#!/usr/bin/env python3 """Customer persistence, complete costs and cash capture — 29 September 2026. Python 3.10+, standard library for all calculations/tests. No network calls. Usage: python capture-calculations.py [--data capture-data.csv] [--figure capture-bridge.svg] The delivery-prefixed sibling filenames work without changes. Matplotlib is required only for --figure. Reported values remain dated inputs. All money inputs are USD millions. Accrual profit is NOT collected cash. Company retention is NOT AI-product retention. Diagnostic expense-envelope sensitivities are not forecasts. The advertiser example holds a reported outcome gain fixed under a hypothetical budget change; it is not an observed campaign cash result. """ from __future__ import annotations import argparse import csv from dataclasses import dataclass from datetime import date import math from pathlib import Path from typing import Mapping DATA_NAME = 'capture-data.csv' DELIVERED_DATA_NAME = Path(__file__).name.replace('capture-calculations.py', DATA_NAME) FIELDS = ('id','entity','metric','period_start','period_end','value','unit','status', 'source_id','source_url','locator','notes') STATUSES = {'reported_actual','company_estimate','company_claim', 'provider_reported_experiment','research_assumption','reported_commitment'} UNITS = {'USD_million','USD_million_per_year','fraction','accounts'} Row = dict[str, str] Inputs = dict[str, Row] def need(ok: bool, message: str) -> None: if not ok: raise ValueError(message) def near(actual: float, expected: float, label: str = '') -> None: need(math.isclose(actual, expected, rel_tol=1e-10, abs_tol=1e-7), f'{label}: {actual} != {expected}') def load_inputs(path: Path) -> Inputs: out: Inputs = {} with path.open(encoding='utf-8', newline='') as stream: reader = csv.DictReader(stream) need(tuple(reader.fieldnames or ()) == FIELDS, 'Unexpected CSV schema') for row in reader: need(None not in row and all(value is not None for value in row.values()), 'Malformed CSV record') key = row['id'] need(bool(key) and key not in out, f'Missing or duplicate ID: {key}') number = float(row['value']) need(math.isfinite(number), f'Nonfinite value: {key}') need(row['unit'] in UNITS, f'Unknown unit: {key}') need(row['status'] in STATUSES, f'Unknown status: {key}') need(bool(row['entity'] and row['metric'] and row['locator']), f'Missing perimeter: {key}') for field in ('period_start','period_end'): if row[field]: date.fromisoformat(row[field]) if row['period_start'] and row['period_end']: need(row['period_start'] <= row['period_end'], f'Reversed period: {key}') if row['status'] != 'research_assumption': need(row['source_url'].startswith('https://'), f'Missing source URL: {key}') out[key] = row need(bool(out), 'Empty data file') return out def value(d: Mapping[str, Row], key: str, unit: str = 'USD_million') -> float: row = d[key] need(row['unit'] == unit, f'Wrong unit for {key}: {row["unit"]}') result = float(row['value']) need(math.isfinite(result), f'Nonfinite input: {key}') return result def f(d: Inputs, period: str, metric: str) -> float: return value(d, 'F_'+period+'_'+metric) def m(d: Inputs, period: str, metric: str) -> float: return value(d, 'M_'+period+'_'+metric) def growth(new: float, old: float) -> float: need(old > 0, 'Growth denominator must be positive') return new / old - 1 @dataclass(frozen=True) class PersistenceBridge: prior_revenue: float existing_increase: float implied_existing_revenue: float residual_new_revenue: float rounded_new_disclosure: float current_revenue: float cost_envelope: float minimum_existing_with_new_unchanged: float def persistence(d: Inputs, half_year: bool = False) -> PersistenceBridge: now, old, suffix = ('H126','H125','H1') if half_year else ('Q226','Q225','Q2') prior, current = f(d,old,'revenue'), f(d,now,'revenue') increase = value(d,'F_existing_growth_'+suffix) existing = prior + increase new = current - existing rounded_new = value(d,'F_new_growth_'+suffix) costs = f(d,now,'cost_revenue') + f(d,now,'operating_expenses') return PersistenceBridge(prior,increase,existing,new,rounded_new,current,costs,costs-new) NONCASH = ('da','contract_amort','noncash_lease','sbc_cf','discount_amort','deferred_tax','other_noncash') WORKING = ('ar_change','capitalized_contract_cost','prepaid_change','ap_change', 'accrued_change','deferred_rev_change','lease_liab_change') def fresh_cash(d: Inputs, period: str) -> dict[str, float]: ocf = f(d,period,'ocf') raw = ocf - f(d,period,'cash_ppe') - f(d,period,'cash_software') result = {'ocf':ocf, 'raw_cash_residual':raw, 'adjusted_fcf':raw+f(d,period,'adjustment_costs_paid')} if period.startswith('H1'): noncash = sum(f(d,period,k) for k in NONCASH) wc = sum(f(d,period,k) for k in WORKING) net = f(d,period,'net_income') result.update(net_income=net, noncash_adjustments=noncash, working_capital=wc, reconstructed_ocf=net+noncash+wc) return result def operating_profit_after_revenue_change(revenue: float, cost_revenue: float, operating_expenses: float, revenue_ratio: float, proportional_delivery: bool = False) -> float: need(revenue > 0 and cost_revenue >= 0 and operating_expenses >= 0, 'Invalid cost envelope') need(math.isfinite(revenue_ratio) and revenue_ratio >= 0, 'Invalid revenue ratio') delivery = cost_revenue * revenue_ratio if proportional_delivery else cost_revenue return revenue * revenue_ratio - delivery - operating_expenses def diagnostics(d: Inputs) -> dict[str, float]: rev, cor, opx, op = (f(d,'Q226',k) for k in ('revenue','cost_revenue','operating_expenses','operating_income')) gp = rev-cor need(op > 0 and gp > 0, 'Positive reference margin required') d_rev = rev-f(d,'Q225','revenue') d_cor = cor-f(d,'Q225','cost_revenue') d_opx = opx-f(d,'Q225','operating_expenses') d_op = op-f(d,'Q225','operating_income') sbc_decline = f(d,'Q225','sbc_expense')-f(d,'Q226','sbc_expense') return dict(revenue_increase=d_rev,delivery_increase=d_cor,opex_increase=d_opx, operating_income_increase=d_op,sbc_expense_decline=sbc_decline, sbc_share_of_operating_increase=sbc_decline/d_op, operating_increase_holding_sbc_change_out=d_op-sbc_decline, gross_margin=gp/rev,operating_margin=op/rev, fixed_cost_revenue_decline=op/rev, proportional_delivery_revenue_decline=op/gp, hosting_increase_share=value(d,'F_hosting_growth_Q2')/d_cor) def meta_analysis(d: Inputs) -> dict[str, float]: ad_new, ad_old = m(d,'Q226','ad_revenue'),m(d,'Q225','ad_revenue') volume,price = value(d,'M_impression_growth','fraction'),value(d,'M_price_growth','fraction') proxy=(1+volume)*(1+price)-1 actual=growth(ad_new,ad_old) result={'ad_growth':actual,'rounded_price_volume_growth':proxy, 'price_volume_rounding_gap_pp':100*(proxy-actual), 'revenue_increase':m(d,'Q226','revenue')-m(d,'Q225','revenue'), 'cost_increase':m(d,'Q226','total_cost')-m(d,'Q225','total_cost'), 'operating_income_change':m(d,'Q226','operating_income')-m(d,'Q225','operating_income'), 'h1_operating_income_change':m(d,'H126','operating_income')-m(d,'H125','operating_income'), 'op_plus_selected_charges':m(d,'Q226','operating_income')+value(d,'M_legal_charge')+value(d,'M_severance_charge')} for period in ('Q226','Q225','H126','H125'): result[period+'_cash_residual']=m(d,period,'ocf')-m(d,period,'cash_ppe')-m(d,period,'finance_lease_principal') return result def buyer_cost_per_outcome(outcome_gain: float, spend_growth: float) -> float: """Relative cost per outcome. Not an estimate of advertiser profit or ROAS.""" need(math.isfinite(outcome_gain) and outcome_gain > -1,'Invalid outcome gain') need(math.isfinite(spend_growth) and spend_growth > -1,'Invalid spend growth') return (1+spend_growth)/(1+outcome_gain) def run_tests(d: Inputs) -> None: for period in ('Q226','Q225','H126','H125'): near(f(d,period,'revenue')-f(d,period,'cost_revenue'),f(d,period,'gross_profit'),period+' gross profit') near(sum(f(d,period,k) for k in ('research_development','sales_marketing','general_admin','restructuring')), f(d,period,'operating_expenses'),period+' expenses') near(f(d,period,'gross_profit')-f(d,period,'operating_expenses'),f(d,period,'operating_income'),period+' op') near(f(d,period,'operating_income')+f(d,period,'interest_other')-f(d,period,'income_tax'),f(d,period,'net_income'),period+' net') c = fresh_cash(d,period) near(c['adjusted_fcf'], f(d,period,'adjusted_fcf'),period+' FCF') if 'reconstructed_ocf' in c: near(c['reconstructed_ocf'],c['ocf'],period+' cash bridge') near(m(d,period,'revenue')-m(d,period,'total_cost'),m(d,period,'operating_income'),period+' Meta op') near(sum(m(d,period,k) for k in ('cost_revenue','research_development','sales_marketing','general_admin')), m(d,period,'total_cost'),period+' Meta costs') for half in (False,True): p=persistence(d,half) near(p.implied_existing_revenue+p.residual_new_revenue,p.current_revenue,'persistence allocation') need(abs(p.residual_new_revenue-p.rounded_new_disclosure) < .1,'Rounded growth bridge mismatch') near(p.minimum_existing_with_new_unchanged+p.residual_new_revenue,p.cost_envelope,'required existing revenue') near(fresh_cash(d,'H126')['working_capital'],-4.059,'Component WC sign, not MD&A prose') near(fresh_cash(d,'H125')['working_capital'],-10.349,'Prior WC') near(fresh_cash(d,'H126')['raw_cash_residual'],107.120,'H1 raw cash') near(fresh_cash(d,'Q226')['raw_cash_residual'],52.011,'Q2 raw cash') near(value(d,'F_cash_start')+f(d,'H126','ocf')+value(d,'F_investing_cf')+ value(d,'F_financing_cf')+value(d,'F_fx_cash'),value(d,'F_cash_end'),'Ending cash') z=diagnostics(d) near(z['revenue_increase']-z['delivery_increase']-z['opex_increase'],z['operating_income_increase'],'increment bridge') near(z['operating_income_increase'],14.716) near(z['sbc_expense_decline'],11.432) rev,cor,opx=(f(d,'Q226',k) for k in ('revenue','cost_revenue','operating_expenses')) for proportional,key in ((False,'fixed_cost_revenue_decline'),(True,'proportional_delivery_revenue_decline')): threshold=z[key] near(operating_profit_after_revenue_change(rev,cor,opx,1-threshold,proportional),0,'zero-profit threshold') need(operating_profit_after_revenue_change(rev,cor,opx,1-threshold-.001,proportional)<0,'Adverse direction') need(operating_profit_after_revenue_change(rev,cor,opx,1-threshold+.001,proportional)>0,'Favorable direction') a=meta_analysis(d) near(a['revenue_increase']-a['cost_increase'],a['operating_income_change'],'Meta increment') near(a['Q226_cash_residual'],784,'Meta cash') near(a['Q225_cash_residual'],8549,'Meta prior cash') near(a['H126_cash_residual'],13170,'Meta H1 cash') q=value(d,'L_conversion_uplift','fraction') near(buyer_cost_per_outcome(q,q),1,'Equal sharing boundary') need(buyer_cost_per_outcome(q,0) None: print('FRESHWORKS: all-product perimeter; USD million; not AI-only cohort cash') for half in (False,True): p=persistence(d,half) print(('H1' if half else 'Q2')+f' growth bridge: existing {p.implied_existing_revenue:.3f}; new residual {p.residual_new_revenue:.3f}; total {p.current_revenue:.3f}; complete expense envelope {p.cost_envelope:.3f}') print(f' Required existing revenue with reconstructed new contribution unchanged: {p.minimum_existing_with_new_unchanged:.3f}') for period in ('Q226','Q225','H126','H125'): print(period, {key:round(val,6) for key,val in fresh_cash(d,period).items()}) for key,val in diagnostics(d).items(): print(key,f'{val:.9f}') print('META: USD million, all company/FoA where stated; not model-attributed profit') for key,val in meta_analysis(d).items(): print(key,f'{val:.9f}') q=value(d,'L_conversion_uplift','fraction') for spend in (0,value(d,'A_spend_growth','fraction'),q): print(f'Conditional outcome gain {q:.3%}, spend increase {spend:.3%}: cost/outcome change {buyer_cost_per_outcome(q,spend)-1:.6%}') print('FY2025 Freshworks unadjusted cash residual:',value(d,'F_FY25_OCF')-value(d,'F_FY25_PPE')-value(d,'F_FY25_software')) print('Seagate buyer-resource context cash residual:',value(d,'S_ocf')-value(d,'S_cash_ppe')) print(f'All accounting, status, unit, threshold and invalid-input checks passed; {len(d)} input records.') def render_figure(d: Inputs, destination: Path) -> None: """One explanatory accrual-profit waterfall; defaults preserve neutral styling.""" import matplotlib.pyplot as plt z=diagnostics(d) changes=[z['revenue_increase'],-z['delivery_increase'],-z['opex_increase'],z['operating_income_increase']] bottoms=[0,z['revenue_increase'],z['revenue_increase']-z['delivery_increase'],0] fig,ax=plt.subplots(figsize=(11.5,6.8)) ax.bar(range(4),changes,bottom=bottoms) labels=['Additional\nrevenue','Additional cost\nof revenue','Additional\noperating expenses','Increase in\noperating income'] ax.set_xticks(range(4),labels) ax.set_ylabel('USD million; Q2 2026 less Q2 2025') ax.set_title('Freshworks: what remains of the additional customer revenue?',pad=22) for i,(change,bottom) in enumerate(zip(changes,bottoms)): top=bottom+change text_y=(bottom+top)/2 if i in (1,2) else top+1.0 ax.text(i,text_y,f'{change:+.3f}',ha='center',va='center',fontsize=12) ax.set_ylim(0,38) ax.axhline(0,linewidth=.7) fig.text(.10,.10,'Whole-company accrual accounting, NOT AI-only revenue or a cash-flow waterfall.\n' 'Operating income rose from -8.656m to +6.060m. An 11.432m stock-compensation decline is already\n' 'inside these costs; do not count it again. Sources: Freshworks June 2026 Form 10-Q; report sections 3–4.',fontsize=10) fig.subplots_adjust(bottom=.29,left=.10,right=.98,top=.85) fig.savefig(destination,format='svg',metadata={'Date':None,'Creator':'Shaduf; capture-calculations.py'}) plt.close(fig) def main() -> None: folder=Path(__file__).resolve().parent default=folder/DATA_NAME if not default.exists(): default=folder/DELIVERED_DATA_NAME parser=argparse.ArgumentParser(description=__doc__) parser.add_argument('--data',type=Path,default=default) parser.add_argument('--figure',type=Path) args=parser.parse_args() d=load_inputs(args.data) run_tests(d) print_results(d) if args.figure: render_figure(d,args.figure) if __name__=='__main__': try: main() except (OSError,KeyError,TypeError,ValueError) as error: raise SystemExit(f'Calculation failed: {error}') from error