Skip to content
Do I Launch

Methodology

How a result is built

Market data first, then fixed calculations, then a written interpretation. The same inputs and the same data always give the same score.

Scoring model v0.2, experimental: thresholds are provisional

From data to verdict

  1. 1. Data

    Search volumes, cost per click and competitors are collected for your market, alongside the figures you enter.

  2. 2. Calculations

    A rules engine turns that data into five scores, a global score, a risk level, a verdict and the financial scenarios.

  3. 3. Interpretation

    An AI model writes the summary, SWOT and plan from the calculated facts. It never computes a score or a figure.

Each figure in a report keeps its source: provider data, a figure you entered, a default assumption or a calculation. A default is always labelled as such and can be replaced.

Sources

Data providers the application calls. They supply data; none of them reviews or endorses the results.

DataForSEO

Monthly Google search volumes, cost per click and bid ranges for the keywords, with 12 months of history. Also the search results pages used to spot online competitors.

Covers searches made on Google only. People who never search, or search elsewhere, are not counted.

Google Places

Businesses listed in the area, with their ratings and number of reviews (local businesses).

A business without a listing is invisible here.

GeoNames

Population of the area, used to compare search volume with the size of the market. A local copy, not a live call.

Administrative boundaries rarely match a real catchment area.

European Central Bank rates (via Frankfurter)

Exchange rates, when the keyword costs come in another currency than your project.

The rate of the analysis day is kept with the report.

OpenAI

Suggests candidate keywords and classifies their intent, then writes the text of the report from the facts supplied by the engine.

Its text can only quote the calculated facts; volumes and amounts never come from it.

Five scores, one global score

Each score is calculated on a 0–100 scale by the engine and shown here out of 10, with one decimal: 50/100 reads 5.0/10. The global score is a weighted average of the five, with weights that depend on the type of business.

Demand

How much people search for what you offer, compared with the population of the area, adjusted for trend, seasonality and concentration on a few keywords.

Commercial intent

The share of those searches that come from buyers: prices, quotes, purchases, local searches, comparisons.

Competition

How accessible the market is, from paid competition, the number of established competitors and the cost of clicks against your price. Higher means more room.

Profitability

Whether the margin on a sale pays for the cost of winning a customer through paid search, with penalties for a break-even above capacity or a loss all year.

Feasibility

Whether your available budget covers the launch budget, and whether your ad budget buys enough clicks.

Weight of each score in the global score
Type of businessDemandCommercial intentCompetitionProfitabilityFeasibility
Local service20%15%25%25%15%
Physical store20%10%25%30%15%
E-commerce30%20%15%25%10%
SaaS30%15%20%25%10%
B2B service15%15%25%30%15%

Scroll sideways to see every column.

If a score cannot be calculated, the others are reweighted as long as at least 4 of the five are available, and the global score is capped at 6.5/10. With fewer, no global score is given. A missing score is shown as “Not enough data”, never as zero.

Verdict rules

The rules are read from top to bottom. The first one that applies gives the verdict.

  1. 1
    Insufficient data

    No global score could be calculated.

    Too much data is missing to reach a conclusion. This is not a negative verdict.

  2. 2
    High risk

    Risk level is high, or global score below 4.0/10, or at least one critical risk.

    Serious risks were detected. Address them before going further.

  3. 3
    Rework the model

    Profitability below 4.0/10, or feasibility below 4.0/10.

    The market may exist, but the economic model does not hold as entered (profitability or feasibility). Rework prices, costs or budget.

  4. 4
    Strong potential

    Global score of 7.0/10 or more and data confidence of 40/100 or more.

    The measured market and your economics support the idea. Confirm it with real customers before investing heavily.

  5. 5
    Validate further

    Otherwise, in every other case.

    No blocking problem was detected, but the signals are not strong enough to conclude. Test the idea further before committing.

Safeguards

  • Demand below 3.0/10 (local service, physical store, e-commerce, or SaaS): no “Strong potential”.
  • Demand below 2.0/10 (B2B service): no “Strong potential”, unless you entered all your own economics.
  • One of the five scores could not be calculated: no “Strong potential”.

When a safeguard applies, the next rule that matches gives the verdict.

Confidence and risk, measured apart

Data confidence says how complete the data behind a result is, out of 100. It says nothing about the quality of the idea: a weak idea can have high confidence.

Data confidence, out of 100

  • 50Data coverageShare of relevant keywords that returned real search data.
  • 20Keyword depthNumber of keywords kept: full points from 15.
  • 15Competitor coverageNumber of competitors found: full points from 3.
  • 15Your figuresShare of the 7 financial figures you entered yourself.

Below 40/100: the result is marked indicative and the global score is capped at 6.5/10.

Below 20/100: no global score, only the scores that could be calculated.

Risk level

Each risk factor that applies adds a point: strong seasonality, falling trend, tough competition, expensive clicks, a budget below the launch budget, low margin, a break-even above capacity, low confidence, among others.

Low
0–1 points
Moderate
2–3 points
High
4 or more points

Some risks are critical on their own, such as a break-even above the number of sales you can handle: one is enough for a “High risk” verdict.

The financial model

The projection covers 12 months and models one channel: paid search. It estimates the clicks your budget buys at the keywords’ cost, then contacts and sales with your conversion rates.

Referrals, word of mouth, map listings, organic traffic, walk-in customers and partnerships are not modelled. For a business that relies on them, the outlook will look harsher than reality.

  • Break-even: fixed monthly costs and ad spend divided by the margin per sale.
  • Launch budget: 6 months of fixed costs, 3 months of paid search, plus the initial investment.

Prudent

Clicks at the high top-of-page bid, conversion rates × 0.6

Realistic

Clicks at the average cost per click, conversion rates × 1.0

Optimistic

Clicks at the low top-of-page bid, conversion rates × 1.4

Defaults used when a figure is not entered
Type of businessVisit to contactContact to saleGross margin
Local service6%35%55%
Physical store4%40%40%
E-commerce2%100%45%
SaaS3%15%80%
B2B service2%20%60%

Scroll sideways to see every column.

Starting points to replace with your own measurements, not industry statistics. Every default used is listed in the report.

What the analysis cannot tell you

  • Google searches are not the whole market.

    Search volume shows interest expressed on Google, not every potential customer.

  • A score is not a probability of success.

    It summarises the data available on the day of the analysis to help you decide what to check next.

  • Thresholds are provisional.

    The scoring model v0.2 is experimental and calibrated on test cases. It will change, and each report keeps the version it was calculated with.

  • Your figures drive the outlook.

    Price, costs and budget come from you. A rough estimate gives a rough result.

  • Text is written by AI.

    It interprets the calculated facts and can be wrong in tone or emphasis. The figures stay those of the engine.

  • Some data can be missing.

    When a provider returns nothing, the report says “Not enough data” and lowers the confidence instead of guessing.

On this site scores are shown out of 10. The application currently shows the same scores out of 100: 5.0/10 here is 50/100 there.

See the method applied to a full report

A recorded test case, every section included.