The high-yield savings setup I'd build today, not the one I built in 2020

My 2020 HYSA stack was a Marcus account holding everything, $200 weekly auto-transfer. It worked. It was also inefficient in three ways I didn't see until I rebuilt it. Here's the setup I'd build today, the math behind each piece, and the part nobody warns you about — what happens when rates drop.

Who actually needs more than one account

Anyone with more than two financial goals running in parallel. If all your money sits in one bucket labelled 'savings' you will, over time, raid the emergency fund to pay for car tyres and then feel bad about it. The mental accounting is real. A second bucket labelled 'car maintenance' that you fund $60/month means the tyres aren't an emergency, they're a planned expense.

Who doesn't need this: anyone in the first eighteen months of building any savings at all. Open one HYSA, put money in it. Multi-bucket discipline is a tool for people who've already mastered the habit. If you don't have a balance yet, optimisation is procrastination.

Other group that doesn't need this: anyone whose only savings goal is retirement. A 401(k) plus a Roth IRA is two accounts already, and adding sinking funds for furniture is over-engineering.

The four-account stack

Account one: operating checking. Keep 1 to 1.5 months of expenses. Everyday spend and bills. The goal is never having a balance below your monthly minimum trigger. A no-fee online checking account is enough. The interest rate on this account doesn't matter — it's your buffer, not your savings.

Account two: primary HYSA for the emergency fund. 3-6 months of expenses. As of 2026, online savings runs 4.0-4.5%, down from a peak of 5.0+ in 2024. Online banks beat brick-and-mortar by a wide margin — your local bank's 0.01% is a tax in disguise. Ally, Marcus, Wealthfront, Synchrony, SoFi all do roughly the same job. Pick the one with the cleanest app, because you'll use it twice a month for the rest of your life.

Account three: second HYSA for sinking funds. This is the bucket I missed in 2020. Sinking funds are pre-planned expenses you know are coming — car repairs ($50/mo), vet bills ($30/mo), annual insurance ($80/mo), Christmas ($50/mo), birthdays ($20/mo). YNAB or Monarch lets you label sub-buckets inside a single savings account, so you don't need ten accounts. One account with software labels works fine. If apps aren't your thing, a paper cash envelope wallet ($18) does the same job.

Account four: brokerage or money market for medium-term. Cash you'll need in 2-5 years — down payment, sabbatical, planned car replacement — shouldn't sit in a 4.5% HYSA when a Treasury money market fund yields similar and a short-term bond ladder yields better. Open this at the same brokerage where your Roth IRA lives so transfers are instant. Fidelity, Schwab, Vanguard all do this well.

What the interest math actually looks like

People dramatically overestimate the importance of rate shopping at this scale. The difference between a 4.0% and a 4.5% HYSA on a $10,000 emergency fund is $50 a year. Worth taking when the apps are equivalent, not worth switching banks every six months for. The bank that lets you set up sub-buckets, automatic transfers, and external account linking in under five minutes will save you more time than the rate difference saves you money.

Where rate matters: large balances. A $50,000 down-payment fund at 4.5% earns $2,250 a year, at 3.5% earns $1,750 — a $500 swing. At that scale, chase the rate. Below $15,000 in any one account, optimise for app quality and integration. The basic principles of getting control of your finances apply regardless of yield.

The frequently overlooked variable: taxes. HYSA interest is taxed as ordinary income at your marginal rate. On a 4.5% nominal yield in a 24% bracket, your after-tax yield is 3.42%. A Treasury money market fund is state-tax-exempt, which for a Californian or New Yorker raises the relative attractiveness substantially. TurboTax Deluxe models this for you in February if you ask it to.

When rates start dropping

The Fed cycle since 2022 has trained a generation of new savers to expect 4-5% on cash. That won't last. When rates drop to 2-3% — which most curve forecasts assume by 2027-2028 — the math shifts. Cash returns less, equity exposure becomes relatively more attractive, the sinking-fund-versus-credit-card calculation tightens.

What not to do: chase yield into less-safe products as HYSA rates fall. The 2019 pattern was investors moving cash into "high-yield" funds that turned out to be junk bond exposure when COVID hit. FDIC insurance on a HYSA is the floor you don't trade away for an extra 0.5%. Read The Psychology of Money if you haven't — Housel makes the case for keeping the cash safe better than I can in three paragraphs.

What to do: lengthen duration on cash you don't need in the next 12 months. A 12-month Treasury bill ladder ETF or laddered CDs lock in current rates before the drop. The reinvestment risk is real — when your CD matures in 2027, the renewal rate will be lower. Build the ladder so something matures every 3 months and you can adjust as you go.

Mistakes I made building the 2020 version

Holding too much cash. My emergency fund in late 2021 was $45,000, which was eighteen months of expenses, which was way too much. The opportunity cost of that cash sitting at 0.5% during the 2021 equity run was substantial. Right-size to 4-6 months, not 12-18, unless your job security is unusually low.

Not automating the transfer. The original setup required me to manually move money each month. I forgot, then I forgot again, then I missed three months. Auto-transfer the day after payday lands.

Treating the HYSA as a brokerage. I checked the balance three times a week. There is nothing to do about a savings account balance besides add to it. Set it up, automate the transfer, look once a quarter.

The setup that actually works is boring. One checking, one HYSA for emergencies, one HYSA for sinking funds with software labels, one brokerage for the 2-5 year money. The yield matters less than the structure. The structure matters less than the habit.