Leak Calculator Methodology

1. The 8 dimensions

SignalScore scores every homepage across 8 dimensions of GTM effectiveness. Each dimension targets a specific failure mode that reduces visitor-to-customer conversion. The calculator uses your self-assessed score on each dimension to attribute leakage.

5-Second Verdict

Measures whether a first-time visitor can answer "what does this company do and why should I care?" within five seconds of landing. The failure mode is a headline that describes the product category or the mechanism rather than the outcome for the buyer. Most B2B homepages fail this test outright.

Story Arc

Measures whether the homepage tells a coherent narrative from problem to solution to proof. The strongest pages create a through-line where each section earns the next. Most pages present features in a random order and call it a homepage.

Mirror Test

Measures whether the copy speaks to the buyer's world or the vendor's world. Pages that use "I/we" language, feature-forward headers, and internal jargon fail. The buyer should see their own situation reflected back at them, not a product brochure.

Status Quo Tax

Measures whether the homepage articulates the cost of doing nothing. The strongest B2B sites quantify what happens if you don't act. Most don't. Urgency without consequences is decoration.

Safety Net

Measures how well the homepage reduces purchase risk. Trust signals, guarantees, integration assurances, and implementation support language all contribute. Missing this dimension makes buyers hesitate at the moment they're ready to act.

Proof Stack

Measures the quality and specificity of social proof. Logos are table stakes. Named customers, specific outcomes, and third-party validation are what actually move buyers. Generic five-star badges carry almost no weight.

Logo Test

Measures whether the page communicates meaningful differentiation from alternatives. If a competitor could swap their logo in and the page would still read correctly, the differentiation score is low. The best pages make substitution obvious and uncomfortable for the buyer.

The Close

Measures the conversion architecture at the end of the page. CTA placement, copy specificity, commitment gradient (high-intent vs. low-intent paths), and friction all factor in. A great page with a weak close leaks buyers at the last second.

2. The corpus

The calculator's benchmarks are based on 163 non-YC B2B homepages scored by SignalScore as of 2026-04-30. Each homepage is scored by SignalScore's automated pipeline across all 8 dimensions, producing an overall score (0-100) and per-dimension scores.

Why YC S25 is excluded: The YC S25 cohort is an early-stage benchmark population with different homepage maturity expectations. Including it would compress the top-quartile benchmarks and distort the leakage estimates for growth-stage companies. YC S25 rankings are published separately at /yc/s25/.

Per-vertical percentiles (q25, median, q75) are computed from the homepages in each vertical bucket. Verticals with fewer than 5 scored homepages are marked with display: false in the benchmark data and fall back to the General B2B SaaS corpus-wide distribution.

Sample sizes by vertical

Verticals shown here have display: true in the benchmark data (n ≥ 5). Verticals with small samples produce noisy q75 estimates; the calculator surfaces a warning when your selected vertical has n=5.

3. The single research claim driving the calculator

Published B2B homepage redesign case studies (CXL Institute, Marketing Experiments, Wynter, Unbounce) consistently report relative conversion lifts in the 10-30% range when companies move a homepage from below-median to top-quartile messaging quality. The typical "before" page in those case studies scores in the 30-45 range on the underlying messaging dimensions; the typical "after" scores in the 60-75 range. So the published lift is calibrated to roughly a 25-30 point gap closure, not to a theoretical 0-to-100 transformation.

We translate that research into a per-point lift coefficient by dividing the case-study lift range by the case-study gap range. That produces three defensible interpretations:

Sources: coefficient research pending peer review - see methodology version stamp. Primary sources include CXL Institute conversion research, Marketing Experiments B2B landing page studies, Wynter messaging research, and Unbounce conversion benchmark reports.

4. The math

The calculator runs five steps. Every input the reader provides maps to a variable below. The headline figure is computed from the overall-score gap; the per-dimension breakdown is an attribution of that total across the 8 dimensions.

We compute the gap between the user's score and the higher of (a) their vertical's top-quartile homepage in our corpus, or (b) a cross-B2B best-practice floor of 85. The floor exists because vertical-specific q75 values cluster between 40-65, but operators scoring above q75 still have meaningful pipeline leakage relative to truly best-in-class B2B homepages.

# Step 1: overall-score gap (measured against best-practice floor)
target         = max(q75_overall[vertical], 85)   # 85 = cross-B2B best-practice floor
gap_overall    = max(0, target - user_score)

# Step 2: total relative conversion lift (one research claim applied once)
total_lift     = gap_overall x lift_per_point[toggle]
# Conservative: 0.0033 | Moderate: 0.0070 | Aggressive: 0.0100

# Step 3: per-dimension attribution weights
# (impact_mid values are attribution weights only - not independent lift claims)
gap[d]              = max(0, q75[d] - user_score)   # for d in 1..8
weighted_gap[d]     = gap[d] x impact_mid[d]
total_weighted      = sum(weighted_gap[d] for d in 1..8)
share[d]            = weighted_gap[d] / total_weighted   # if total_weighted > 0

# Step 4: translate relative lift to monthly dollars
current_pipe   = visits x visitor_to_won x deal_size
leakage_month  = current_pipe x total_lift
leakage_year   = leakage_month x 12
leakage[d]     = leakage_month x share[d]

# Step 5: rounding pass
# Round headline and per-dimension rows to nearest $100.
# Force-correct the largest row by residual if sum != headline.

The rounding pass runs once at the end. Math.round(x / 100) * 100 is the JavaScript implementation. If the sum of per-dimension rows diverges from the headline by more than $100 (rare due to float precision), the largest row absorbs the residual so the breakdown reconciles exactly.

The per-dimension rows are sorted descending by dollar leakage. The dimension with the highest dollar attribution is labeled "Biggest single contributor" in the headline. This is not the dimension with the biggest raw gap - it is the dimension whose gap, weighted by its attribution importance, produces the largest share of the total.

Worked example

Inputs: Marketing Automation vertical, q75_overall=62, user_score=38, visits=27,500/mo, deal_size=$25,000, visitor_to_won=0.3%, Conservative toggle.

target         = max(62, 85) = 85
gap_overall    = max(0, 85 - 38) = 47
total_lift     = 47 x 0.0033 = 0.1551   (15.51% relative lift)

current_pipe   = 27,500 x 0.003 x $25,000 = $2,062,500/mo
leakage_month  = $2,062,500 x 0.1551 = $319,894 -> rounded to $319,900
leakage_year   = $319,900 x 12 = $3,838,800

Edge case - user_score=65:
  gap = max(0, 85 - 65) = 20   (was $0 when target was q75=62; now $136,100/mo)

Edge case - user_score=85:
  gap = max(0, 85 - 85) = 0 -> $0/mo leakage

5. Vertical-specific gap vs. cross-B2B impact reconciliation

There is a deliberate inconsistency in the math that careful readers will notice: the score gap is measured against your vertical's top-quartile benchmark (our data), but the conversion-lift coefficient is sourced from cross-B2B research averages (published case studies). These two inputs live at different levels of specificity, and combining them is a v1 approximation.

Why we do it anyway: per-vertical lift case-study data does not exist publicly today. We do not have post-fix conversion data from B2B cybersecurity homepage redesigns separate from B2B CRM redesigns. So we apply the cross-B2B coefficient to the vertical-specific gap as the most defensible available method, then default the calculator to Conservative to partially compensate for the approximation error.

What this means for your number:

• Verticals where homepage messaging carries above-average conversion weight (cybersecurity, where trust and Status Quo Tax messaging are primary buying levers) are likely under-stated by this math.
• Verticals where homepage messaging is a smaller part of the buying decision (enterprise infrastructure, where relationships and RFPs dominate) may be slightly over-stated.
• For most mid-market B2B SaaS verticals, the cross-B2B coefficient is a reasonable approximation.

Per-vertical lift coefficients are a v2 priority once we have post-fix conversion data from re-engaged customers. Until then, the methodology version stamp lets returning readers know which coefficients produced the number they saw previously.

6. Dimension attribution weights

The attribution weights (called impact_mid) are derived from cross-B2B research on which homepage dimensions most strongly correlate with conversion-rate differences. They survive in the v1 math purely as weights for the breakdown table - they do not independently drive the headline figure.

These weights sum to 59.5% rather than 100% because they are cross-B2B research midpoints drawn from ranges, not a closed-form allocation. The math uses them only as relative weights: each dimension's weighted gap is divided by the sum of all weighted gaps to produce a share. The absolute values of the weights do not affect the final attribution - only their relative ratios matter.

7. Caveats

The calculator does not model the following, and in some scenarios these factors will dominate the homepage messaging gap:

• Traffic mix and demand-gen quality. A homepage converting 0.3% of cold paid traffic and 0.3% of warm brand search traffic are very different situations. The calc uses a single blended rate. If your traffic is primarily high-intent, you're probably already outperforming the cross-B2B baseline.
• Sales process, product-market fit, seasonality, competition. None of these are modeled. A perfectly scored homepage attached to a product the market doesn't want produces zero conversion.
• Diminishing returns at extreme score gaps. The math is linear. Real-world messaging improvements have compound and saturating effects we don't model. A user scoring 10 against a q75 of 70 produces a 60-point gap and a very large number at Aggressive. Treat extreme outputs as directional, not literal.
• Per-vertical impact differences. Covered in section 5. The cross-B2B coefficient is an approximation for all verticals.
• Noisy benchmarks at n=5. Verticals with exactly 5 scored homepages have wide error bars on q75 (roughly +/-15-20 points at 95% confidence). A single outlier homepage shifts the benchmark materially. The calculator surfaces a warning for these verticals.
• Self-score ego bias. Visitors consistently rate their own homepages higher than external scorers do. The Conservative default partially mitigates this - a generous self-score combined with a 0.33% coefficient produces a number closer to reality than Aggressive would. The best fix is a real SignalScore from the free-score flow.

8. Update policy

Benchmarks refresh monthly via scripts/generate_calc_benchmarks.py. The script pulls all non-YC scored homepages from Supabase, recomputes percentiles per vertical, validates the output against the committed JSON Schema, and writes the updated benchmarks.json to the deploy directory. The last_refreshed date and corpus_size in the metadata strip above update automatically on each refresh.

Methodology coefficients (the 0.33% / 0.70% / 1.00% per-point figures) change only via a formal revision to this design document. Each revision increments methodology_version. The version stamp appears in the calculator's "show your work" expansion, so a returning reader who sees a different number can check whether it changed because new benchmark data came in (same version, different q75) or because the underlying math changed (new version).

Current methodology version: 1.0.0. Current benchmarks as of: 2026-04-30.

Ready to run the numbers on your own homepage?

Run the Leak Calculator