LeadersToolset

Methodology

Last updated: July 9, 2026

LeadersToolset promises that every figure in a report is itemised — inputs, assumptions, and the maths between them. This page explains how the calculators work, which constants the models rely on, and where every benchmark we ship comes from, so you can audit the numbers rather than take them on trust.

How the calculators work

Each calculator takes the figures you enter and passes them through a pure, versioned formula. Every formula is a standalone, tested module with golden tests: a fixed set of inputs must always produce the same outputs, and any change to a formula requires a version bump — outputs never change silently.

The formula produces three scenarios (see below), and the report presents the working as an itemised, receipt-style ledger. Every line in that ledger is traceable either to a figure you entered or to an assumption the report states explicitly. There are no hidden adjustments between your inputs and the headline numbers.

Scenarios

Projection reports (forecasting before you invest) run three scenarios: Conservative at 70% of the modelled improvement, Moderate at 100%, and Optimistic at 130%. Impact reports (measuring after implementation) run a tighter ±10% band, framed as measurement variance rather than forecast uncertainty.

To be plain about it: these spreads are editorial model defaults, not statistical confidence intervals. We do not have the data to compute a real confidence interval for your organisation. What you get instead is transparency — because the full delta-driven maths is itemised, you can see exactly how sensitive the result is to the spread and re-run the model with figures you trust.

Constants the models use

  • Working hours per FTE-year: 1,800. This matches the OECD figure for average annual hours actually worked in the US (1,800.2 in 2025 — OECD Employment database). Be aware that it overstates hours in some other economies — the same dataset puts the UK at roughly 1,533 hours and Germany at roughly 1,332 — so a US-derived default will inflate labour-time savings there. Regional defaults are on the roadmap.
  • Fully-loaded labour cost: salary × ~1.4, as a floor. Per the BLS Employer Costs for Employee Compensation series, wages are about 70% of total compensation in US private industry (BLS ECEC news release), which implies a ~1.4× multiplier on wages before recruiting, training, equipment, and office overhead. Multiplying salary by 1.4 therefore floors — it does not overstate — the true cost of an hour of labour.
  • Annualisation: monthly × 12, undiscounted. Single-year headline figures are simple annualisations by design — no discounting, no ramp. Discounting happens in the multi-year panel, which computes NPV at a discount rate you set, IRR, discounted payback, and an adoption ramp — and every one of those assumptions is user-editable.

Benchmark bibliography

Where a calculator offers benchmark guidance, the figure comes from one of the verified external sources below — each entry links to the publisher. One honesty note: where no credible public source exists for a commonly quoted figure, we ship no benchmark rather than an invented one. Several famous numbers in circulation (uncited cost-per-ticket averages, unverifiable deflection rates) fail that test and do not appear in the product.

Support economics

  • Live service channels average $8.01 per contact; self-service about $0.10. — Gartner, 2019. Press release
  • US inbound calls average $7.20 — 47% more than email, 23% more than web chat. — ContactBabel, US Contact Center Decision-Makers' Guide, 2026. Report
  • UK inbound calls average £5.58 — 57% more than email, 83% more than web chat. — ContactBabel, UK Contact Centre Decision-Makers' Guide, 2024. Press release
  • Only 14% of customer service issues are fully resolved in self-service; even for very simple issues, only 36%. — Gartner consumer survey (n=5,728), 2024. Press release
  • 81% of customers attempt self-service before contacting a live representative (attempts, not successes). — Harvard Business Review, Dixon et al., 2017. Article
  • Klarna's AI assistant handled two-thirds of customer service chats in its first month (company-reported best case; Klarna later rehired human agents). — Klarna, 2024. Press release
  • Intercom Fin's average AI resolution rate grew from 30% to 76%; its money-back guarantee threshold is 65% (vendor-defined resolution — we model on 30–65%). — Intercom, 2026. Guarantee terms
  • 18% of US web chats are handled entirely without human agents (up from 6% in 2020). — ContactBabel, US Contact Center Decision-Makers' Guide, 2026. Report
  • Industry-wide average handle time is about 6 minutes 3 seconds. — Call Centre Helper, 2026 (190,000+ calculator entries). Article
  • Practitioner consensus puts live-chat concurrency at 2–3 simultaneous chats per agent. — Call Centre Helper practitioner poll, 2013. Poll

Retention & churn

  • Average subscription churn is 3.27% per month (B2B ~3.8%, direct-to-consumer ~6.5%). — Recurly Research (billing-platform data). Benchmarks
  • Annual B2B churn varies widely by industry — from ~11% (energy/utilities) to ~56% (wholesale); annual figures must never be mixed with monthly ones. — CustomerGauge, State of B2B Account Experience. Report summary
  • T-Mobile postpaid phone churn was 0.89% per month in Q3 2025 — a realistic floor for telecom's stickiest segment. — T-Mobile earnings, 2025. Earnings release
  • US premium streaming (SVOD) churn averaged 4.6% per month in 2025. — Antenna, 2025 Year in Review. Report
  • Median net revenue retention for private B2B SaaS ($25k–50k ACV) is 102%. — SaaS Capital retention survey, 2025. Survey
  • A 5% retention increase produced 25–95% profit increases — for e-commerce companies, in the dot-com era; not a universal law. — Reichheld & Schefter, E-Loyalty, Harvard Business Review, 2000. Article
  • In financial services, a 5% retention increase produces more than a 25% increase in profit. — Reichheld, Bain & Company, 2001. Paper (PDF)
  • Acquiring a new customer is 5–25× more expensive than retaining one — an HBR-published rule of thumb the article itself leaves uncited. — Gallo, Harvard Business Review, 2014. Article
  • Retention elasticity is 3–7: improving retention by 1% improves customer and firm value by 3–7% (peer-reviewed; the strongest retention-value anchor we use). — Gupta, Lehmann & Stuart, Journal of Marketing Research, 2004. Paper (PDF)
  • Targeted win-back of lost customers was shown to be profitable in a study of 53,000+ lost telecom customers. — Kumar, Bhagwat & Zhang, Journal of Marketing, 2015. HBR reprint

Digital & web performance

  • A 0.1s mobile speed improvement lifted retail conversions +8.4% and travel conversions +10.1% (37 brands, 30M sessions; observational). — Deloitte / Google, Milliseconds Make Millions, 2020. Report (PDF)
  • A 100ms delay can hurt conversions by up to 7%; 53% of mobile visitors leave pages taking over 3 seconds. — Akamai / SOASTA, State of Online Retail Performance, 2017. Announcement
  • Every 1s of page-speed improvement produced up to a 2% conversion increase at Walmart (single company, 2012). — Walmart page-speed study, 2012. Original deck
  • Core Web Vitals case studies: Vodafone +8% sales from a 31% LCP improvement; Rakuten 24 +33.1% conversion (company-reported, illustrative range). — Google web.dev. Case studies
  • Global average e-commerce conversion is 2.74%, ranging from 0.71% (luxury) to 5.37% (beauty); mobile now converts at least as well as desktop in this panel. — Dynamic Yield (Mastercard), rolling benchmarks. Benchmarks
  • Median landing-page conversion is 6.6% across industries (SaaS 3.8%) — landing pages, not site-wide e-commerce conversion. — Unbounce Conversion Benchmark Report (41k pages, 464M visitors). Report
  • Google Ads search benchmarks 2025: CTR 6.66%, CPC $5.26, conversion rate 7.52%, cost per lead $70.11 (16,446 US campaigns). — WordStream / LocaliQ, 2025. Benchmarks
  • Facebook Ads lead campaigns 2025: conversion rate 7.72%, cost per lead $27.66 (US campaigns, SMB skew). — WordStream / LocaliQ, 2025. Benchmarks
  • Median blended e-commerce ROAS was 2.04 in 2024. — Triple Whale. Analysis
  • Email marketing returned £42.24 per £1 spent (self-reported by UK marketers; the famous “$42 per $1” is a currency-mangled version of this figure). — UK DMA, Marketer Email Tracker, 2019. Report
  • Average documented cart abandonment is 70.22% (meta-average of 50 studies, 2006–2025); extra costs are the top stated reason (39%). — Baymard Institute. Research

Workforce & automation

  • Average annual hours actually worked, 2025: US 1,800.2 · UK 1,532.9 · Germany 1,332.2 · France 1,498.3 · OECD average 1,736.1. — OECD Employment database. Dataset
  • The US federal government computes hourly rates on a 2,087-hour year — gross paid hours, before leave, holidays, and sickness. — US Office of Personnel Management. Fact sheet
  • In US private industry, wages are about 70% of total compensation — implying a ~1.43× multiplier on wages, and that is a floor for fully-loaded cost. — BLS, Employer Costs for Employee Compensation. News release
  • Training produces medium-to-large effects on organizational results (Cohen's d ≈ 0.62 across 26 studies) — evidence that training works, not a percentage ROI. — Arthur et al., Journal of Applied Psychology, 2003. Meta-analysis (PDF)
  • Firms in the top half of training spend returned 37% to shareholders vs 20% for the bottom half (correlational, dot-com era; directional only). — ASTD, Profiting from Learning, 2000, via Computerworld. Coverage
  • RPA implementers measured a 16% average overall cost reduction; one bank cut an error rate from 30% to almost nothing. — Deloitte UK, The Robots Are Here, 2017. Report (PDF)
  • Automation implementers reported a 32% average cost reduction, while pilot payback lengthened from 16 to 22 months (self-reported; optimistic bound, not a default). — Deloitte, Intelligent Automation Survey, 2022. Survey
  • About half of paid work activities are technically automatable; data processing is 69% and data collection 64% automatable — technical potential, not realized savings. — McKinsey Global Institute, A Future That Works, 2017. Report
  • Processing an invoice costs $9.40 on average all-in; best-in-class teams pay $2.78 vs $12.88 for the rest. — Ardent Partners, AP Metrics That Matter, 2025. Report (PDF)
  • Accounts-payable cost per invoice: top quartile ≤$2.07, median $5.83, bottom quartile ≥$10 (n=1,485; different cost scope from Ardent — never mixed). — APQC, via CFO.com, 2018. Column
  • Manual knowledge-work error rates run 1–5% (average cell error rate 3.9% across 14 studies); 94% of inspected real-world spreadsheets contained errors. — Panko, EuSpRIG, 2015. Paper (PDF)

What the outputs are not

  • The outputs are estimates produced by models. They are not financial advice, and you should not treat them as a guarantee of results — see our terms of service.
  • Results are only as good as the inputs. A model fed rough guesses produces a rough guess with more decimal places.
  • Defaults and benchmarks are starting points. Wherever you have measured data from your own operation — call volumes, handle times, churn rates, loaded costs — replace our defaults with it. Your numbers beat our benchmarks every time.