In spite of the fact that "doing rebalancing on a per account basis is costly and inefficient," Daryanani says that many firms are doing exactly that. When asked why, he offers two possible explanations. First, he says, "If you locate assets properly, rebalancing is tough." Daryanani estimates that it takes 18 to 20 minutes to rebalance a typical Regent Atlantic "household" manually. Second, portfolio management software may play a role in the decision making process. "Centerpiece [now Portfolio Center] allows you to create asset allocation targets for households; however the base version of Advent Axys does not. To generate the household information in Axys, I'm told you must buy an additional reporting package that costs in the neighborhood of $15,000."
Setting asset allocation targets at the household level is only a starting point, however. As Daryanani correctly points out, rebalancing is part art and part science. For an automated system to work, it not only has to be rules based, it has to have some "artificial intelligence" built into it. According to Daryanani, iRebal is the only software program on the market that combines a rules-based approach with artificial intelligence.
Daryanani created the intelligence through trial and
error. Each week, he developed multiple scenarios, had the program
generate solutions and then shared the solutions with investment
professionals at the three firms. If the program came up with solutions
that matched the expert's expectations, it was a success. If the
results differed, the rules were "tweaked" in order to generate the
appropriate solutions.
Daryanani says that getting the software to
recognize when not to rebalance was a challenge. For example, if in
order to be perfectly rebalanced you had to buy and sell three asset
classes, but you could get very close to "perfection" by just buying
and selling one asset class, advisors in "real life" would chose the
latter course. After over a year of tweaking, the developers now
believe that the program can generate "real life" results for about 99%
of the cases they handle.
The rules based system has a number of factors that
influence recommended actions. For example, holdings are categorized at
the asset-class level (large-cap stocks), the subasset-class level
(large-cap value, blend or growth), and the individual-holding level
(ABC fund and XYZ index fund). Most advisors would agree that getting
the asset-class weightings right is most important, the subasset class
somewhat less important and, unless there is some specific issue
regarding an individual holding, the individual-holding level is least
important. Hence, the program assigns more weight to asset-class
decisions than to individual-holding decisions.
One of four "flags" can be associated with each
ticker on the sell side: must sell, desired sell, can sell, don't sell.
So, if there is a holding that you don't want to sell due to a zero
cost basis or client instructions, you can code it as such. Buy-side
transactions are usually easier: Most firms have one or two "preferred"
investments for each subasset class. Other factors that influence each
recommendation include taxes, location and transaction costs.
If a rebalancing trigger is tripped, the program
will first rank every possible sale in order of desirability, suggest
the "best" sale(s), then assume you have the cash in a given location
and suggest the "preferred" buys. The software is smart enough that it
knows not to do certain things, such as purchase a muni bond in an IRA.
You can also set rules at the household level so that, for example, if
a rebalancing generates too great a capital gain, the program will
automatically scale back the rebalance to an acceptable tax threshold.
The software itself is surprisingly simple to
operate. It only takes a couple of mouse clicks to launch the program
and, depending on how many accounts are involved and the complexity of
the solutions, the results can be delivered in as little as a couple of
minutes (the subset of Regent Atlantic accounts we worked with during a
live demo comprised about 95 "households" and the results came back in
a little over two minutes).
Users have the option of running all the firm's
accounts at one time or dividing them into smaller groupings. iRebal
can be instructed to alert the user when an operation is completed, so
that they can be productive while awaiting the program's results.
When the program completes its run, the information
can be viewed in a number of different ways. The best way to start is
with the "Client Details" report, which is actually a summary of all
the households. It includes information in spreadsheet form, such as
date of last rebalance, a dollar amount of the positions that are out
of balance, and the implications of a rebalance (gains and losses,
trading costs, and an approximation of the dollar benefit likely to be
generated by the rebalancing). There is also a check box next to every
account so that a trader or manager can approve the rebalancing.
Clicking on any individual "household" brings up a
screen that contains all of the details for that household. This screen
contains parameters for the household, the individual accounts being
monitored, overall positions, a summary of each asset class along with
the recommended trades and much more. In order to get so much
information on a single screen, the developers had to make ample use of
abbreviations. Whenever you need more information, you can hold the
mouse over an item and additional information about the item appears.
When all suggested actions are reviewed and
approved, modified or rejected, the recommended trades can
automatically be uploaded to the custodian(s) for execution.
In Summary
iRebal is expensive, but it could be a bargain for
firms large enough to benefit from its power. For example, if your firm
could potentially eliminate one skilled full-time position by deploying
iRebal, it is almost certainly worth the cost. If it encourages you to
improve on your asset location and tax-loss harvesting, and your
clients achieve better results due to the program's recommendations,
that alone might also justify the cost. As word of iRebal spreads, I
suspect that there will be a waiting list for the limited number of
licenses Mr. Daryanani and his friends will be selling over the next
couple of years.
Joel P. Bruckenstein publishes Virtual Office News (www.virtualofficenews.com). Reach him at [email protected].