Glossary

This page defines the 78 terms that these docs and the product use in more than one place. The terms are in alphabetical order, and each entry links to the page that covers the term. For the terms of the API, see also the API types.

A B C D E F G H I L M N O P R S T V W

A

Actual to date

The figures of a plan from the start of its planning period to today. The spend is the reported spend, and the outcome is the outcome that the models attribute to the campaigns of the plan.

See also: Expected to date, Modeled, Reported, Check-in.
Learn more: Saved items and goal pacing

Admin

The role that can change a workspace. An Admin manages the members, the connections, the display currency, the API keys, and the billing.

See also: Viewer, Workspace.
Learn more: Roles and members

Allocation type

A setting on one row of the media of a plan. Optimize lets the optimizer move the row freely, Max spend sets a ceiling, Min spend sets a floor, and Fixed takes the row out of the optimization.

See also: Spend flexibility, Optimized.
Learn more: How Media Forecaster builds a plan

API key

A secret that authenticates requests to the Prescient GraphQL API. A key gives access to the data of one workspace, and an Admin creates and revokes keys on the API keys tab of the Settings page.

See also: Channel scope, Workspace.
Learn more: Create and revoke an API key

B

Backtest

A view of the Model Health tab. It shows a separately trained model that did not train on the most recent period, so the view ends where the test period of that run ends.

See also: Current Fit, Non-trained Accuracy, Model Accuracy.
Learn more: Accuracy and fit

Base

The part of a modeled outcome that the spend of a row drives directly. The product calls this the first-order effect.

See also: Halo, Modeled.
Learn more: Modeled and reported figures

Baseline

The reference-period spend mix of a plan, scaled to the budget of the plan. It is the starting point before the optimizer moves budget, and the models forecast it over the same planning period as the optimized plan.

See also: Optimized, Reference period, Lift.
Learn more: How Media Forecaster builds a plan

C

Campaign

One advertising campaign of an advertising channel. A campaign is the most detailed level in Attribution and in an Optimization plan.

See also: Channel, Tactic.
Learn more: Compare channels, tactics, and campaigns

Channel

One thing that a workspace measures, for example an ad platform, a store, a retailer, or a web analytics account. Each channel has a type, and the type decides what Prescient uses the data for.

Channel Reported

A figure that each ad platform reports for itself, before any modeling. Each platform takes full credit, so the figures of the platforms overlap and their total is typically more than the modeled total.

See also: Modeled, Reported.
Learn more: Modeled and reported figures

Channel scope

The channels that an API key can read. A key with no channel selected has access to every channel.

See also: API key.
Learn more: Create and revoke an API key

Chat

One live conversation with Prescient AI. A chat changes each time a message is sent, and it stays in the history of the person who started it.

See also: Saved conversation, Prescient AI.
Learn more: Prescient AI

Check-in

The Monday comparison of a tracked goal. It adds up the expected figure and the actual figure from the start of the plan to that day and names the pace.

Confidence band

The shaded range on a saturation curve. It shows where the actual outcome is likely to be at each spend level, and a narrower band means a more certain estimate.

See also: Saturation curve, Model confidence.
Learn more: Explore a spend scenario

Config

One configuration of a model. A model can have more than one config, and each config has its own results.

Connection method

The way that the data of one channel reaches Prescient. Prescient selects the method when the channel is added, and the setup dialog shows the form for that method.

Connection status

The state of the connection of a channel, shown in the Status column of the Channels tab. The status is separate from the date of the most recent data.

See also: Most Recent Data, Freshness warning.
Learn more: Connection statuses

Current Fit

A view of the Model Health tab. It runs through the latest modeled date and includes data that the current model has already seen.

See also: Backtest, MMM Fit, In-sample.
Learn more: Accuracy and fit

Cursor

The value in the endCursor field of a paged API response. A request sends the cursor as the after argument to get the next page.

See also: API key.
Learn more: Pagination

D

Data patch

A Google Sheet of extra rows that Prescient lays over the data of a connected channel. The patch adds rows that the connection cannot give, or corrects rows that are wrong at the source.

See also: Channel, Source.
Learn more: Data patches

Data QA

The check of the numbers of a channel by a person. The Data QA column of the Channels tab shows the date of the approval and the approver.

See also: Onboarding, Most Recent Data.
Learn more: Data dates and freshness

Default config

The config with the name Default. A model card on the Model Center overview shows this config, and the Avg Accuracy tile uses it.

See also: Config, Model.
Learn more: Configs and the Preference Score

Display currency

The currency that labels each money figure of a workspace, for example spend, revenue, and cost-per metrics. It does not convert stored amounts.

See also: Workspace.
Learn more: Change the display currency

E

Efficiency

A figure that compares an outcome with spend. A ratio such as ROAS is the outcome for each $1 of spend, where a higher value is better, and a cost such as CAC is the spend for each unit of outcome, where a lower value is better.

See also: Target, Marginal efficiency.
Learn more: Modeled and reported figures

Expected to date

The values of a plan for each day or week that has passed, added up through today. The figure follows the shape of the plan, not an even daily split.

See also: Actual to date, Pace.
Learn more: How Media Forecaster builds a plan

F

Forecast plan

A plan type of Media Forecaster that works on channels and tactics. Media Forecaster suggests it for a planning period of 42 days or more.

See also: Optimization plan, Plan.
Learn more: How Media Forecaster builds a plan

Freshness warning

A badge next to Most Recent Data on the Channels tab. It shows when a source that a person updates by hand has no new data for longer than its allowance.

See also: Most Recent Data, Source, Connection status.
Learn more: Data dates and freshness

G

Granularity

The time unit of the data of a model, Daily or Weekly. The granularity sets the periods that a page offers, and it is not how often the model updates.

See also: Model, Modeled through.
Learn more: Accuracy and fit

H

Halo

The part of a modeled outcome that the spend of a row drives indirectly, by lifting other channels and sources. The product calls this the second-order effect.

See also: Base, Modeled.
Learn more: Modeled and reported figures

Holdout

The label of the marker on the Backtest chart for the non-trained period.

See also: Non-trained period, Backtest.
Learn more: Accuracy and fit

I

Inactive campaign

A campaign that spent nothing in the last days of the reference period of a plan, after it ran earlier in the period.

See also: Reference period, Campaign.
Learn more: Media and run settings

In-sample

Data that a model has already seen. The model learned from this data, so a close match is expected.

See also: Out-of-sample, Current Fit.
Learn more: Accuracy and fit

Integration

A warehouse or a file location that carries the data of many channels on one credential. The Integrations tab of Data Center lists BigQuery, Snowflake, SFTP, and Azure Blob Storage.

L

Library

The product area that holds saved conversations. Each item is a fixed copy of a chat, with a title, a visibility, and one or more versions.

See also: Saved conversation, Version, Tracked goal.
Learn more: Library

Lift

The difference between the forecast of the optimized plan and the forecast of the baseline over the same planning period.

See also: Baseline, Optimized.
Learn more: How Media Forecaster builds a plan

M

Marginal efficiency

The modeled return from the next unit of spend. It is the modeled return from the next $1 for a return metric such as ROAS, and the modeled cost of one more result for a cost metric such as CAC.

See also: Efficiency, Saturation curve.
Learn more: How Media Forecaster builds a plan

MMM

Marketing mix model. In the product, a figure with MMM in its name is a modeled figure, for example an MMM revenue figure.

See also: Model, Modeled.
Learn more: Modeled and reported figures

MMM Fit

A score from 0 to 100 in the Current Fit view. It shows how closely the predictions of the current model follow the reported totals in the selected period, and a higher score is better.

See also: Model Accuracy, Current Fit, Non-trained Accuracy.
Learn more: Accuracy and fit

Model

A marketing mix model of a workspace. A model measures one target for one channel.

See also: Config, Target, Granularity, MMM.
Learn more: Model Center

Model Accuracy

A score from 0 to 100 that the product shows in two places. On a card of the Model Center overview it is the fit of the current model for the selected period, the same figure as MMM Fit, and in the Backtest view it is the fit of the backtest model on the dates that it trained on.

See also: MMM Fit, Non-trained Accuracy, Backtest.
Learn more: Accuracy and fit

Model Center

The product area that shows each model of a workspace, with its scores, its health, and its attribution.

See also: Model, Model Comparison.
Learn more: Model Center

Model Comparison

The page that puts two configs of a model side by side and gives each one a Preference Score.

See also: Config, Preference Score.
Learn more: Compare two configs

Model confidence

The trust in a saturation curve at the selected spend. The product shows it as Strong, Moderate, or Directional.

See also: Confidence band, Saturation curve.
Learn more: Explore a spend scenario

Modeled

A figure that comes from the marketing mix models of a workspace. The modeled outcome of a row is Base plus Halo.

See also: Reported, Channel Reported, Base, Halo, MMM.
Learn more: Modeled and reported figures

Modeled through

The last day that a model has results for. With 2 or more models selected, Attribution uses the earliest of these days, and its default period ends on that day.

See also: Model, Granularity.
Learn more: Find how current a model is

Most Recent Data

A column of the Channels tab. It gives the last day that has rows for the channel, and it belongs to the channel, not to the connection.

See also: Freshness warning, Connection status, Data QA.
Learn more: Data dates and freshness

N

Non-trained Accuracy

The score of the backtest model for the most recent dates that it did not train on. It is the one Model Center score for dates that a model did not learn from, and it does not change with the selected period.

See also: Non-trained period, Model Accuracy, Backtest, Out-of-sample.
Learn more: Accuracy and fit

Non-trained period

The most recent period of about 30 days, or about 4 weeks, that the backtest model did not train on.

See also: Holdout, Non-trained Accuracy.
Learn more: Accuracy and fit

O

Onboarding

The five stages that a new workspace goes through before every product area opens. Data Center shows the stages in the list Your onboarding.

See also: Data QA, Workspace.
Learn more: Your first days

Optimization plan

A plan type of Media Forecaster that works on live campaigns. Spend flexibility applies to each campaign, and the results have a Campaign tab.

See also: Forecast plan, Plan.
Learn more: How Media Forecaster builds a plan

Optimized

The budget split that Media Forecaster recommends, and what the models forecast that split to deliver over the planning period.

See also: Baseline, Lift, Spend flexibility.
Learn more: How Media Forecaster builds a plan

Out-of-sample

Data that a model did not train on. A close match on this data shows that the model can predict, not only describe.

See also: In-sample, Non-trained Accuracy.
Learn more: Accuracy and fit

P

Pace

The difference between the actual figure and the expected figure of a tracked goal, as a percentage of the expected figure. Within 5% is On track, more than 5% above is Ahead of plan, and more than 5% below is Behind plan.

See also: Check-in, Tracked goal, Expected to date.
Learn more: Saved items and goal pacing

Performance Table

The table of Attribution that compares results at the levels of the marketing hierarchy: channel, tactic, and campaign.

See also: Channel, Tactic, Campaign.
Learn more: Performance Table options

Plan

One budget recommendation of Media Forecaster for one goal and one period. The settings of a finished plan are locked, because the results depend on the settings.

See also: Optimization plan, Forecast plan, Baseline, Optimized.
Learn more: Media Forecaster

Planning period

The dates that a plan allocates budget across. The planning period starts 14 days before today or later, and it gives the forecast the seasonality of those dates.

See also: Reference period, Plan.
Learn more: How Media Forecaster builds a plan

Powered by

A column of the Channels and Integrations tabs. It names the service or the method that carries the data of the row.

Preference Score

A number from 0 to 100 that Model Comparison gives to each of the two configs. The product says that the score measures model quality across different criteria.

See also: Config, Model Comparison.
Learn more: Configs and the Preference Score

Prescient AI

The product area where a person asks questions about the data of a workspace in a chat. Its screen also shows the state of the connections and the models and the recommendation of the day.

See also: Chat, Recommendation card, Library.
Learn more: Prescient AI

R

Recommendation card

A card on the Prescient AI screen that summarizes the daily Media Forecaster plan of the workspace, which runs with default settings. The card appears only when the plan found a gain.

See also: Plan, Prescient AI.
Learn more: Where the daily recommendation comes from

Reference period

Dates in the past that set which media is eligible for a plan and the starting spend of each row. The reference period does not limit what the models know, because the models train on the full history.

Reported

A figure that comes from the connected sales sources of a workspace. It is a total for the whole workspace and is not divided by channel.

See also: Modeled, Channel Reported.
Learn more: Modeled and reported figures

S

Saturation curve

A curve that shows the modeled outcome of a campaign, a tactic, or a channel at each level of spend. The Explore Spend Scenario panel of Attribution and an answer of Prescient AI can show one.

Saved conversation

The copy of a chat that the Library holds. The copy keeps the messages, the charts, and the tables as they were on the day of the save, and it does not change when the chat continues.

See also: Chat, Library, Version, Visibility.
Learn more: Saved items and goal pacing

Shared credential

The credential of an integration. A person enters it one time, and each channel that arrives through the integration uses it.

See also: Integration.
Learn more: Set up shared credentials

The one-time code that Prescient AI emails to a person who signs in with an email address. The product calls it a one-time code.

See also: Single sign-on.
Learn more: Sign in to Prescient AI

Single sign-on

A setup in which an organization controls access to Prescient AI from its own identity provider. People sign in there, and membership and roles come from the identity provider.

See also: Sign-in code, Admin, Viewer.
Learn more: Single sign-on

Source

The place where the data of a channel comes from: a platform, a sheet, a table, or a file location.

Spend flexibility

A setting of a plan that sets how far the optimizer can move each row from its baseline, as a band such as 0.5× to 1.5×. It changes how the budget is split and does not change the total budget.

Subscription

The billing state of a workspace. It controls access to the product areas for each member.

See also: Workspace, Admin.
Learn more: Subscriptions and access

T

Tactic

A label that a person gives to a campaign in Data Center. Attribution can show one row for each tactic, and a tactic row can contain campaigns of more than one channel.

See also: Campaign, Channel.
Learn more: Label campaigns with tactics

Target

The outcome that a model measures, for example revenue. The selected target sets the efficiency metric that a page shows, for example ROAS.

See also: Model, Efficiency, Target of a plan.
Learn more: Overview reference

Target of a plan

The total modeled figure that a plan must reach over its planning period. It is not a reported number, and it is not growth on top of the current figure.

See also: Plan, Target, Modeled.
Learn more: Plan settings

Tracked goal

The outcome that the optimized plan expects over its whole planning period, which the Library reads from the plan when the tracking starts and keeps fixed.

See also: Check-in, Pace, Plan.
Learn more: Saved items and goal pacing

V

Version

A new fixed copy of a saved conversation that a person adds when the item runs again. A version does not replace the earlier one.

See also: Saved conversation, Library.
Learn more: Saved items and goal pacing

Viewer

The role that can use the product areas. A Viewer cannot change the display currency, the members, the connections, the API keys, or the billing.

See also: Admin, Workspace.
Learn more: Roles and members

Visibility

The setting of a saved conversation that says who can open it. Private is for the person who saved it and Prescient staff, and Shared with your team is for each person in the workspace.

See also: Saved conversation, Workspace.
Learn more: Share and export

W

Workspace

The unit that holds the data, the models, the plans, the members, and the subscription of one organization in Prescient AI.

See also: Admin, Viewer, Subscription, Display currency.
Learn more: Workspaces and organizations

Other words on screen

The product shows more than one word for some terms. These docs use the term in the last column.

Word on screen	Where	Term in these docs
Account admin	Some messages of Data Center	Admin
Company	Some messages of the API keys tab	Workspace
Connector	Header of the first column of the Channels tab	Channel
Direct connector	Text of the Add integration dialog	Integration
One-time code	The sign-in page and the Invite member dialog	Sign-in code
Organization	The menu Switch organization, the page No organization access, and some messages	Workspace