What your own records can and cannot say about a market
The market entry and prioritisation guide scores candidate markets on four signal groups, demand, competition, regulation and cost to serve, and its desk evidence for demand, from local-language search volume to import statistics, comes from outside the business. A business that already sells online holds a second kind: parcels sent abroad, enquiries in other languages, distributors asking to stock the range, and search and analytics data from countries nobody planned for.
Desk researchResearch from existing sources, the business's own and published ones, done first so that interviews and surveys cover only what nobody has answered. begins with that internal evidence, and the step called "Start with what the business already holds" in the desk research tutorial lists the sources to gather. This tutorial does the country-level reading the list leaves open.
An order is a choice made with money at stake; an enquiry costs the sender nothing. The distinction between stated and revealed preference applies directly: what people in a country did weighs more than what they said they might do.
The records describe the people who happened to find the business, not the market. They say nothing about a market's size, which belongs to the market sizing guide, or about local competitors and rules, so the reading informs which markets are scored and how; it does not choose the market.
Before you start
- The scoring model. The market entry guide's criteria and weights, agreed before this reading begins, so the evidence is gathered for the scores rather than the scores bent to fit it.
- Order data. An export from the business's own order system, with delivery and billing country and the customer behind each order.
- Contacts and data access. The inboxes, forms, chat and support tool, approaches from distributors and shops, and the site's Search Console and analytics properties.
- An agreement on handling the data. Order exports and contact records hold personal data. Agree with whoever handles data protection what is extracted, who sees it and when the working copies are deleted.
- A named reader. The owner of the market scoring, who will take the result into it.
The steps
Write down what the reading is for
Before pulling any data, write one paragraph. Name the decision the reading feeds, which countries enter the market scoring and with what evidence; the period to be read, long enough for seasonal gifting and one-off campaigns to show up as what they are; and what the reading will not decide: market size, entry or a test.
Then write down what would count. A country joins the list for scoring when its evidence points to people living there who chose the product and could be served at a cost the business can carry. Any line expressed as a count comes from the business's own economics, never from a figure found elsewhere.
Pull orders by delivery country with their returns and failed deliveries
Take orders from the order system itself. The cross-market measurement guide compares markets on backend data, meaning orders, revenue and customers from the business's own systems, and treats platform figures as directional.
Group the export by delivery country and record, for each, orders, customers, revenue, products, first and repeat orders, refunds, returns and failed or refused deliveries. Keep billing country beside delivery country, and count customers as well as orders, because one buyer reordering can look like a market. Note anything in the period that could have produced orders on its own, such as a bulk order or a press mention.
Collect enquiries, support contacts and partner approaches by country and language
Gather what people abroad have sent: messages through every channel, support tickets and approaches from distributors, retailers and agents. The guide to supporting customers across languages and time zones tags each contact with its language and its market from the first exchange. Where those tags are missing, tag the period's contacts by hand, taking the country from the address given or the edition used rather than from the language.
For each contact, note what was asked. Questions about delivery cost, duties, returns or payment speak to cost to serve and barriers; questions about stock or price speak to interest. Keep partner approaches in a list of their own.
Filter Search Console by country and split brand from non-brand queries
GSCGoogle's search performance and indexing tool. holds the query evidence. The Search Console guide, drawing on Google's documentation of the performance report, describes clicks and impressions that can be narrowed by country, query, page and device, and it warns that site totals mix brand and non-brand searches with every country and device. Filter to one country at a time, then split its queries by excluding the brand terms.
Brand queries come from people who already know the name and say little about whether strangers there want the category. The tutorial on keyword research for a new market uses the same filter for a market already chosen, observing that a market's queries, even those typed in the home language, reveal what it is already looking for, and it insists that terms are researched from the market up rather than translated. A short list of non-brand queries is therefore a pointer for later research, not a term list. The Search Console guide also notes that the long tail is anonymised, so a country's query table can be incomplete. Record clicks and impressions by query group for each country, with the period.
Read analytics location as a pointer, not a count
The guide to reports and dimensions in Google Analytics, drawing on its help centre, describes comparisons built from dimension conditions, such as users from one country, and filters that restrict a report to matching rows. Use one to see where a country's visitors land and whether they reach the delivery page.
Do not turn those visitor figures into a demand count: the cross-market measurement guide treats visitor location as a weak stand-in for the market. Read location as a prompt for where to look, and let the orders, enquiries and queries carry the evidence.
Find out who the buyers abroad are
Before a country's orders count as its demand, find out who placed them. Orders billed at home and delivered abroad with a gift message point to home customers sending presents; an account whose address moved abroad points to a customer who relocated; repeat orders in trade quantities point to a reseller. Orders billed and delivered in the same country, by customers with no home history, are the closest to local demand.
Then talk to some of the buyers. The customer interviews tutorial keeps each conversation on what the person actually did, such as what they compared and what nearly stopped them; here, also ask how they found the business. Before each call, explain to the participant what is being recorded and for what purpose, who has access to it, for how long it is kept and how they can withdraw, and record nothing until they have agreed, as that tutorial sets out. For a participant abroad, the customer research guide for new markets writes the participant information for that jurisdiction and has it checked locally.
Count what serving them cost and who gave up
For each country still on the list, add up what its orders cost to serve: delivery paid against delivery charged, returns, refunds, failed deliveries and support time. The guide to delivery promises observes that sending orders abroad brings in other services, distances and costs that vary from one market to the next. Check the Delivery PromiseA store's claim about what delivery will cost and when an order will arrive, made on product pages, at checkout, in confirmations and in ads. too: whether the dates and charges shown at checkout held for these orders.
Then look for the people the orders leave out: baskets with a foreign address abandoned at the delivery step and enquiries that asked about cost and went quiet. Survivorship bias is the error of reading the ones who got through as the whole picture, and the people who saw the delivery charge and left may describe the market better than those who paid it.
Read each count against its size and its history
Counts from a country the business does not market in can be small, and small counts move by chance. The guide to reading small numbers puts the count beside every rate, a range around every rate and a pooled view beside a short one; apply all three here. Ranking countries by rate invites the league table trap that guide describes, where the extremes belong to the smallest counts.
Read each country against its own earlier periods, and check whether a change lines up with an event noted when the orders were pulled. A count too small to judge alone still goes into the scoring, flagged as small, with its period. Unlike a planned test, whose Sample SizeThe number of users or sessions an experiment needs to detect a given effect reliably. is fixed before it starts, these records were never sized for any question, so a country with thin evidence may need research or a test first.
Enter the evidence into the market scoring with its source
Take each country that met the bar written at the start into the market scoring. The weighted scoring matrix asks for a recorded source behind every score and a label on any guess, and checks whether the ranking survives a change of weights. Own-records evidence fits two of the market entry guide's signal groups: demand, from orders by buyers who live there, non-brand queries and local customers' enquiries, and cost to serve, from delivery cost against charge, returns, failed deliveries and support load. Brand queries and gift orders from home customers do not go in as that market's demand.
Write the source beside each score: the export, the period, the count and any small-count flag. Score competition and regulation from desk research and local questions. A country with own-records evidence earns no bonus for it, and one with none is not marked down, because the records follow where the business happened to be visible.
A fresh test of whether people in a shortlisted market would commit at a given price is the job of the demand test tutorial. A country the business once paused or closed comes back through the guide on returning to a paused or closed market, which starts from the pause record rather than from fresh orders.
Decide how those customers are served until the decision is taken
The people already buying from abroad are customers now, whatever the scoring concludes. The international SEO hub advises that a market which would get only a machine translation is usually better left on the primary edition, with its shipping information made clear, and it records Google's guidance that unreviewed machine translation offered up as indexable pages may fall foul of its spam policies as auto-generated content. No local edition goes up while the scoring runs.
Make the home edition honest for those countries instead: delivery costs, times, returns and support hours stated before checkout, and contacts routed by their language and market tags. Decide with operations whether to keep accepting orders for a country whose failed deliveries or returns cost more than the business can carry. Then date and record the reading, so the next review compares each country with its own earlier figures.
Sources of evidence and what each can mislead about
| Source | What it can show | What it can mislead about | Check before scoring |
|---|---|---|---|
| Orders by delivery country | People who paid to have the product sent there | Gifts from home customers, a single reseller or a one-off spike passing as a market | Billing beside delivery country, customers beside orders, events in the period |
| Returns and failed deliveries | Part of the cost to serve | A country looking cheap to serve because few orders have reached it | Read against the order count and flagged as small |
| Enquiries and support contacts | Interest, and the barriers people raise | Questions are statements, and one language can span several countries | Country from the address or edition, and what each contact asked |
| Partner approaches | That a business there sees an opportunity | A partner's interest standing in for customers' | Kept in a separate list and not counted as demand |
| Brand queries in Search Console | People who already know the name | Awareness the business created, read as a market's need | Filtered out of demand and read as reach |
| Non-brand queries in Search Console | People looking for the category | An incomplete table and terms that only resemble the market's own | Treated as a pointer for research from the market up |
| Analytics visitors by country | Where on the site a country's visitors go | Travellers, cross-border shoppers and VPN users blurring the picture | Used to decide where to look, not as a count of demand |
| Abandoned baskets with a foreign address | People who wanted the product and stopped | Ordinary abandonment read as a delivery problem | The step where they left, against the country's other checkout steps |
A beeswax wrap maker reads its overseas orders (illustrative)
Jupplevorn Wraps reads its orders from Market A and Market C
Jupplevorn Wraps, an invented maker of beeswax food wraps, sells from its home site and markets only at home. The order export for the last 12 months shows 184 orders delivered to Market A and 71 to Market C.
Billing addresses change the picture. Of Market A's 184 orders, all were billed at home and 112 carried a gift message. Market C's 71 orders were all billed and delivered there, from 58 customers, 9 of whom reordered. Search Console filtered by country shows 1,240 clicks from Market A, 88% of them from queries containing the brand name, and 430 from Market C, 61% of them from category searches in the local language.
Short calls with 8 buyers, 4 in each market, each agreed to in advance, match the records: all 4 Market A buyers were home customers, 3 of them sending a present to relatives, and all 4 Market C buyers lived there and had found the brand through a search. Market C's checkout records show 47 baskets with a Market C address abandoned at the delivery step, more than at all the other checkout steps together, and 16 of the 23 enquiries from Market C asked about delivery cost or returns. Market C had 2 failed deliveries and 3 returns, and delivery paid against delivery charged on its 71 orders left a gap the team judged it could carry.
Read against its own history, Market C's orders went from 22 in the previous 12 months to 71; the team flags both counts as small and finds no promotion or press mention in either period. Market C goes into the scoring with its demand score sourced from its own orders and non-brand queries, the counts written beside each score, and its cost to serve scored from delivery costs, returns, failed deliveries and the support load of its 23 enquiries, with the abandoned baskets written beside it. Market A stays off the shortlist: its orders are home demand delivered abroad. Until the scoring is decided, the home edition gains delivery, returns and support information for both countries, and contacts from both are tagged by language and market.
Common mistakes
- Reading the biggest country in the export as the answer. Orders follow where the business happened to be seen. Find out who the buyers are first.
- Counting gifts as local demand. Orders billed at home and delivered abroad come from home customers.
- Treating brand queries as category demand. People typing the name already know the business. Split brand from non-brand first.
- Using visitor location as a count. Location is a poor proxy for the market. Use it to decide where to look.
- Forgetting the people who gave up. Abandoned baskets and enquiries that went quiet belong in the evidence.
- Ranking countries by rate on small counts. Show the count, a range and the country's own history.
- Building a partial edition while the scoring runs. Serve those buyers from the home edition with clear delivery, returns and support information.
The international strategy hub sets this reading beside the market entry, customer research, pricing and pausing guides around it.
