How Oraqul decides.
For anyone who has to judge the numbers. Every score is worked out by code from the taste data it returned, by the rules on this page, and every rule is written into the receipt that uses it.
How a lift is computed
lift = score(x) ÷ score(baseline)
Both scores come from the same Qloo ranking call: your audience in a place, against a neutral audience in the same place. Above 1 means your audience likes it more than most.
Measured, sourced, computed, written.
Measured
Taste and place come from Qloo: who the audience is, what they like and where they gather, ranked for each city.
Sourced
Dated market facts, such as rivals, release windows and distributors, come from web research and link to their page.
Computed
Code works out every score, verdict, ranking and confidence, and checks every line of the plan against them.
Written
The model reads your brief, chooses which questions to ask, and writes the plan. Launch craft is labelled and cites nothing.
Numbers come only from values Qloo returned in the same run. If the model names a title or place that is not in that run’s Qloo data, the output is rejected and the step runs again. A fact with no page behind it is not stated.
The taste data changes the decision, not just the wording.
Qloo is part of the planning loop. Oraqul uses its audience taste data to test the brief, rank real options and decide which recommendations can be stated. The model cannot add a taste claim after the fact or set the result itself.
Tests the brief
Audience, market and partner assumptions receive a verdict from published rules.
Changes the order
Places, dates and partners are found and ranked from the returned data before the plan is written.
Leaves a receipt
Every grounded recommendation links back to the finding and Qloo call behind it.
Verdicts
Each belief in a plan gets one of four verdicts, always shown as a glyph and a word. The exact rule used is written in that belief’s receipt.
- HoldsThe values Qloo returned support the belief by its rule.
- WeakThere is support, but below the rule’s strong threshold.
- BreaksThe values point the other way, or another option ranks clearly above it.
- Could not testQloo returned no signal. It is never counted as Weak, and nothing is guessed.
Lift
How much more your audience likes something than a neutral audience in the same place, from one ranking call. Strong at 1.5 and above, usable from 1.2 to 1.5, weak below 1.2.
Fit
A score from 0 to 100 for a partner Qloo scored against mass-market names in one call. 75 and above is recommended, 50 to 74 is usable, and below 50 is not recommended.
Rival overlap
The shared audience between your title and a rival, from one scoring call. Lower is better: below 70 is low, 70 to 85 is moderate, and above 85 is high overlap.
Direction calls
A verdict tests a belief. A direction says what to do about a decision: an opening date against its rivals, a territory as a place to open, a partner among the brands Qloo ranked. Code sets it from a value Qloo returned, with the confidence of the finding behind it.
- OkayOverlap below 70, a place at 1.2 or above, a partner in the top third.
- RiskyOverlap 70 to 85, a place from 1.0 to 1.2, a partner in the middle third.
- AvoidOverlap above 85 (high overlap), a place below 1.0, a partner in the bottom third. For a live show, a rival on the same night or week is at least risky.
A call on something you chose keeps your choice. An avoid on your own city says to open it after the stronger ones, and never to drop it. Something Qloo did not rank gets no call.
Opportunities
A nearby city is shown under Worth a look only when it ranks above your weakest city in the same Qloo call, and an area only when a city’s neighbourhood reading ranks it strongest. Each comes with why it fits, which cites its finding, and three lines of launch craft that claim nothing about taste: what the data does not show, a first test and what to measure. Code writes all four, asks Qloo nothing more, and never changes your plan: an opportunity is a lead beside it until you add it. If the evidence is thin there are none, and the page does not fill a list. Qloo shows who likes what, not that a market will sell.
Your results
Results are your own evidence, kept apart from Qloo’s findings and from web facts. You say the period they cover and where they came from, and Oraqul keeps each only when your own words contain it. Code works out each place’s share of the total and sets it against the place the plan gave it. For a campaign test, code compares your figures for the two angles: fewer than 30 of what you counted for either, or figures within 20% of each other, are inconclusive, and otherwise it says which angle recorded more and by how much. Those two thresholds are chosen conventions, not a significance test. Results show what happened. A line about them never states a cause as fact, and a taste score never stands in for a result.
What changed
When you ask, Oraqul asks Qloo and the web again with your plan’s own brief and sets the answers beside what the plan saved. A finding the plan had is moved when a number differs by 10% or a name changes place, and otherwise unchanged. A page is changed only when the plan saved a fact from the same page and it now says something different, and each fresh fact carries the day it was read. A finding or page it does not return is not a change, because the check asks the brief again and does not replay the exact calls. Your plan stays as it is until you ask for a revision.
Launch rules
The craft around the data is written down per category and versioned: how a calendar is worked back from the release date, the one signal a plan is judged on, the channels that lead, and the partner tiers. Each rule carries its source and the date it was set, and each plan records the version it used. A target is the goal you set, or a benchmark of Oraqul's own where there is one, and it is never a forecast. The rules are on the guide.
Route and topics
Route score, for tours
A weighted mean of lift over the stops, weighted 1, 0.9, 0.8 and so on by position, so the first cities count most. The best order by lift is shown beside it for comparison.
Topic ownership, for podcasts
One comparison of your audience with the rival shows’ audiences. A topic only your side lists is yours; one on both sides is shared. The weight orders and sizes the topics and never decides who owns one.
Confidence
Set for each line from what stands behind it: how many comparable titles, people and themes Oraqul could use, how well Qloo knows each, and whether they agree. A launch that is not in Qloo yet is the normal case and does not lower confidence by itself. Few or poorly known comparisons do. It is a word, never a percentage, because a percentage would claim a precision the data does not have.
Held back
Some signals stay out of the bottom line until they are validated: the values signal and the early adopter score and premium tolerance. None of them is used in a plan today, and nothing here is ever a price.
What it measures
Oraqul measures audience taste: who likes what, and where. Sales, streams, ticket numbers and prices are yours to bring, and Oraqul compares them with the plan when you do. It drafts briefs and next steps for you to send; it never sends anything itself.