App de Shopify
Analytics, logs, insights and A/B tests
The Analytics page and savings estimate, the Logs and order history, AI insights with backtests, and provider A/B tests on your own orders.
Address Validation iO shows what it checked, what happened to the addresses with a problem, and whether another provider would fit your shop better. This page covers the Analytics page, Logs, the Insights page (AI insights package) and provider A/B tests.
Analytics
Open Analytics in the app menu. Pick the period at the top: 7 days, 30 days (the default), 90 days or 12 months.
| Tile or chart | What it counts |
|---|---|
| Address checks | Every address AVIO checked, in checkout and on orders, re-checks included. Billing addresses and tests aren't counted |
| Valid addresses | The share of checks with the result Valid |
| Addresses corrected | Orders whose address changed from Needs checking or Invalid to a valid one |
| Customer follow-up | Of the orders with a problem: how many got an email, were corrected or were confirmed unchanged |
| Fulfillment holds | Orders held and released |
| By country, By provider | Checks per country, and requests per provider. Backup provider attempts count too, so one check can be two calls |
Email and phone check totals appear when those checks are on.
Analytics counts address checks, including checkout checks and re-checks. The Plan page counts billed orders, and an order is billed once however often it is checked, so the two numbers differ.
The savings estimate
AVIO counts every corrected address as one failed delivery avoided. The formula is shown under the chart:
corrected addresses × cost per failed delivery − AVIO charges = estimated savings
It is an estimate, and a careful one: customers who fix their address during checkout aren't counted. It can be negative when the charges are higher than the avoided costs.
The cost per failed delivery starts at $15. Click Change cost basis under the savings chart to open the Cost calculator and set your own.
Logs
Logs answers "what happened to this order, and why?" and "who changed what?". It shows a chart, "Orders and checks", and a table of events, newest first.
| Type | What it records |
|---|---|
| Validations | An address was checked, was incomplete, was skipped (with the reason) or got a new result |
| Address changes | Old and new value per field, and who changed it |
| Corrections | The customer corrected or confirmed the address, or it was corrected in the admin |
| Emails | Correction emails, reminders and email checks, including bounces |
| Phone checks | A phone check; the number is masked |
| Holds | A fulfillment hold placed or released |
| Tag updates | Order tags AVIO added or removed |
| Flow triggers | Whether AVIO sent a trigger to Shopify Flow |
| Billing | Charges, the spending cap and subscription changes |
| Settings changes | Which settings changed, before and after, and by whom |
| App events | For example installed, validation paused or resumed |
| Errors | A provider couldn't check an address, or a Shopify update failed |
The Source column says where a check came from, for example Checkout, Order created, Admin, Bulk or Customer.
Filter by date, type, source, order number and result, or show only problems. Export CSV downloads up to 10,000 matching events. View on a row opens the details, such as the provider calls or a before-and-after table.
Events are kept for 90 days. Names, phone numbers and full email addresses are never stored in Logs; emails and phone numbers appear masked.
Order history and the Orders list
Click an order number in Logs, Log next to an order under Logs → Orders, or View history in the admin order block to open the order's history. Now shows the current result, the customer follow-up, any hold, the address and when it was last checked. History lists every event of the order, oldest first.
Logs → Orders lists your checked orders with their result, customer follow-up, provider, country and fulfillment hold. It is also where you Re-check orders by filter; see Order checks.
Insights (AI insights package)
The Insights page shows where address problems slip through and what to change. It needs the AI insights package; see Plans. Open Analytics → Open insights and pick a period.
- Overview: orders checked, flagged, corrected after a request and escaped problems, also by country and by provider.
- Escaped problems: orders that passed validation but had a problem later: the address was changed, the delivery failed, or a later check found a problem.
- Ask AVIO: a short written summary with recommendations, often with a button such as "Set up backtest". Until you click Generate AI summary, the page shows a summary calculated from your orders.
Only totals and rates per country and provider, plus your provider setup, go to the AI model. Addresses, names, emails and order numbers never leave AVIO. Nothing in your shop changes unless you click a button.
Backtests
A backtest answers "would another provider have caught the problems that slipped through?". Pick a provider and a country, review the estimate and confirm. AVIO re-checks up to 100 escaped orders and counts each as Caught, Missed or No answer. If the provider caught enough of them, a button such as "Use Loqate for Germany" makes it that country's provider.
A backtest shows which problems a provider catches, not how often it flags good addresses. For that, use an A/B test. Backtest re-checks are billed like verified orders and count toward the spending cap.
Provider A/B tests
An A/B test compares two or three providers on your own new orders. Your normal providers still decide every order, so a test never changes a result, a tag, an email or a hold, and never slows down checkout.
Open Providers → Compare providers → Open A/B tests and fill in "New A/B test":
| Field | Rule |
|---|---|
| Name (optional) | Up to 60 characters |
| Providers | Two or three |
| Countries | Up to 20; each needs at least two of the chosen providers |
| Share of orders to test | 1 to 100 % |
| Budget | $5 to $2,000 |
| End date | At most 90 days ahead |
The form estimates the monthly cost from your last 30 days. One test can run at a time, and a test checks at most 2,000 orders. Each test provider checks the address on its own, without a backup provider.
Every provider answer is billed like one verified order, without the repeat-address discount; provider errors are free. The test pauses when the budget or the spending cap has no room for the next order. You can Pause, Resume, Edit and Stop test. A stopped test can't be resumed, but its results stay.
Reading the results
- Providers compared: for each provider, the share of answers marked as a problem (Needs checking or Invalid), errors, response time and the amount charged. A wide "likely between" range means too few orders to tell the providers apart yet.
- Agreement: how often each pair of providers gave the same result.
- What happened later: "Caught" (marked, then corrected or the delivery failed), "Missed" (accepted, then corrected or failed) and "Likely false alarms" (marked, but the customer confirmed the address).
- By country and Orders: the results per country, and each tested order with every provider's answer.
Routing suggestions
When your A/B tests or backtests show that another provider let significantly fewer address problems through in a country, the Providers page shows a suggestion with Apply. Without enough evidence it says there are no suggestions yet.