Platform objects
Dataset
Dataset
A project-level container that holds the data sources for one body of work. All operations — data, models, predictions — happen inside a dataset.
Data source
Data source
A connection to where your data lives. Allyy supports files (CSV, JSON), warehouses and databases (Google BigQuery, Microsoft SQL Server), storage and transfer (Google Cloud Storage, SFTP), and marketing tools (Agillic, ActiveCampaign, Salesforce). Sources can be batch (synced on demand or on a schedule) or streaming (real time).
Model
Model
A predictor trained on your history to answer one question. Each model has a type (e.g. propensity, expected amount, churn, lifetime value) and is trained on a recipe of features derived from your data.
Prediction (score)
Prediction (score)
The output of a model applied to a population — a probability (0–1, for classification models like propensity or churn) or an amount (for regression models like expected gift or lifetime value).
Decision
Decision
One or more model scores combined and optimised into a ranked, budget-aware action list — for example “mail these 12,000 supporters, in this priority order, with this ask amount.”
Export
Export
A decision delivered to your tools — synced to a connected destination or retrieved via the API.
Workflow
Workflow
Automation that chains steps — sync data, train, score, export — on a schedule, so the pipeline runs without manual clicks.
The Allyy data model
When you map your data, you map it onto these entities. See The Allyy data model for detail.| Entity | What it represents |
|---|---|
| Contacts | The people (or organisations) you engage — your supporters. |
| Offers / Content / Treatments | The things a contact can interact with — a mailing, a call, an appeal. |
| Responses | An interaction between a contact and an offer — a gift, a click, a refusal. |
| Subscriptions | A standing agreement — a recurring gift / direct debit, with a start and (optionally) end date. |
| Lists | Collections of contacts used as the population for a model or a campaign. |
Modelling terms
Feature
Feature
A signal the model learns from, derived from your data — e.g. “days since last gift”, “number of gifts in the last 90 days”, “average gift size”. Allyy builds these for you from recipes; you don’t compute them by hand.
Classification vs regression
Classification vs regression
Classification predicts a probability (will they respond? will they churn?). Regression predicts a number (how much will they give? what is their lifetime value?).
Score group
Score group
Supporters ranked by score and split into bands (e.g. top 10%, next 10%, …). Dashboards often show response rate and income per score group so you can see how well the ranking separates good prospects from poor ones.
Optimisation
Optimisation
Choosing how far down the ranked list to go. Contacting fewer, higher-scoring supporters usually lowers cost and raises ROI for a small drop in total income — the profit curve shows the sweet spot.
Donor tiers
Allyy describes supporters with a consistent value vocabulary:| Tier | Meaning |
|---|---|
| Major | Annual giving at or above your major-donor threshold. |
| Middle | The upper band of single-gift donors, below major. |
| Standard | Single-gift donors below the upper band. |
| Regular giver | On a recurring direct debit / standing order — a giving cadence, not a value tier. |