> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tilt.now/llms.txt
> Use this file to discover all available pages before exploring further.
# How optimization makes tradeoffs
> Understand how targets, risk, taxes, costs, and constraints shape a proposed portfolio.
Optimization turns an account's saved holdings, effective model allocation, and settings into a proposed portfolio and trades. A proposal can differ from the model because reaching the target also involves taxes, transaction costs, and account restrictions.
This page explains how to interpret those tradeoffs. [Review a rebalance](/advisor/review-rebalance) covers running and reviewing a proposal.
## Targets, preferences, and limits
Three kinds of input play different roles:
| Input | Role | Example |
| - | - | - |
| Target allocation | Describes the portfolio the account is trying to implement. | Sleeve weights and the holdings used to implement each sleeve. |
| Preference | Changes which feasible portfolio the optimizer favors. | More emphasis on tax benefits or a larger penalty for trading. |
| Constraint | Limits the portfolios or trades considered in a solve. | A position bound, turnover limit, cash requirement, or buy restriction. |
A preference is a tradeoff, rather than a promise about the final metric. Increasing tax emphasis can change the proposal, but does not by itself guarantee a specified tax bill. Similarly, a cost penalty does not mean estimated transaction costs must remain below that value.
Effective settings can come from the organization, model portfolio, account, and applicable Dynamic Rules. Read [settings and rules](/advisor/configuration-and-rules) to understand which values apply to the account.
## What closeness to the target means
Tilt supports different ways of penalizing departures from the ideal portfolio. In the mean-variance configuration, the optimizer uses modeled factor and asset-specific risk to assess weight differences. In the weight-distance configuration, it penalizes squared differences in portfolio weights. The chosen configuration affects which departures are expensive to the optimizer.
Two proposals with the same total weight difference can therefore receive different risk penalties. Correlated holdings may partly offset each other in a risk model, while a concentrated exposure can contribute more risk.
The optimizer's distance from its ideal weights and a report's tracking error against a benchmark are related measures, but their reference portfolios and definitions need to be checked. [Models and benchmarks](/advisor/models-and-benchmarks) explains the references; [monitoring explained](/advisor/monitoring-explained) explains ex ante tracking error and drift.
## Why exact replication may be undesirable
Consider a hypothetical account with an overweight position that has a large unrealized gain. The following are possible choices, not outputs from an actual optimization:
| Choice | Potential advantage | Tradeoff to inspect |
| - | - | - |
| Sell the entire excess | Move closer to the target immediately. | Realize more gains and incur trading costs. |
| Sell part of the excess | Reduce the deviation with fewer sales. | Retain some drift and modeled active risk. |
| Keep the position for now | Defer the gain from a sale. | Continue carrying the overweight exposure. |
Restrictions and limits determine which choices are feasible. Preferences then influence which feasible choice the optimizer favors. A loss-harvesting opportunity elsewhere can also change the balance. [Tax-aware management](/advisor/tax-aware-management) explains the lot and tax assumptions behind that calculation.
## Turnover and costs measure different things
A turnover limit controls the amount traded. Tilt's optimizer measures two-way turnover as the sum of absolute traded portfolio weights, excluding the cash residual. Selling 5% of account value and buying another 5% uses 10% two-way turnover in this measure.
Transaction-cost penalties influence the optimizer's choices. Reported estimated transaction costs are calculated separately, so a cost penalty and the reported cost estimate are not interchangeable. Neither an estimate nor a saved proposal establishes the eventual execution price.
## What the proposal establishes
A successful proposal gives you a candidate portfolio under the run's inputs and configuration. Check the resulting holdings, proposed trades, drift, modeled risk, tax estimates, costs, and exceptions together. An infeasible run means the attempted combination could not produce an acceptable solution; it does not establish that the model allocation itself is inappropriate.
A monitoring breach identifies an account for attention. It does not establish that an immediate rebalance is the best tradeoff, and running optimization does not execute trades. Use [reports and backtests](/advisor/reports-and-backtests) to inspect evidence before accepting a proposal.
***
Wash-sale rules may disallow a loss if substantially identical securities are bought within 30 days before or after the sale, including in the client’s other accounts, such as IRAs or a spouse’s accounts. The Platform’s wash-sale checks cover only accounts on the Platform, so confirm the client’s full position with a qualified tax professional.
Tilt provides portfolio technology, not investment, tax or legal advice, and is not a registered investment adviser or broker-dealer. The advisor makes all investment decisions, including whether to accept a proposal and submit trades, directly or through approval rules the advisor configures. Trades are executed through the client’s custodian. Estimates depend on the data and assumptions recorded for the account; confirm the client’s tax treatment with a qualified tax professional.
See [IRS Publication 550](https://www.irs.gov/publications/p550) for U.S. wash-sale rules.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.