Grocery Delivery Best Practices for Local Stores

Most grocery delivery best practices come down to five decisions: which delivery model you run, how you pick and substitute orders, how you keep cold food cold, what you charge, and how you route the day. Get those five right and delivery becomes a profitable second storefront. Get them wrong and you are paying someone to hand your customer a bag of warm milk.

This guide is for the independent store deciding how to offer delivery on its own terms, rather than handing the customer relationship to a national app. The demand is not speculative. Online grocery orders passed 19% of the category’s total sales in Q1 2026, and sales have grown more than 20% year over year for six straight quarters, according to Brick Meets Click data reported by Digital Commerce 360. Around 80% of those delivery orders are same-day.

The Bottom Line

• Pick your model before your software. In-house, third-party fleet, and marketplace apps have different costs and give you different amounts of control over the customer.
• Picking accuracy is the whole game. High-performing operations run 96-98% order accuracy, and a substitution policy is what protects the rest.
• Cold food has a legal clock. The FDA Food Code wants cold TCS food at 41°F or below, and perishables should not sit out for more than two hours.
• Price delivery against your real cost per drop. Keep delivery cost under roughly 15-20% of average order value, then use minimums to get baskets above that line.
• Fewer, tighter windows beat all-day availability. Batched windows let you build dense routes, which is the only durable way to cut cost per delivery.

Choosing a grocery delivery model: in-house, third-party fleet, or marketplace

There are three ways a local store can get groceries to a doorstep, and the choice drives everything downstream.

In-house delivery means your staff picks the order and your driver delivers it in your vehicle. You keep the margin, the customer data, and the brand experience. You also take on vehicles, insurance, payroll, and the awkward reality that delivery demand is lumpy while a driver’s shift is not.

A third-party fleet puts drivers on your deliveries while you keep your own storefront and your own checkout. Your team still picks and packs, so quality control stays in the store. You pay per delivery instead of per hour, which matters on slow days.

A marketplace app lists your store in someone else’s shopping experience. It brings reach you cannot buy on your own, and it takes a commission plus the customer’s loyalty. Commission rates for third-party ordering platforms are commonly quoted at 15-30% of the order, and climb higher once add-on fees land.

Most independent grocers land on a hybrid. They run their own van for the dense neighborhoods and standing weekly orders, hand the scattered outlying drops to a fleet partner, and treat a marketplace listing as paid discovery rather than as their delivery operation. Delivery platforms built for local businesses sit in that middle lane. Metrobi, for example, provides multi-stop route optimization and lets a store work with the same drivers over time, with a dedicated support rep on every delivery, which is the part that keeps a hybrid model from feeling like a lottery.

One warning about marketplaces. When the app’s shopper picks your shelves, your store gets blamed for the substitution you never chose. If you go that route, keep your inventory feed accurate and tight.

How to pick and substitute online grocery orders accurately

Order picking is where local grocery delivery quietly succeeds or fails. A late delivery annoys a customer. A wrong one makes them stop ordering.

High-performing fulfillment operations maintain 96-98% order accuracy, and grocery ecommerce benchmarks from Local Express put most organized grocery operations at 98% or better. Grocery is harder than general retail: identical cartons with different milk fat percentages, multipacks that differ by one unit, expiry dates that matter, and three temperature zones in a single basket.

A few practices do most of the work:

• Pick in zone order, not list order. Sequence the pick list to match your aisles, then finish with chilled and frozen so cold items spend the least time out.
• Scan, don’t eyeball. A barcode scan at the pick catches the 2% milk that should have been whole.
• Batch small orders, single-pick big ones. Two or three small baskets can be picked together with color-coded totes. A 60-item order deserves its own run.
• Write the substitution policy down. Same brand, different size beats same size, different brand for most categories, and nothing gets substituted in baby formula, allergen-sensitive items, or alcohol.
• Ask before you swap. Substitution acceptance rose from 66% to 75% when retailers used data-driven substitution rules, per the same Local Express analysis, and a text message to the customer beats a guess.

Track pick time per order too. If it takes 25 minutes to pick a $70 basket, your labor cost per order is already competing with your delivery cost, and no routing improvement will fix that.

Keeping groceries cold in transit

Temperature is the one area where getting it wrong is a health issue rather than a service issue. The FDA Food Code requires cold holding of time and temperature control for safety foods at 41°F or below, and the two-hour rule means perishable food should not sit at room temperature longer than two hours, cut to one hour when the air is above 90°F.

That gives you a clock, and a practical setup for it:

• Stage cold last. Chilled and frozen items come off the shelf immediately before the van leaves, not at the start of the pick.
• Use insulated totes with gel packs, separated by zone. Frozen, chilled, and ambient in the same bag means all three arrive wrong.
• Log a temperature at pack-out and, for long routes, at drop-off. A cheap probe and a clipboard is enough to show you did it.
• Put perishable-heavy stops early in the route. Route sequence is a food safety tool, not just an efficiency tool.
• Set a hard cutoff for unattended drops. Leaving groceries on a porch in July is how you lose a customer and a health inspection at once.

If you deliver ice cream, frozen seafood, or raw proteins, cold capacity should decide how far your delivery radius extends. Draw the boundary on the cold chain, not on the map.

Setting grocery delivery windows customers will actually choose

Grocery delivery windows are a capacity tool disguised as a convenience feature. All-day availability sounds generous and produces routes with one stop in each direction.

Offer two or three windows a day, not eight. Two-hour windows are the sweet spot for local grocery delivery: narrow enough that the customer will stay home, wide enough that you can sequence six or eight stops inside them. Cap each window at the number of orders one shift can actually pick and drive, then close it and push the next order to the following slot. A window you blow through is worse than a window you never offered.

Two more details worth building in. Set an order cutoff far enough ahead that picking is not a panic. Most stores need two to four hours between the last order and the first drop. Release windows in line with your restock too, so a customer ordering Wednesday morning is shopping a shelf that actually has product on it.

Same-day matters here. With roughly 80% of online grocery delivery orders being same-day, a store that only offers next-week slots is not really in the market. But same-day does not have to mean one hour. A morning cutoff with an afternoon window is same-day, and it is a schedule a small team can actually hit.

What to charge for grocery delivery

Grocery delivery fees fail in two directions, and the second one is harder to spot. A fee set too high suppresses orders, which at least shows up in your order count. A fee set too low keeps the orders coming and loses money on each one.

Start with your own numbers. Grocery delivery runs expensive per package because of temperature control, fragility, and time constraints, and the Local Express benchmarks cited above put the sustainable ceiling at 15-20% of average order value. Cost per drop is not a constant either. The same delivery costs meaningfully more in a Friday rush or in bad weather than on a Tuesday at noon.

So build the price from three levers:

• A minimum basket. This is the strongest lever you have. Online grocery baskets often exceed $100, and a $50 or $60 minimum turns the ratio of cost to order value in your favor without raising a single fee.
• A flat delivery fee, tiered by distance or basket size. Waive it above a higher threshold instead of discounting it everywhere.
• A membership or subscription for regulars. Prepaid unlimited delivery smooths your demand and buys loyalty from exactly the customers you want.

Compare that stack against the commission on a marketplace order before deciding which is expensive. A 25% commission on a $90 basket is $22.50, which is often more than the fully loaded cost of driving it yourself.

Routing and dispatching a day of grocery deliveries

Once picking is reliable and windows are batched, cost per delivery is a routing problem. Density is the answer: the more stops you fit into the same hour and the same few square miles, the cheaper each one gets.

Practical routing habits for grocery delivery routes:

• Group by window, then by geography. Never let one driver zigzag across two windows.
• Sequence with software, not memory. Multi-stop route optimization handles the traffic, turn restrictions, and stop ordering that a person guessing at a map will not.
• Assume a realistic service time per stop. Groceries take three to seven minutes at the door, longer with stairs, and routes planned on drive time alone always run late.
• Keep a buffer stop count. Planning a van to 100% of capacity means a single traffic jam breaks every promise behind it.
• Give the driver the cold instructions, not just the address. Which bag is frozen, whether the drop can be left, and where the side entrance is.

Track on-time delivery rate by window. If one window runs consistently late, the problem is usually the cutoff or the capacity cap, not the driver.

The metrics that tell you grocery delivery is working

Five numbers are enough to run this. Review them weekly:

• Order accuracy. Orders with zero errors, divided by total orders. Aim for 98%.
• Substitution acceptance rate. What share of your swaps the customer keeps. A rising number means your rules are getting smarter.
• On-time rate inside the promised window. The metric customers actually judge you on.
• Delivery cost as a percent of average order value. Under 15-20% is the line for sustainable economics.
• Repeat order rate for delivery customers. Delivery is a habit business. If they do not come back, something upstream is broken.

Frequently asked questions

Is grocery delivery profitable for a small store?

It can be, but only if the basket is big enough to absorb the drop cost. Stores that succeed use minimums, batch orders into a few windows, and keep routes dense. Stores that lose money usually offer free delivery on small baskets across a wide radius.

Should I use my own drivers or a delivery partner?

Use your own when volume is steady enough to fill a shift in a tight radius. Use a partner when volume is uneven or the stops are scattered, since you pay per delivery rather than per hour. Many stores run both.

How far should we deliver?

Let the cold chain set the limit. Map how long your insulated packing keeps product in safe range, subtract your pick and load time, and draw the radius inside that. For most stores that is 20 to 45 minutes of driving.

How do we handle out-of-stock items?

Publish a substitution policy, exclude the categories where a swap is never acceptable, and contact the customer while the picker is still on the floor. Refunding a missing item is always better than delivering the wrong one.

Do we need software to start?

No. Start with a shared order sheet, insulated totes, and two delivery windows. Add picking and routing tools when order volume, not ambition, makes the manual version break.

Grocery delivery rewards stores that treat it as an operation rather than a feature. Pick a model, write the substitution rules down, protect the cold chain, charge what the drop actually costs, and batch the day into routes that hold together. That is the whole playbook, and it works at five orders a day as well as at fifty.