Free Shipping Threshold: Set It From Your Average Order Value

Set a free shipping threshold a step above your average order value (the example below tests about 20 and 50 per cent higher), then check the sums. It only works if the extra goods bought to reach it, or the orders you would otherwise lose, earn more than the shipping fee you stop charging.

That last clause is the part most guides skip. A threshold is a trade: you give up the fee on every order that already qualifies, and you hope to win it back through larger baskets and fewer abandoned carts. The order export you already have tells you how big the giveaway is before you change anything.

Follow me on Instagram@sagnikteaches

What the order export has to contain

Open your shop platform's order report and export the last 90 days. Fewer than 90 days misses a slow week or a promotion; much more than six months risks old prices. You need one row per order with these columns, or the closest your platform offers:

Connect on LinkedInSagnik Bhattacharya

• Order ID and date, to count orders and spot seasonal bumps.
• Items subtotal, before shipping and tax. Note whether discounts are already taken off.
• Shipping charged to the customer, including zero.
• What you paid to ship it, carrier label cost plus packaging. If the platform does not have this, add it by joining your carrier invoice or use an average per parcel.
• Cost of goods per order, or your product cost list so it can be worked out.

Remove test orders, cancelled orders and fully refunded orders. Keep partial refunds as they are, but note them. Check that the columns hold numbers, not text with currency symbols, because a model will happily average a column it has read wrongly.

Subscribe on YouTube@sagnikteaches

If you have never run a calculation on a spreadsheet with an AI assistant, ChatGPT for Excel covers how to hand over a file and check what comes back.

A gift shop's 1,200 orders: the numbers

This is an illustration with made-up figures. Say an online gift shop has 1,200 orders in the last 90 days, totalling $58,845 in goods, so the average order value (AOV) is $49.04. Its gross margin on goods is 45 per cent. It currently charges a flat $5.95 for every order, and the average shipping cost to the shop is $7.40 per parcel, so shipping already loses $1.45 per order.

Group the orders into bands by value. The shape of the table matters more than the average.

Just over 62 per cent of orders sit below $50, so a $50 threshold would sit right in the thick of the crowd and give away a lot; a $100 threshold would be out of reach for almost everyone. The AOV of $49.04 times 1.2 gives about $59, and times 1.5 about $74. That is where the two candidates come from: $60 and $75.

Price the giveaway before changing anything

First, what does the shop make today? Each order earns margin on goods (45 per cent of its value), minus the $7.40 parcel cost, plus the $5.95 fee. Across 1,200 orders that is $26,480 of margin, less $8,880 of shipping cost, plus $7,140 of fees: about $24,740 of profit after shipping over the 90 days. That is the baseline to beat.

Now apply each threshold to the same orders, with no change in behaviour.

• A threshold of $60: orders at $60 and over become free. That is 130 + 120 + 80 = 330 orders, each losing the $5.95 fee. Giveaway: $1,963.50, so profit falls to about $22,777 before any behaviour change.
• A threshold of $75: orders at $75 and over, 200 of them, become free. Giveaway: $1,190, so profit falls to about $23,550.

The raw giveaway is the figure to keep in view. The threshold has to win back at least that much through the two routes below, or it loses money however good it looks on the cart page.

The two ways a threshold earns its keep

Bigger baskets. A customer who would have spent $54 adds a $9 item to reach $60 and stops paying the fee. For the shop that is $9 × 45 per cent = $4.05 of extra margin, minus the $5.95 fee it no longer collects: a loss of $1.90. A top-up only pays if the extra spend times the margin exceeds the fee, so the extra must be more than $5.95 ÷ 0.45 = $13.22. A nudge to a customer who is $5 short of the line, who will add $6 to $9, loses money on that order. This one sum explains why many "you are $5 away" prompts do not help.

Orders that would not have happened. Extra costs are a common reason people leave a cart: Baymard Institute's checkout research puts "extra costs too high (shipping, tax, fees)" at the top of the reasons shoppers give. If a threshold brings back orders you would have lost, each one earns margin minus the shipping cost. At $66.50 average value an extra order earns 66.50 × 0.45 − 7.40 = $22.53. At $85 it earns $30.85.

Because you cannot see lost orders in the export, turn the question round: how many extra orders would break even? For the $60 threshold, add the small effect of top-ups (suppose 30 per cent of the 120 orders in the $50 to $59.99 band, so 36, add $9: extra margin $146 minus $214 of fees lost, so −$68). The total shortfall is about $2,032, so the shop needs 2,032 ÷ 22.53 = about 90 extra orders in 90 days, a 7.5 per cent increase, to break even.

For the $75 threshold, suppose 30 per cent of the 130 orders in the $60 to $74.99 band (39 orders, average gap $8.50) add $11. That earns $193 of margin and costs $232 of fees: −$39. The shortfall is about $1,229, and each extra order earns $30.85, so the shop needs about 40 extra orders, a 3.3 per cent increase.

The $75 line is a cheaper bet: it gives away less and needs a smaller conversion lift to pay. Its risk is that it is 53 per cent above AOV, so fewer customers can reach it; the $60 line is closer to the typical basket but costs more. A shop should not pick the winner by thinking. It should test them in turn.

Ask ChatGPT or Claude to do the banding for you

Do not paste your whole export into a chat and ask "what should my threshold be?" Models do arithmetic across hundreds of rows unreliably, as why AI is bad at maths explains. Ask for the method and the formulas, or use a plan that can run code on an uploaded file, and check the totals against your platform's own report.

I am uploading an order export (CSV). Columns: order_id, date, items_subtotal,
discount, shipping_charged, shipping_cost, cogs.

Please:
1. Remove orders with items_subtotal of 0 and report how many you removed.
2. Report order count, total items_subtotal, mean order value and median order value.
3. Group orders into these bands by (items_subtotal - discount): under 20, 20-34.99,
   35-49.99, 50-59.99, 60-74.99, 75-99.99, 100 and over. For each band give order count,
   average value, average shipping_cost and average cogs ratio.
4. For thresholds of 60 and 75, work out how many orders would qualify for free
   shipping and how much shipping_charged income I would give up.
5. Show the calculation steps or the code so I can check them, and flag any column
   that looks wrong (blank cells, text in number columns, negative values).
Do not recommend a threshold yet; only produce the tables.

A sample first sentence of a reply (illustrative): "I removed 14 orders with a zero subtotal. The remaining 1,200 orders total $58,845.00, with a mean of $49.04 and a median of $41.00." Compare the median with the mean. Here the median sits about $8 below the mean, a sign that a tail of larger orders is lifting the average.

When the table comes back, check three things yourself: the band counts add up to the order total; the sum of band values equals the total revenue; and one band, picked at random, matches a filter in your spreadsheet. If any of those fail, ask the model to show its code and rerun.

Mean, median and the team-order trap

An illustration with a sports club's online shop shows why the median matters. Suppose the club sells small items to members at $15 to $40 and, twice a season, takes team-kit orders of $900 or more. A handful of those lifts the mean AOV to $62 while the median stays at $27. A threshold built on $62 would sit at $75 or $80, out of reach for nearly every member buying a $27 scarf.

The fix is to set the threshold from the median or from the band where most orders cluster, and to handle bulk orders separately. Exclude orders above a sensible cap (say three times the median) from the AOV calculation, record how many you excluded, and keep a flat shipping rule for team orders.

A coworking space selling merchandise to members faces the opposite problem: tiny volumes, with perhaps 40 orders in a quarter. Bands of ten orders are noise. Use a single sentence rule such as "free shipping when the basket covers two items" and revisit once there are a few hundred orders.

Running the two tests fairly

You cannot easily split the same visitors between two thresholds on most platforms, so run them one after another. Keep everything else the same.

1. Baseline window. Record the last four weeks of conversion rate, orders per day, AOV and shipping profit per order. You already have these from the export.
2. Test one, four weeks. Set the lower-risk threshold ($75 in the example). Do not run it in a sale week, change ads, or add a discount code in the window.
3. Read the result against the break-even. If orders per day rose by more than the 3.3 per cent break-even and profit per visitor rose, the test passes. If it did not, return to baseline.
4. Test two, four weeks. Only run $60 if test one's conversion lift looks strong, because $60 needs a lift more than twice as big.
5. Look at the bands. Compare how many orders land within 10 per cent below the line. A pile-up just above it means customers are topping up; a pile just below it means they are giving up.

Measure profit per visitor, not just AOV. A higher AOV with a lower conversion rate can leave the shop worse off. Per-visitor figures capture both effects in one number: total profit after shipping divided by sessions.

A cart message that tells the customer what to do

If you add a progress message, have a model write three variants and keep the plainest. Sample copy (illustrative, for a shop with a $75 threshold):

Before:  Shipping: $5.95
After:   You are $22.50 away from free shipping. Orders of $75 and over ship free.
At goal: Your order ships free.

Notice that the message uses the real gap, not a vague "almost there". It also helps if you suggest items at around the gap size, because the customer who is $22.50 away is likelier to add a $24 product than to browse for one. Give the shop's best-selling items between $15 and $30 a "good with your basket" shelf in the cart, built from the order export rather than guessed.

Check the wording against your shipping page. If the cart says "orders over $75" and the policy says "$75 or more", a customer with exactly $75.00 will email you. A model can draft consistent text across the cart, policy and email footers; the shipping policy generator tutorial covers wording for delivery times, costs and returns.

How the shipping threshold goes wrong in practice

• Counting the discount twice. A shop sets the threshold on the pre-discount subtotal, then runs a 20 per cent code. A $75 basket becomes $60 after the code and still ships free, so the shop pays the fee and discounts the goods. Decide which number counts and say so on the page.
• Ignoring heavy products. A single heavy item above the threshold can cost more to send than the free-shipping order brings in. Check the shipping cost of your five heaviest products against the threshold and carve them out if needed.
• Averages hiding the real shape. If 40 per cent of orders land between $44 and $49, a threshold at $50 may suddenly make a large part of your shop eligible. Always look at the table, not only the mean.
• Forgetting returns. A free-shipping order that is returned costs you the outbound shipping and, in many shops, the return label. Orders that only reached the threshold by adding a cheap extra item often come back as a pair. Look at return rates in the near-threshold band.
• Changing the carrier rate card mid-test. A carrier price change of 40 cents per parcel shifts break-even by hundreds of dollars across 1,200 orders. Date your inputs.
• Treating the threshold as permanent. Costs, product mix and AOV move. Rerun the bands each quarter.

For the margin side of the sums, the dropshipping profit margin spreadsheet shows how to lay out fees and refunds so the 45 per cent margin figure above is not a guess, and the peak season ecommerce checklist covers what to adjust before a busy period, when a threshold test is a poor idea.

Items that should not count towards the line

Some lines in an order change the arithmetic and should be handled before you band anything. An event venue that also sells its candle range online is a good illustration: a $50 digital gift voucher in the basket adds $50 to the subtotal but never needs a parcel, so a customer could reach a $75 threshold with a voucher and a $25 candle and still get the candle shipped free. Decide which items count towards the threshold, and exclude vouchers, services and downloads.

The same goes for pre-orders that ship weeks later, oversized items with their own carrier surcharge and anything sold at a loss as a promotion. In the export, add a column flagging these lines, ask the model to recalculate the bands on the qualifying subtotal only, and compare with the first run. If the qualifying AOV is $4 or $5 lower than the headline AOV, that is the figure to build the threshold from.

Alternatives that sometimes beat a flat threshold

A single dollar threshold is the simplest rule, but a few variants can fit a shop's costs better.

• Flat-rate shipping with a higher threshold. Charge $4.95 under the line and make it free above. The lower fee reduces cart abandonment while keeping some of the giveaway.
• Item-count threshold. Free shipping when the basket has three items, which suits shops with many low-price items, where the basket's dollar value says little about how much goes in the parcel.
• Member or subscriber rule. Free shipping for customers on a list or loyalty tier. It rewards repeat buyers and builds an email list, though it adds complexity.
• Weight-based tiers. Where your product weights vary a lot, three weight bands follow your costs more closely than a value threshold.
• Free shipping on selected products. Use it for a headline product with a high margin and keep the rule off for the heavy line.

Each of these can be modelled on the same order export. Ask the model to compute the shop's profit after shipping under each rule using the same banded method as above, then compare it to the baseline.

Keeping the answer current

Write the result in one line at the top of your shipping sheet: threshold, date set, the AOV and median it was built from, the break-even order lift, and the review date. Put the review in your calendar for three months ahead. At that point, rerun the prompt on the latest 90 days and ask for two things: whether the near-threshold bands have shifted, and whether profit per order after shipping matches what the sums predicted.

If you want help applying this to your own export and carrier rates, that is the sort of job my AI implementation consultation covers; the contact page is the place to start.

Questions about free shipping thresholds

Should the threshold use mean or median order value?

Look at both. The mean is pulled up by a few large orders, such as bulk or team purchases, so it can make the threshold too high for most customers. The median shows the typical order. If they differ by more than about 20 per cent, build your bands from the distribution instead.

Should the threshold count before or after discounts?

Decide it once and state it on the site. Counting after discounts protects your margin, because a heavy discount cannot tip an order over the line. Counting before discounts feels generous but can hand free shipping to orders that barely cover their own costs. Check which figure your platform's shipping rule actually reads before you announce the line.

How long should I run each threshold test?

Run each for at least two to four weeks, or until you have a few hundred orders in the near-threshold bands, whichever is longer. Avoid sales and holidays inside the window, and change nothing else, so the comparison is fair. Small samples make random swings look like results.

What if my shipping cost varies a lot by parcel?

Use the average shipping cost per band, not one flat figure. Heavier or bulkier items cost more to send and often sit in the higher-value bands. If cost varies by destination zone, calculate the threshold for your largest zone first, then check the rest against it.

Is a free shipping bar worth adding to the cart?

Usually it is a cheap thing to try. A progress bar showing how far the customer is from the threshold costs little and tells them what to do. It only pays if the typical gap is large enough that the extra goods cover the shipping fee you stop charging.

Further reads

• Dropshipping Returns and Refunds: Policy Wording and AI Replies — Returns wording that sits beside your shipping offer.
• Landed Cost Calculator in Excel: Test Tariff Changes With AI — A cost model for the products behind these orders.
• AI for a Dropshipping Business: Where It Helps and Where It Fails — Where AI helps and fails in a running store.
• AI for Shipping Companies: Courier Quotes, Tracking Replies and Claims — How couriers quote, and what to ask them.
• Advance Shipping Notice: Prepare an ASN From Pick Lists With AI — Tidy outbound paperwork once volume rises.
• Abandoned Cart Recovery in WooCommerce: Plugin Setup and AI Copy — Install a free cart recovery plugin, set consent and timing, have AI draft three emails, and measure what the sequence really recovers.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Baymard Institute cart abandonment research (extra costs the most-cited reason for abandoning a checkout), read 5 November 2026; worked example uses illustrative figures only; method is general order-export arithmetic. Check your own carrier rate card and platform settings.