Home / Guides

Calculator teardown and rebuild spec

What was inspected

• Endowus: https://endowus.com/en-hk/investment-plan-calculator (inline on page)
• Bowtie: https://www.bowtie.com.hk/en/insurance/claim-calculator, which iframes the actual widget from https://app.bowtie.com.hk/claim-calculator/

Endowus teardown

Ten inputs, read straight off the DOM:

Structural findings beyond the form itself:

• Zero <label> elements on the entire page. Screen readers get nothing bound to any input.
• Roughly nine H1s, two of which are completely empty. Directly violates the one-H1 rule in seo-geo-requirements.md.
• Six H2s whose text is literally "1", "2", "3", "4", "5", "6." Meaningless to a crawler, an LLM, and a screen reader alike.
• Money entered as type="text" in two places but type="number" in another. Inconsistent keyboards on mobile.

Bowtie teardown

Four inputs. All of them are facts about yourself, not forecasts:

Output renders immediately with no submit step: a named real scenario with the actual procedures listed (PET-scan, MRI, fine needle aspiration, chemotherapy, radiotherapy), a total bill of $380,808, then three plans compared side by side at 101% / 41% / 35% coverage, and the number people actually came for stated plainly: out of pocket expense, $0. It closes on "Get Plan Recommendation" and grounds itself with "calculated based on the real claims data of Bowtie."

What Bowtie gets wrong, and what the rebuild should not copy: the widget is iframed from app.bowtie.com.hk, so the parent page carries almost no crawlable content (body text on the parent page is under 900 characters). For a GEO strategy that depends on being cited, that's a real cost. Its label coverage is also thin.

Rebuild spec

{
  "id": "endowus-calculator-rebuild",
  "principle": "Eliminate, automate, simplify. Ask only for facts the user already knows; model everything else.",
  "inputs": [
    {
      "name": "goal",
      "label": "What are you saving for?",
      "type": "choice_chips",
      "options": ["Retirement", "A home", "My child's education", "Just growing my money"],
      "default": "Retirement",
      "rationale": "Replaces investmentType. Plain language, tap not type, and it sets sensible downstream assumptions."
    },
    {
      "name": "currentAge",
      "label": "How old are you?",
      "type": "select",
      "range": [18, 80],
      "default": 35,
      "rationale": "A fact about yourself. Replaces all three date fields, since the model derives horizon from age and goal."
    },
    {
      "name": "monthlyAmount",
      "label": "How much can you put in each month?",
      "type": "slider_with_numeric_entry",
      "inputMode": "numeric",
      "min": 1000,
      "max": 50000,
      "step": 500,
      "currency": "HKD",
      "default": 5000,
      "rationale": "Slider gives an instant answer without typing; numeric entry stays available for precision."
    },
    {
      "name": "startingAmount",
      "label": "Anything to start with?",
      "type": "slider_with_numeric_entry",
      "inputMode": "numeric",
      "min": 0,
      "max": 2000000,
      "step": 10000,
      "currency": "HKD",
      "default": 0,
      "optional": true,
      "rationale": "Explicitly optional so it never blocks a first result."
    }
  ],
  "modelled_not_asked": [
    { "field": "expectedReturn", "handling": "Derived from the selected portfolio's historical range. Shown as a disclosed assumption the user can open and change, never as a required input." },
    { "field": "assumedInflation", "handling": "Defaulted to the published HK figure, disclosed in the same assumptions drawer." },
    { "field": "yearsOfCashflow", "handling": "Derived from goal plus current age." },
    { "field": "endingAmount", "handling": "Removed. Reported as an output, not collected as an input." }
  ],
  "output": {
    "renders": "immediately on load with defaults, before any interaction",
    "primary_number": {
      "label": "What you could have by 65",
      "note": "One headline figure, in HKD, no jargon. This is the Bowtie 'out of pocket expense: $0' equivalent — the single number the visitor came for."
    },
    "comparison": {
      "type": "three_portfolios_side_by_side",
      "columns": ["Flagship", "IncomeUp", "CashUp"],
      "shows": ["projected value", "expected range", "risk level"],
      "rationale": "Bowtie's strongest move: three plans compared at a glance rather than one abstract result."
    },
    "secondary": [
      "Total you'd contribute vs projected growth, split visually",
      "A plain-language sentence stating what the projection assumes"
    ],
    "assumptions_drawer": {
      "collapsed_by_default": true,
      "contains": ["expected return", "inflation", "fee assumptions"],
      "editable": true,
      "rationale": "Meets the regulatory need to disclose without making disclosure the price of entry."
    },
    "grounding_line": "Modelled on Endowus's own published portfolio performance data.",
    "cta": "See the portfolio behind this number"
  },
  "structural_requirements": {
    "labels": "Every input gets a real bound <label>. Non-negotiable, and the single biggest a11y gap in the current build.",
    "headings": "Exactly one H1. No numeric-only H2s. Section headings written as real questions so they are extractable by AI answer engines.",
    "crawlability": "Rendered inline in the page DOM, server-side where possible. Do not iframe it from an app subdomain the way Bowtie does, or the content becomes invisible to crawlers and LLMs.",
    "mobile": "inputMode='numeric' on every currency field so the numeric keypad appears.",
    "schema": "Wrap the surrounding explainer copy in FAQPage schema so the calculator page can be cited in AI answers, not just used."
  }
}

Sixty word paragraph

Endowus's calculator asks ten inputs, including an expected return figure most retail investors can't supply. Bowtie's asks four, all facts about yourself, and answers instantly. NN/g's cognitive-load research is blunt about why that matters: every field a user must interpret raises abandonment risk. The rebuild eliminates the guesswork fields, defaults the rest, and shows the number people actually came for.