> ## 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 monitoring policies work > Understand model drift, ex ante tracking error, cash bands, factor exposure, and the monitoring results they produce. Monitoring policies help you identify client accounts and households that need review. **Model drift** checks allocation alignment, **sleeve tracking error** estimates benchmark-relative return variability, **cash bands** check cash levels, and **factor exposure** checks asset allocations or risk exposures against bounds. Each check measures a different part of the portfolio. A policy sets the scope, conditions, and limits for a check. Tilt applies those limits to a dated observation and records the result. A warning or breach points you toward a review, raising cash, or deploying cash. You make the subsequent proposal and trade decisions. | Check | Question it answers | Measurement | | - | - | - | | Model allocation drift | Have the amounts allocated to the models moved away from their targets? | Absolute weight differences, in percentage points. | | Ex ante sleeve tracking error | How much return variability relative to the benchmark is estimated for the current sleeve? | Annualized standard deviation of active returns, shown as a percentage. | | Cash band | Is the account's cash inside its permitted range? | Cash in USD or as a percentage of account value. | | Factor exposure | Are asset allocations or risk exposures inside their permitted ranges? | Absolute exposure and exposure relative to a benchmark, in the selected factor's units. | A **model portfolio** defines target allocations. A **sleeve** is a component of that portfolio, such as its US equity allocation. A **benchmark** defines the reference constituent weights for a sleeve. A constituent's **active weight** is its current weight within the sleeve minus its benchmark weight. ## Model allocation drift measures changes in allocation Tilt attributes current eligible holdings to the top-level model allocations, then compares each allocation's current share with its target share. If a ticker belongs to more than one sleeve, Tilt uses its benchmark contributions to distribute the holding across sleeves. The current weights use eligible invested holdings and any excess cash as their denominator. Tilt excludes manually held assets, the legacy non-tradable balance, and the account's cash reserve from the model drift basis. The cash reserve is the cash an account keeps on purpose: Tilt takes the larger of the account's cash buffer and the model's cash target in US dollars and in Canadian dollars, and adds the two. Cash above the reserve is **excess cash**. It stays in the current denominator with a target of zero, like an off-model holding. Tilt normalizes target allocations over their investable share. Check the reported drift basis and excluded values when comparing drift with total account value. Eligible holdings that Tilt cannot attribute to a model appear as **off-model** exposure. They remain in the current denominator and have a target of zero, so unmatched holdings contribute to total drift. When an account has excess cash, it also gets its own breach check. Tilt compares the excess cash weight with the individual allocation limit and suggests deploying cash when it is strictly above that limit. This check differs from a cash band. A cash band compares an account's total cash with a permitted range; the excess cash check compares only the cash above the reserve with the allocation limit. The cash buffer and the cash band are separate settings, so either check can breach without the other. ### Total drift avoids counting the same shift twice Consider this illustrative portfolio: | Allocation | Target weight | Current weight | Absolute difference | | - | - | - | - | | US equity | 60% | 67% | 7 percentage points | | Fixed income | 40% | 33% | 7 percentage points | | Off-model | 0% | 0% | 0 percentage points | Total drift is half the sum of the absolute model allocation differences, the off-model weight, and the excess cash weight. In this example, half of 7 plus 7 is **7% total drift**. Dividing by two avoids counting the same shift into one allocation and out of another twice. The individual allocation check measures the absolute difference for each model. With an illustrative total breach limit of 6% and an individual allocation limit of 5 percentage points, this portfolio breaches both checks. If the same targets instead have current weights of 54% US equity, 36% fixed income, and 10% off-model, total drift is half of 6 plus 4 plus 10, or **10%**. Tilt reports the off-model holdings in the drift detail; it does not create a separate off-model breach check. Excess cash counts the same way. Consider a \$500,000 account with a \$10,000 cash reserve that holds \$40,000 in cash and \$460,000 in holdings split 60% US equity and 40% fixed income. Its drift basis is \$490,000: the holdings plus \$30,000 of excess cash. US equity is 56.33% of that basis, 3.67 percentage points below target. Fixed income is 37.55%, 2.45 percentage points below target. Excess cash is 6.12%, against a target of zero. Total drift is half of 3.67 plus 2.45 plus 6.12, or **6.12%**. With the illustrative limits above, total drift breaches its 6% limit, and the excess cash check breaches its 5 percentage point limit and suggests deploying cash. ### Allocation differences use percentage points A target of 60% and a current weight of 67% differ by **7 percentage points**. Tilt uses that absolute weight difference. A relative calculation of 7 divided by 60 measures a different quantity. ## Ex ante tracking error measures estimated return variability Tilt evaluates tracking error for positive-weight sleeves against their resolved benchmarks, whether or not direct indexing is enabled. It compares current attributed constituent weights, normalized within each sleeve, with the sleeve's benchmark weights. The monitoring observation uses the current holdings. Tilt combines the active weights with estimated variances and covariances. Covariance describes how constituent returns move together. The square root of the resulting active-return variance gives tracking error, expressed on an annualized basis. The factor-model estimator combines common factor exposures, factor covariance, and constituent-specific risk. A policy can also select a historical covariance estimator, which uses aligned daily USD total returns to estimate those relationships. If factor-model coverage is unavailable or incomplete, Tilt tries historical covariance for that sleeve when its benchmark and attribution allow evaluation. The same policy limits apply after that estimator switch. Different estimators can produce different values for the same holdings, so inspect the reported estimator, fallback reason, and coverage when comparing results across dates. Historical covariance still evaluates the current fixed active weights. It is an **ex ante** estimate informed by history. **Ex post** tracking error measures the variability of the portfolio's actual realized return differences over a historical period. ### The resolved benchmark defines the comparison Tracking error helps you assess how closely a sleeve follows its intended reference portfolio. Check that its resolved benchmark and policy limits match the sleeve's mandate, including intentional customization. An intentionally active sleeve can have higher tracking error while remaining aligned with its investment objective. Tilt treats ETFs in a configured alternative set as equivalent for this check. It combines their current weights and benchmark weights into a representative ETF before calculating tracking error. Differences between those approved alternatives therefore do not contribute to the measured tracking error. Read the result as a comparison against the resolved implementation benchmark. ### The same weight differences can produce different risk Consider a sleeve with these illustrative weights: | Constituent | Benchmark weight | Current sleeve weight | Active weight | | - | - | - | - | | Ticker A | 50% | 60% | +10 percentage points | | Ticker B | 50% | 40% | −10 percentage points | Assume each ticker has annualized volatility of 20% and their return correlation is 0.80. These active weights produce tracking error of approximately **1.26%**. With illustrative warning and breach limits of 1% and 2%, the sleeve triggers a warning and passes the breach check. With the same weights and volatilities but zero correlation, tracking error rises to approximately **2.83%** and breaches the 2% limit. The risk relationships determine how weight differences translate into return variability. ### Interpret tracking error at the sleeve level Tracking error of 2% describes estimated annualized variability of returns relative to the benchmark. It gives you a measure of dispersion, rather than a forecast of underperformance or a maximum return gap. Future realized tracking error can differ from the estimate. The estimate can change when the risk model or return history changes, even if holdings stay constant. A warning or breach prompts you to review the cause. Compare holdings, benchmark changes, and estimator changes before deciding whether a rebalance is appropriate. A higher estimate can reflect changed market risk for the same holdings. The metric uses weights normalized within the sleeve. It does not measure that sleeve's contribution to total account tracking error. Calculating an account total would also require the risk relationships between sleeves. ## Drift and tracking error can signal different issues A portfolio can keep its target allocation between models while holding different constituents inside a sleeve. Its allocation drift can be low while tracking error is high. Conversely, model allocations can move away from target while each sleeve continues to replicate its benchmark closely. Use drift to examine allocation alignment. Use tracking error to examine the return variability associated with constituent differences inside a sleeve. The example limits on this page illustrate the calculations; inspect the policies that apply to your account for its actual limits. ## Cash bands measure cash against a permitted range A cash-band policy sets a lower and upper cash limit. It can express those limits as USD amounts or as fractions of account value. For a percentage band, Tilt divides the observed cash balance by account value. The dollar check uses the cash balance valued in USD. For an illustrative account worth \$100,000 with a 2% to 5% cash band: | Cash balance | Cash as a percentage of account value | Result | | - | - | - | | \$1,000 | 1% | Lower bound breaches; the suggested action is to raise cash. | | \$2,000 | 2% | At the lower bound; both checks pass. | | \$4,000 | 4% | Inside the band; both checks pass. | | \$5,000 | 5% | At the upper bound; both checks pass. | | \$6,000 | 6% | Upper bound breaches; the suggested action is to deploy cash. | A dollar band keeps the same dollar limits as account value changes. A percentage band changes the corresponding dollar range with account value. Cash bands produce breach checks for each boundary, without a separate warning level. A missing or invalid cash observation makes the check unevaluated; a percentage check also needs a positive account value. ## Factor exposure measures absolute and benchmark-relative bounds A factor-exposure policy checks each selected exposure against an absolute range and an active range. **Absolute exposure** is the measured exposure itself. **Active exposure** is absolute exposure minus the selected benchmark's exposure. Each range has a minimum and maximum, and Tilt records a result for each boundary. ### Asset-class types can define exposure bands **You can use an asset-class type as the factor in a factor-exposure band.** Select a type such as equity, fixed income, or cash to monitor the combined exposure to that type across the holdings. For example, an equity band of 50% to 70% checks the portfolio's total equity exposure; an active equity band of −3 to +3 percentage points checks its difference from the benchmark's equity exposure. Asset-class bands are available for account, model portfolio, and organization policies. The policy scope determines whether Tilt measures the account or the household, as described below. The worked equity example on this page illustrates both absolute and active asset-class bands. ### Scope determines the portfolio being measured | Policy scope | Evaluation level | Supported exposure checks | | - | - | - | | Account or model portfolio | Each applicable client account. | Asset-class exposures, such as equity, fixed income, and cash, including unknown exposure. | | Organization | Household. | Asset-class exposures and factors from the selected risk model. | An account or portfolio policy measures the account's asset-class exposures from its holdings. This differs from model drift: drift measures allocation to the assigned models, while asset-class exposure measures the holdings' classifications. Two models can both contribute equity exposure. For organization policies, Tilt combines the household's account positions in proportion to their account values. An account worth USD 300,000 has three times the weight of an account worth USD 100,000. Household checks require complete member-account evaluation and position snapshots; an incomplete household cannot establish an in-range result. Risk-factor values use the selected model's loading scale. Read those values in their factor units. Asset-class exposures are fractions that can be read as percentages; a risk-factor loading does not necessarily represent a percentage allocation. ### Absolute and active limits can give different results Suppose an account has 65% equity exposure and its policy benchmark has 60%. Its active equity exposure is **+5 percentage points**. | Check | Illustrative permitted range | Measured value | Result | | - | - | - | - | | Absolute equity exposure | 50% to 70% | 65% | Both absolute checks pass. | | Active equity exposure | −3 to +3 percentage points | +5 percentage points | Active maximum breaches. | At exactly 70% absolute exposure or +3 percentage points active exposure, the corresponding maximum check passes. Values strictly outside a range breach. Factor-exposure checks have no separate warning level and suggest reviewing a rebalance when they breach. Coverage applies to each measurement. Missing factors, insufficient classification or factor coverage, and unavailable benchmarks can leave checks unevaluated. An absolute check can still be evaluable when the benchmark needed for its active check is unavailable. Unknown asset-class exposure records holdings without a classification; inspect it alongside the coverage detail. ## Applicable policies determine the limits Policies can belong to the organization, a model portfolio, or a client account. Their entries can depend on a date or condition. Tilt evaluates applicable enabled entries, so more than one policy can produce results for the same metric. Drift, tracking error, and cash bands apply across those scopes to account checks. Factor-exposure policies use the account or household evaluation levels described above. For example, an applicable organization policy with a 2% tracking error breach limit and an account policy with a 3% limit can both evaluate a sleeve. A measurement of 2.5% breaches the organization check and passes the account check. The account still has a breach to review. Disabling an account or portfolio policy leaves any applicable broader policies available for evaluation. Inspect each result's source, enabled state, conditions, and evaluation date to understand which checks apply. [Manage monitoring policies](/advisor/monitoring-policies) explains how to inspect and change a policy. ### Warnings include the limit and breaches exceed it | Check | Triggers when | At exactly the limit | | - | - | - | | Total drift warning | Measured drift is at or above the warning limit. | Warning. | | Total drift breach | Measured drift is strictly above the breach limit. | Breach check passes. | | Individual allocation breach | Absolute allocation difference is strictly above the limit. | Breach check passes. | | Excess cash breach | Excess cash weight is strictly above the individual allocation limit. | Breach check passes. | | Sleeve tracking error warning | Measured tracking error is at or above the warning limit. | Warning. | | Sleeve tracking error breach | Measured tracking error is strictly above the breach limit. | Breach check passes. | | Cash or factor minimum | Measured value is strictly below the minimum. | Minimum check passes. | | Cash or factor maximum | Measured value is strictly above the maximum. | Maximum check passes. | At exactly 6% total drift, a 6% breach check passes. A separate 3% warning check still triggers. Above the breach limit, both rows can trigger; the account is classified as breached. ## Read results together with coverage A monitoring report is a dated snapshot. Check the evaluation date, holdings freshness, and price freshness before relying on its values. The [daily briefing](/advisor/daily-briefing) brings the report's account status, coverage, and possible next steps into one review. | Account status | Meaning | | - | - | | Breached | At least one evaluated check breached. Other checks may still be unavailable. | | Drifting | A warning triggered, with no breach and no unevaluated check in the classification. | | Holding | The evaluated checks produced no warning or breach. Confirm which checks were applicable. | | Skipped | Tilt could not establish a complete result. Inspect the stated reason. | Tilt assigns the account status in order: **Breached** if any evaluated check breached; otherwise **Skipped** if any check was not evaluated; otherwise **Drifting** if any check warned; otherwise **Holding**. An account with no breach, a warning, and an unevaluated check is therefore Skipped, and the warning remains in its check results. Tilt marks a rule **not evaluated** when it cannot calculate a reliable observation. Reasons include missing benchmark weights, incomplete model attribution, unavailable return history, unsupported benchmarks, or invalid inputs. If factor-model coverage is insufficient and historical covariance also cannot produce a result, tracking error remains unevaluated. Household factor results have their own classification. A household is **breached** if any evaluated factor bound breaches, **skipped** if there is no breach but a required check is unavailable, and **holding** when its evaluated checks pass. Confirm which policies applied before interpreting a holding result. A missing value does not establish that the account is within its limits. Investigate the reason and check report coverage before drawing a conclusion. ## A monitoring signal leads to a review Inspect the account or household, date, measured value, threshold, comparison operator, and policy source. Then examine the affected allocations, sleeve constituents, cash balance, or factor exposures and their coverage details. Consider why the deviation exists, including intentional customization, restrictions, cash needs, and tax considerations. Account and portfolio asset-class policies can also contribute bounds to proposal construction when classification coverage is sufficient. Monitoring still evaluates all applicable policies, while construction uses the effective account or portfolio policy. Inspect the reported construction-constraint status to see whether those bounds were applied to a proposal. The monitoring calculation can save optimization proposals for review. Saving a policy or producing a report does not submit orders. [Review a rebalance](/advisor/review-rebalance) explains the separate proposal and trade decisions. ***
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.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.