Advisers who run model portfolios spend a surprising amount of time on work that is neither advice nor investing: reimplementing a rule in a spreadsheet, rerunning it when the month ends, updating a fact sheet, and translating the result into the format the custodian wants. ENSEMBLE takes that layer over. Describe a model in plain language, get a documented, backtested implementation, and let it rerun nightly with alerts, fact sheets, and allocation exports when something changes. This page explains what that looks like in practice and where the platform's role ends.
What a model portfolio is on ENSEMBLE
The platform has one building block, and it nests.
A model holds assets, other models, or both. A 60/40 allocation is a model. So is a tactical rule that moves from equities to Treasury bills when the market is below its 200-day average. Each model has its own tear sheet, its own generated code, and its own history.
A model that holds other models is a single model with a single tear sheet: if you run a core allocation and a tactical sleeve, one that weights them 70/30 is the client model, and the platform computes the look-through holdings so you can see the actual asset weights at any date. There is no limit on how deep this goes. A fixed core of index funds alongside one or more rule-based sleeves, each of which may hold sleeves of its own, is still one model rebalanced on a schedule you set.
Because every model is built from a description, the methodology is written down by construction. That matters for compliance review, for client communication, and for the moment a client asks why their allocation changed last month.
Building and documenting a model
Building a model takes a sentence and about a minute. The platform writes the code, runs the backtest in a sandbox against twenty years of daily data, and returns a tear sheet with the equity curve, drawdowns, holdings over time, and the metrics defined on our metrics explained page. The generated Python is readable and downloads as a standalone script, so the implementation is not a black box to you or to anyone reviewing it, and the code stays yours whether or not you keep using the platform.
The how to backtest a portfolio walkthrough covers the process end to end, and the strategies section has worked examples of the rules advisers most often ask about, from the 60/40 portfolio and its variants to risk parity and the 200-day moving average.
On the Business plan, each model can produce a white-label PDF fact sheet with your firm's branding, the model's description and rules, current target weights, the hypothetical backtest, and the disclosures. Fact sheets regenerate from live data, so the version you send in March and the version you send in September are consistent in format and current in content.
Keeping models current
The part of model management that consumes the most time is the part that has the least judgment in it: rerunning the rule when new data arrives. Every ENSEMBLE model reruns after each US trading session, on every plan, and the platform is deterministic, so the same rule on the same data produces the same result every time.
When a rerun produces a new target allocation, three things can happen, depending on how you have configured the model:
- An email alert tells you the model made a new allocation decision, with the old and new weights. You are notified only on changes, not on every run.
- A webhook on the Business plan delivers the new target allocation to your systems as JSON, so a rebalancing tool, a CRM note, or an internal dashboard can react without anyone copying numbers.
- An allocation export produces a file in a custodian's format: Altruist's custom model upload and the Interactive Brokers TWS Rebalance Portfolio import today, with more custodians coming. The same data is available through the REST API for any other integration.
None of this places a trade. The platform's output is a target; the decision to act on it and the execution remain with you and your custodian.
Compliance and framing
ENSEMBLE is research software. It is not a registered investment adviser, does not manage assets, and does not make recommendations. Every tear sheet is labeled a hypothetical backtest, every fact sheet carries the disclosure that past performance does not guarantee future results, and the platform never describes a model as something a client should hold.
That framing is deliberate and it is also convenient for you, because it means the platform's output slots into your own compliance process as research rather than as advice. The methodology is documented on the methodology page, including data sources, cost assumptions, and how re-runs are kept deterministic, so a reviewer can see exactly how each figure was produced.
Agents and automation
Some advisory firms are starting to use AI assistants for research, and ENSEMBLE was built with that in mind. The same models you build in the browser can be built, read, and revised by Claude, Cursor, or a custom agent through the hosted MCP server or the REST API. A common pattern is to let an assistant draft and backtest variations of a model during research, then review and adopt one through the normal process. The AI agents section covers the setup.
Getting started
The Business plan is the one built for advisory use: it includes $200 of monthly usage that rolls over, Max effort builds, webhooks, white-label fact sheets, broker allocation files, and fee modeling. Pricing is on the pricing page, and every account starts with $100 in credit and no card required, which is enough to build and document several models before deciding.
If you would like to talk through a specific workflow, such as documenting an existing set of models or wiring alerts into your systems, email support@ensemble.markets and we will set up a call.