Skip to content

Assessment

Assessment turns the facts you collect in a hearing into a result — a quote, a fit score — and tells you which unanswered question would change that result the most.

It is built for hearings that never run top to bottom. What is worth asking next depends on what you already know, so the questions are not a fixed list:

  • You get an answer from the first question on. While anything is still open the result is a range, and it narrows as you fill facts in.
  • “Unknown” is an answer. The question stops being asked, and it is listed as something to confirm.
  • The order of questions is worked out for you. The next question is the one whose answer would narrow the result the most. Questions that stop mattering because of earlier answers drop out.
  • What you know stays with you. How much each thing adds to the result, and your rates, live in the catalog. None of it leaves Minion — including through the integration API.

Assessment is experimental and off by default. Ask us to enable it for your workspace if you do not see it in the sidebar.

TermWhat it is
CatalogA reusable model for one kind of decision: its facts, contribution items, measures and, optionally, a verdict. Found under Assessment → Catalogs
FactSomething you find out. Yes / No, a number, or a choice
Contribution itemWhat goes into the result, and when it applies — “multi-language routing: 16 hours, when multi-language is needed”. This is your know-how, and it stays inside Minion
MeasureAn output of the catalog: initial cost, monthly fee, fit score. Shown as currency or as a number
VerdictOptional. A pass / borderline / fail call made from a measure, for uses such as screening
AssessmentOne case: the answers so far and the result they give

The quickest way in is Add a sample, which copies a finished catalog into your workspace — a logo design quote, a web production quote, deal triage, or hiring screening. Read it, then change it into your own.

A catalog is edited as a draft and published before it takes effect. Only the published version is used — by assessments in Minion and by integrations alike.

Starting from blank asks for an identifierweb-production, say — alongside the name. It is derived from the name when the name is written in English, and you fill it in yourself otherwise. It is worth a moment’s thought: an identifier cannot be changed after the catalog is created, because integrations pin it to check they are talking to the catalog they expect, and letting it move would turn a rename into a false alarm. You can see it later under Integration, and under each name in the catalog list. Samples bring their own.

A catalog can be driven from outside Minion: a quote form on your website, a chat bot, or your own back office. Visitors answer the questions there, and each hearing lands in your assessment list for someone to review.

Leaves MinionNever leaves Minion
The questions: label, description, type, unit, choices, topicContribution items and how much each one adds
The result as presented: a range for each measureThe conditions that decide when an item applies
The verdict’s label and levelHourly rate, overhead rate, target margin
Which facts are still openThe breakdown, and hours before they are turned into a price

One thing to plan around: whoever holds the key can read every question in the catalog, and the questions alone show what you price. The key belongs on the server of the site you build, never in a browser.

  1. Publish the catalog. A key cannot be issued for a catalog that has never been published.
  2. Open the catalog and choose the Integration view.
  3. Under Intake API keys, name the key after where it will be used — “Estimate form on the corporate site” — and press Issue key.
  4. Copy the key now. It is shown only once; Minion keeps nothing but a hash of it.

If the key comes with a warning that some numeric facts have no sample values, the key still works, but the order of questions will be less accurate. Set Sample values on those facts and publish again.

How keys behave:

  • A catalog with at least one active key can be used from outside. Revoke every key and it is closed again. There is no separate switch.
  • Issue one key per place it is used. A key only reaches the hearings it started, so your website and your chat bot never see each other’s hearings — and nothing created inside Minion can be reached through any key.
  • Revoking a key stops that integration immediately. The hearings it started stay in your list.
  • Deleting a catalog deletes its keys. The confirmation tells you how many are still active.
  • Turning Assessment off for the workspace stops every key.

A key cannot be issued while the catalog’s verdict points to an implementation that does not exist; the catalog editor flags this.

A hearing started through a key is an ordinary assessment in your list, marked External. You can narrow the list by catalog, by origin (Created here / From integration) and by status. Assessments without a single answer — mostly visitors who left at the first question — are hidden by default; Show unanswered brings them back.

StatusMeansWho acts next
In progressStill being filled inThe visitor, or whoever is filling it in
Awaiting reviewThe site has ended the hearingYou: check the answers and finalize
FinalizedA person checked it, and the result is frozen
  • You are notified when a site ends a hearing that has at least one answer. Every workspace member gets it in the app; email is off by default and can be turned on in your notification settings.
  • Finalizing is always a person’s decision. An integration cannot finalize. Once you finalize, the site can still read the hearing, but it can no longer change or delete it.
  • Reopening for editing puts a hearing the site had ended back to Awaiting review, not In progress, because it still needs a person’s check.
  • An assessment awaiting review is the same record as any other — open it, correct answers, add what the visitor did not know, then finalize.

Changing a catalog while hearings are running

Section titled “Changing a catalog while hearings are running”

Until it is finalized, an assessment follows the catalog’s published version. When you publish a change, hearings in progress pick it up on their next request: new facts can be asked, deleted facts stop being asked, and answers already given stay. A finalized assessment keeps the version it was finalized with, including when it is read through the API. Unpublished draft edits affect nothing.

If you delete a choice that a visitor has already picked, their answer is kept as it was.

Answers are yes / no, numbers and choices; there is no free-text answer. The only text an integration stores is the title it gives a hearing when it starts it.

If your site also collects names or email addresses, it keeps them itself. To match them up, the site can put something recognisable in the title — a company name, say — as long as you intend that to be stored in your workspace.

To remove a hearing, the integration can delete the ones it started, unless they are finalized. Anyone in the workspace can delete an assessment from the list.

  • Catalog text is in one language — the one you wrote it in. For a site in several languages, create a catalog per language and give each its own key.
  • The API is for servers. It sends no CORS headers, so a browser cannot call it directly, and the key must never be shipped to a browser.
  • Rate limits, per minute: 300 requests per key, 60 new hearings per key, and 20 changes to any one hearing.