# Test ladder

URL: /business/test-ladder · Id: `page:business/test-ladder`

> Every way to grade a Hypothesis, cheapest first, and the rule that sends each one to the lowest rung that can settle it.

The test ladder lists every way to grade a [Hypothesis](/business/nouns/hypothesis), cheapest first. Each Hypothesis goes to the lowest rung that can yield the grade its threshold needs.

## The rungs
Each rung has a typical cost, a time to a [Readout](/business/nouns/readout), and the strongest [Evidence](/business/nouns/evidence) grade it can yield.

| Rung | Cost | Latency | Strongest grade | Reaches people | What it does |
| --- | --- | --- | --- | --- | --- |
| <a id="prior-lookup"></a>Prior lookup | $0 | 1h | Opinion | No | Prior lookup reads what the Portfolio already learned about the same Product Template, Motion and state. |
| <a id="offline-replay"></a>Offline replay | $0 | 4h | Opinion | No | Offline replay scores a Variant on the kept log of past choices before anyone sees it. |
| <a id="interview"></a>Interview | $0 | 3d | Commitment | Yes | Interview asks ICP members about the last time the Problem hit, and counts the introductions and time they offer. |
| <a id="smoke-page"></a>Smoke page | $0 | 1w | Behavior | Yes | Smoke page publishes an Offer at its real Price before it delivers and counts who tries to accept. |
| <a id="fake-door"></a>Fake door | $0 | 1w | Behavior | Yes | Fake door adds a coming-soon Offer to a live Site and counts the clicks on it. |
| <a id="pre-sale"></a>Pre-sale | $0 | 2w | Commitment | Yes | Pre-sale takes a refundable deposit on a reserve page, priced against what the ICP hires today. |
| <a id="concierge"></a>Concierge | $100 | 2w | Behavior | Yes | Concierge delivers the outcome by hand to a few paying Customers, in the open. |
| <a id="wizard-of-oz"></a>Wizard of Oz | $100 | 2w | Behavior | Yes | Wizard of Oz delivers behind the product’s real interface while people do the work. |
| <a id="paid-probe"></a>Paid probe | $100 | 1w | Behavior | Yes | Paid probe spends $100 on one ad venue to measure what a click and a Customer cost. |
| <a id="single-capability"></a>Single capability | $100 | 4w | Behavior | Yes | Single capability ships the one capability the approach rests on and measures use. |
| <a id="flag-split"></a>Flag split | $0 | 2w | Behavior | Yes | Flag split randomizes Variants of the built thing behind a Flag and reads each against the control. |

## The cheapest decisive rung
A rung is decisive when its strongest grade meets the threshold. Among decisive rungs, the Hypothesis takes the one with the lowest cost per day to a verdict.

The two rungs that reach no one run first, on every Hypothesis. They grade Opinion, so they prune weak candidates and never pass a gate. A Prior lookup that finds the question settled across the [Portfolio](/business/nouns/portfolio) opens the Hypothesis at `Testing`, not `Open`.

## Where each slot starts
| Slot | Grade it needs | Cheapest decisive rung |
| --- | --- | --- |
| customer | Commitment | [Interview](/business/nouns/interview): introductions and time offered |
| problem | Commitment | Interview: the last time the [Problem](/business/nouns/problem) hit |
| approach | Behavior | Smoke page at the real [Price](/business/nouns/price) |
| switch | Commitment | Pre-sale: a [reserve-first](/business/reserve-first) deposit |
| differentiation | Behavior | Fake door, one per axis |

Before [Live](/business/lifecycle/states/live), the ladder settles most questions in days. A [Flag](/business/nouns/flag) split waits for a built thing and the traffic to read it.

## Lines fixed before the first exposure
Every test states its pass and kill lines before anyone sees it: the [Metric](/business/nouns/metric), the window and the minimum sample. A line can tighten after launch and never loosen. [Test demand](/workflows/processes/test-demand) runs the rung and records the verdict.

## Next
[Experimentation](/business/experimentation) explains why every run is a test, and what no test varies.