How to Forecast Demand for Global Ecommerce in 2026

Forecasting demand for a new country means estimating how much revenue it will produce before you commit to it — and doing it with inputs that reflect cross-border behavior rather than domestic. That distinction is where most forecasts break. Conversion, average order value (AOV), and repeat purchase rate all shift once an order crosses a border, and landed cost changes the price your forecast is built on.

Choosing a market comes down to four things: whether there's real demand, whether landed cost supports a viable price, your tariff and de minimis exposure, and how much operational complexity the market adds. Our guide to picking your first global markets covers all four. This guide goes deeper on the first two, because those are the ones you can put numbers against — a forecast turns "is there demand?" into a revenue range and "does landed cost work?" into a margin outcome. The other two remain risk screens, and a market has to clear those separately no matter how good the forecast looks.

In short: a cross-border demand forecast works stage by stage down your own funnel — sessions, add-to-cart, cart-to-checkout, checkout completion — using your international rates rather than your domestic ones, then multiplies through by a market-appropriate AOV and subtracts landed cost. Run it twice: once at the completion rate your domestic checkout achieves, which is the ceiling, and once at the rate a market with unpriced duties is likely to achieve. The distance between those two numbers is the opportunity, and the decision is which of them you intend to hit.

How do ecommerce merchants forecast demand before entering a new country?

Merchants forecast demand by working down their own funnel rather than down from total market size. Seven steps.

step 1 — establish a baseline from existing international traffic.

Pull sessions from the target country for a recent trailing window — at least 90 days, and long enough to smooth out a promotion or a quiet stretch. Organic and direct only: paid traffic reflects your media buying, not local demand. Use the same window for every market you compare, or the comparison isn't one. 

step 2 — step down the funnel using your international rates.

Apply your own add-to-cart rate and cart-to-checkout rate for international sessions, not your site-wide averages. Both are usually close to domestic, and sometimes better — international shoppers who reach a cart tend to be high-intent. Skipping these stages and applying one conversion rate to raw sessions is the most common way a forecast loses its footing. 

step 3 — choose a checkout completion rate, and be explicit about which one.

This is the decision that sets the whole forecast. Use your domestic completion rate — the share of shoppers who finish checkout after starting it — as the ceiling: what this market reaches once cost surprises are removed. A market where duties are collected at delivery will land well below it. Run both, a parity case at the ceiling and a friction case beneath it, and forecast the range rather than a single number. Which of the two belongs in your forecast is covered below. 

step 4 — apply an aov appropriate to the market. 

Basket size is driven more by product mix and local purchasing power than by anything you control, and it tends to run inversely to engagement — the regions where shoppers move through checkout most readily are not the ones with the largest baskets, so assuming a market will be strong on both is how a forecast ends up double what it should be. Middle East orders averaged $170 while North America averaged $96 across the same network in full-year 2025, per FlavorCloud's 2026 State of Cross-Border Commerce report. 

step 5 — calculate landed cost per representative order, then decide who pays it.

Take three to five representative SKUs, classify them to the correct HS (Harmonized System) code for the destination, and calculate duties, taxes, and fees on the full order value including shipping. Then choose: absorb it into your product price, pass it to the shopper at checkout, or split it. A Landed Cost Engine returns the figure in real time; how to set country-specific pricing covers the pricing decision that follows. 

step 6 — layer in retention.

Retention determines which threshold the market has to clear. If a meaningful share of buyers order again inside the forecast window, acquisition cost is recovered across multiple orders and a thin first-order margin is survivable. If they don't, the market has to pay for itself on the first order alone. That difference is large in cross-border: full-year 2025 repeat purchase rates ranged from 36% in North America to 13% in Africa by region, and from 52% for health and wellness to 15% for assorted consumer goods by category, per FlavorCloud's 2026 State of Cross-Border Commerce report. Leaving retention out doesn't make a forecast conservative — it silently applies the harder threshold to every market. 

step 7 — annualize the result, then set a threshold and schedule the checks.

Scale the window result up to a year and record it as an assumption, not a fact — it holds only if the window wasn't unusually strong or weak. Then write down what result justifies further investment and what triggers an exit, and set the dates to check each input as it becomes readable: traffic within a month, completion rate by 90 days, retention not before six. Forecasts without a kill criterion get defended rather than tested.

what a demand forecast covers and what it leaves out?

A forecast quantifies two of the four things that predict a good market:  

• Is there real demand? The forecast answers this as a revenue range, built from your traffic, your funnel rates, and a market-appropriate AOV. 

• Does landed cost support a viable price? It answers this as a margin outcome for each cost-handling option — absorbed, passed through, or split. 

It does not answer the other two. Tariff and de minimis exposure is a volatility question — how likely is this market's cost base to move under you. Operational complexity is a capability question — tax registration, returns, language, local payment methods. Both are screens rather than numbers, and both can disqualify a market that forecasts well. Run the forecast to size the opportunity; run the screens to decide whether you can hold it. 

which inputs go into a cross-border demand forecast?

Six inputs go into a cross-border demand forecast, and all six come from data you already have. What varies is whether you can measure each one for this specific market or have to carry it over from your international business as a whole — and a forecast built from your own funnel is far more defensible either way than one assembled from third-party averages. 

The inputs to label as assumptions are the ones you carried over rather than measured. For a market you already sell into, that's usually none of them. For a market you've never shipped to, it's everything downstream of sessions — which is worth stating plainly in the forecast rather than leaving implied.

why isn't traffic data enough to forecast demand?

Traffic data can't carry a forecast on its own, for four reasons:

1. It measures curiosity, not willingness to pay your landed price. A shopper browsing from Brazil may have no intention of paying your price plus import taxes. Sessions capture interest; completion captures willingness.
2. It's contaminated by your own activity. Paid traffic, influencer geography, and press coverage all inflate country-level sessions without indicating organic local demand.
3. It reflects the site you have, not the site that market needs. A store quoting shipping in USD only, or shipping Delivered Duty Unpaid (DDU) with duties collected at the door, suppresses completion everywhere. Low conversion from a country may be measuring your checkout rather than their appetite.
4. It stops short of the stage that matters. Sessions tell you people arrived. Whether they added to cart, reached checkout, and paid are three different questions, and the drop-off between the last two is where cross-border forecasts usually go wrong.

One thing sessions do capture, and most brands overlook: traffic from countries you don't ship to. Blocking checkout doesn't stop people arriving — it stops them buying. Those sessions are recorded demand your store can't currently act on, and they are directly sizeable.

Sizing them is what white space analysis does. FlavorCloud's Market Intelligence takes a merchant's own funnel — sessions, add-to-cart, cart-to-checkout, and checkout completion, measured by region over a recent trailing window — and compares each market against the completion rate it could reach with landed cost friction removed. The output is trapped revenue quantified country by country, split between markets where shipping is already live and markets where it isn't yet enabled.

should you use your domestic conversion rate or a regional benchmark?

Use both, for different jobs. Conflating them is what produces forecasts that are either wildly optimistic or quietly defeatist.

Your domestic completion rate is a ceiling. It represents what your checkout achieves when nothing surprises the shopper — no unexpected duty at the door, no unfamiliar currency, no delivery uncertainty. Applied to an international market, it answers a specific question: what would this market do if the cross-border friction were removed? The gap between that number and actual performance is the trapped revenue, and it is the number worth acting on for a market you already serve.

A regional benchmark is an expectation. It reflects how shoppers in a region actually behave across many merchants — including all the friction. For a market you've never sold into, it's the more honest starting point, because you have no local performance to reason from.

Practically: use the ceiling to size what a fix is worth, and the expectation to forecast a cold entry. Then state which one your number came from. A forecast that says "$85K at parity" and a forecast that says "$48K at current friction" are both defensible. A forecast that says "$85K" without saying which is not.

what are the average order value and repeat purchase rate benchmarks by region?

The regional figures below are a cross-check on your own numbers, and the fallback for a market where you have no order history to measure.  All figures are from FlavorCloud's 2026 State of Cross-Border Commerce report, full-year 2025, drawn from 600+ merchants shipping to 220+ destination countries.

Source: FlavorCloud 2026 State of Cross-Border Commerce report, full-year 2025. Repeat purchase rate is a within-period measure: the share of cross-border buyers who placed two or more orders in the same period. For directional context, global checkout abandonment sits at roughly 70%, per Baymard Institute.

what does a cross-border demand forecast look like with real numbers?

Every figure below is invented to demonstrate the arithmetic. It is not FlavorCloud data and not drawn from any merchant. A US health and wellness brand sizing Germany, using a trailing 90-day window:

Stepping down the funnel

1. Sessions: 4,500 organic and direct sessions from Germany over the window.
2. Add to cart at 7.5%, the brand's own international rate → 338 carts.
3. Cart to checkout at 88%, also international → 297 shoppers reach checkout.
4. Checkout completion — run twice. At the brand's domestic rate of 62% (the ceiling): 184 orders. At 32%, reflecting duties collected at delivery: 95 orders.
5. AOV of $115, the brand's international average nudged up for a supplements mix, at an assumed 60% gross margin — $69 per order before duty. Annualized figures below multiply the 90-day window by four.

What that produces

Three things this makes visible that a single number would hide.

Absorbing the duty costs $30 per order and buys 89 additional orders per window. Whether that trade pays depends on your gross margin: above roughly $62 per order on a $115 basket, the absorbed path produces more annual contribution, and below it pass-through does. At the $69 assumed here, absorption wins by about $2,500 a year — a narrower victory than the near-2x revenue gap suggests, and a reminder that this comparison turns on your margin rather than on the duty. Run it with your own.

The $40,940 revenue gap between the two paths is the trapped revenue in this market. It is not a projection of growth; it is the cost of the checkout experience you'd be launching with.

This is first-order only. At a 52% repeat purchase rate for health and wellness, the absorbed-duty path compounds meaningfully within the same window while the pass-through path does not — which usually widens the gap further than the first-order numbers suggest.

how does landed cost change a demand forecast?

Landed cost changes both sides of the model at once, which is why it can't be handled as a line item after the revenue estimate is done. Absorb duties and taxes and your margin per order falls while completion holds. Pass them through and margin holds while completion falls — Baymard Institute finds 39% of shoppers abandon checkout over extra costs added at the end. Either way the forecast moves, so the landed cost figure has to exist before the completion assumption is locked.

The size of that effect grew over the last year. De minimis thresholds that let low-value parcels clear duty-free have been dismantled in the two largest corridors — the US $800 exemption ended in August 2025, and the EU €150 threshold closed on 1 July 2026, replaced by a flat €3 duty charged per HS6 line rather than per parcel, so a three-category order carries €9. The point for forecasting is narrow: any model that assumed a share of orders would clear duty-free now needs that share re-costed, and multi-category baskets need it re-costed per line.

Building the cost into the product price up front is one way to protect the completion assumption, and it is the pricing decision the forecast should be run against — absorbed, passed through, or split — rather than settled after the fact.

how does product category change a demand forecast?

Product category changes the retention input more than any other input. Repeat purchase rate by vertical, full-year 2025, from the same report:

1. Health and wellness: 52% — replenishment-driven, the highest-loyalty cross-border category
2. Beauty and cosmetics: 40% — consumable, with strong brand affinity
3. Apparel and fashion: 20% — discretionary, tied to seasonal and collection cycles
4. Assorted consumer goods: 15% — deal-cycle driven, low inherent stickiness
   

Practically: if you sell supplements, your forecast can reasonably include repeat orders inside the window, and the market can clear its threshold on lifetime value. If you sell apparel, it can't, and the market has to clear on first-order contribution margin alone. Same country, same traffic, entirely different verdict.

how much international revenue should an ecommerce brand realistically expect?

Most brands generate around 16% of revenue internationally, per Shopify. FlavorCloud's Commerce Intelligence is built to move that share toward 40–60%.

For forecasting purposes, the useful read on that gap is that 10–20% is usually an execution ceiling rather than a demand ceiling. Brands plateau there because unpriced landed cost, DDU delivery, and blocked checkouts cap completion — not because the demand runs out. So if your forecast lands at the low end, ask which of those three is binding before concluding the market is small.

It also helps to express the target as a share rather than an absolute — international revenue as a percentage of total, before and after the fix you are proposing. A share moves only when the underlying experience changes, which makes it harder to hit by accident and easier to hold a plan against.

five cross-border demand forecsasting mistakes to avoid

1. Applying one conversion rate to raw sessions. Step down the funnel stage by stage, using your international rates at each stage.
2. Not saying which completion rate you used. A parity forecast and a friction forecast are different documents. Label them.
3. Treating a region as a market. Regional figures are reference points. Customs regimes, VAT rates, and per-parcel fees are set nationally, and neighboring countries can differ sharply.
4. Forecasting revenue without forecasting landed cost. A revenue number with no duty assumption isn't a forecast.
5. Building the case on first-order revenue. Repeat purchase rate ranges from 13% to 36% by region and 15% to 52% by category. It belongs in the model, not in a later optimization phase.

how do you know whether to trust your forecast?

A forecast is only useful if it names its own weakest input. In cross-border that's almost always the completion rate, because it depends on a shopper experience you may not have delivered in that market yet, or the landed cost, because it's country-specific and recently changed.

Both are addressable with data rather than judgment. White space analysis closes the first by measuring your funnel market by market and quantifying the distance to parity, including in markets where shipping isn't enabled yet. Accurate HS classification and guaranteed Delivered Duty Paid (DDP) shipping close the second, by making the landed cost in the forecast the same number the customer pays at checkout.

Once the forecast clears your threshold, the remaining two criteria — tariff exposure and operational complexity — decide whether you can hold the market. The full dataset behind this guide, including segment and category breakdowns, is in the 2026 State of Cross-Border Commerce report.