost of capital, or (b) the opportunity cost of capital deployed elsewhere. A bootstrapped founder facing the choice between deploying $1 into acquisition or $1 into product engineering should use the marginal return on engineering hours as the floor. That number is rarely under 20% annualised for early product work, meaning a 33-month nominal LTV of $3,000 is closer to $2,250 in present value. That 25% adjustment alone flips a 3:1 ratio into a 2.25:1 ratio.

The discount rate also kills the case for very long-tail LTV claims. Anything beyond month 60 contributes little to present value. Founders quoting "lifetime value over seven years" are inflating a number that, properly discounted, mostly resolves in the first 36 months anyway.

6. A pre-seed-honest LTV/CAC, with confidence intervals

If you must compute LTV/CAC pre-seed, compute it as a range and propagate the confidence intervals from each input. The structure:

Inputs (with confidence intervals):
  ARPU                = $X ± SE(X)
  Gross Margin        = G ± SE(G)
  Monthly Churn       = p ± SE(p), where SE(p) = sqrt(p(1-p)/N)
  Discount Rate       = r (annual)
  CAC by channel      = {C_paid, C_organic, C_referral, ...}

LTV (low)   = (ARPU × G_low)  × Σ_t (1 - p_high)^t / (1 + r)^t
LTV (mid)   = (ARPU × G_mid)  × Σ_t (1 - p_mid)^t / (1 + r)^t
LTV (high)  = (ARPU × G_high) × Σ_t (1 - p_low)^t / (1 + r)^t

Ratios reported:
  LTV (low)  / CAC_marginal     ← decision floor
  LTV (mid)  / CAC_blended      ← headline
  LTV (high) / CAC_organic-only ← optimistic ceiling

This framework forces three ratios instead of one. The decision floor (pessimistic LTV against the CAC of the channel you would scale next) is the number that matters for budget calls about acquisition spend. The headline is the number for board updates. The ceiling is for sanity checks; if even the ceiling looks bad, the unit economics are not salvageable through optimisation.

Most pre-seed businesses with healthy unit economics produce a decision floor between 1.5:1 and 2.5:1 and a ceiling between 4:1 and 8:1. A point estimate of 3:1 is neither here nor there until you show the spread.

7. Three numbers that survive volatility

Until cohort data stabilises, three metrics give you decision-relevant signal without extrapolating a retention curve.

Payback Period. Months for cumulative gross-margin contribution from a customer to recover CAC. Formula: Payback = CAC / (ARPU × Gross Margin). Depends on cash you have actually observed, no extrapolation. OpenView's 2024 benchmarks put median CAC payback at 18 to 24 months for mid-market B2B SaaS, 12 to 18 for SMB ^[3]. For a bootstrapped business, anything over 12 months is a runway risk regardless of LTV/CAC. The CAC payback calculator handles the gross-margin adjustment that most back-of-envelope versions skip.

Magic Number. Bessemer's sales efficiency metric: Magic Number = (Net New ARR × 4) / Sales & Marketing Spend in the prior quarter. Above 1.0 means you're getting back more than you're spending within a year: invest more. Between 0.5 and 1.0: cautious optimisation. Below 0.5: stop scaling spend until conversion or retention improves ^[5]. Magic Number requires no LTV assumption and aggregates across channels and customer types.

Net Dollar Retention. NDR measures revenue retained from a starting cohort, including expansion and contraction, after a fixed period. Formula: NDR = (Starting MRR + Expansion − Churn − Contraction) / Starting MRR. Healthy SaaS shows NDR above 100%: expansion outpaces churn even before new logos. Below 90% means the funnel is leaking faster than expansion can compensate, and growth from new acquisition is masking a retention problem ^[3][4]. NDR sidesteps cohort-size issues because it works on dollars, not customers, and gives you a leading indicator long before churn rates stabilise.

If all three are healthy (payback under 18 months, Magic Number above 0.7, NDR above 100%), unit economics are working regardless of LTV/CAC. If one is bad, fix it before optimising the others. If all three are bad, no amount of LTV math will save you.

8. Worked example: 47 users, 6 months, $50 ARPU

A representative pre-seed scenario. The product is a vertical SaaS at $50/month ARPU with 75% gross margin. After 6 months it has 47 paying users, of whom 5 have churned. CAC over the period was $4,200 of paid spend plus $1,800 estimated value of founder time on outbound, divided across 47 acquisitions for a blended CAC of $128. The paid-only CAC across the 22 customers from ads was $191.

Observed inputs:
  ARPU                  $50/month
  Gross Margin          75%
  Customer-months obs.  ~155 (47 customers, average ~3.3 mo each)
  Churn events          5
  Observed monthly churn p = 5 / 155 = 3.2%
  SE(p)                 sqrt(0.032 × 0.968 / 155) = 1.4%
  95% CI on p           ~0.5% to 6.0%

Optimistic LTV (p = 0.5%):
  Lifetime              200 months → cap at 60 (discount kills tail)
  Discounted LTV @ 18%  ~$1,840

Midpoint LTV (p = 3.2%):
  Lifetime              31 months
  Discounted LTV @ 18%  ~$880

Pessimistic LTV (p = 6.0%):
  Lifetime              17 months
  Discounted LTV @ 18%  ~$540

Ratios:
  Optimistic LTV / blended CAC   $1,840 / $128 = 14.4x  (ceiling)
  Midpoint   LTV / blended CAC   $880   / $128 = 6.9x   (headline)
  Pessimistic LTV / paid CAC     $540   / $191 = 2.8x   (decision floor)

Three-numbers check:
  CAC payback (paid)    $191 / ($50 × 0.75) = 5.1 months
  Magic Number          (assume 22 paid acquisitions × $50 × 12 / $4,200) ≈ 3.1
  NDR (6-mo)            ~95% (5 churned, no expansion yet, no upgrades)

The point estimate LTV/CAC of 6.9x looks excellent. The decision floor of 2.8x is solid but not spectacular. The 95% confidence interval on monthly churn alone produces a 3.4x spread in headline LTV. The three-numbers view tells the cleaner story: payback is fast (good), Magic Number is high (good, keep spending on this channel), NDR is below 100% (warning, expansion mechanics are missing). The action item is not "scale paid acquisition based on a 6.9x LTV/CAC." It is "build a usage-based upgrade path or expansion seat structure to lift NDR above 100% before the cohort gets large enough that early-period churn dominates the economics."

None of that comes from the LTV/CAC ratio. All of it comes from the three-numbers view and the confidence intervals.

9. Which metric to use at which stage

The metric to rely on changes as data thickens.

• 0 to 50 paying users, <6 months. Use payback period only. Cohort is too small for any retention-curve estimate. Watch week-4 retention as a leading indicator. The profit margin calculator matters more here than any LTV tool because gross-margin discipline is the input that makes payback feasible at all.
• 50 to 150 paying users, 6 to 12 months. Add Magic Number, start tracking NDR. Compute LTV as a range, not a point. Use the decision-floor ratio (pessimistic LTV / marginal CAC) for budget choices. The SaaS pricing strategy calculator becomes useful for testing whether a price change moves the payback metric without breaking conversion.
• 150 to 500 paying users, 12+ months. Cohort decomposition becomes feasible. Compute LTV by cohort age band (0 to 90 days, 90 to 365 days, 365+ days). The headline LTV/CAC starts having signal. Continue tracking the three numbers as the primary scoreboard.
• 500+ paying users, 18+ months. Full LTV/CAC framework with confidence intervals, cohort segmentation, and channel-level attribution. The textbook 3:1 rule becomes a useful test rather than a fiction.

The mistake is using stage-4 metrics at stage 1. Pre-seed founders citing 4:1 LTV/CAC ratios from 30-customer cohorts are pattern-matching to investor expectations, not analysing their business. The investor knows the number is noise. The founder usually doesn't. "Payback is 9 months, Magic Number is 1.4, NDR is 102%, sample is too small for a stable LTV but the three numbers point the same direction" is a stronger pitch than a fabricated ratio, because it shows you know which numbers are real.

As of 2026-Q2, the standard-of-care for pre-seed unit-economics reporting: lead with payback period and Magic Number, report NDR with at least 6 months of cohort data, and treat LTV/CAC as a directional check with explicit confidence bands rather than a single number. Founders who run the math this way make better acquisition-spend decisions and survive longer when one input drifts.