Shopify knows what you sold. Meta knows what it spent. Neither can tell you what you made. picoask lands every one of those systems in a warehouse you own, then answers questions about them the way an analyst would — with a hypothesis, a test, and the evidence attached.
Every channel you sell through and every system you run on, landed in one place on one timeline — in a warehouse you keep.
Ask in plain language. picoask proposes the competing explanations, tests each one against your data, and reports which survived.
The metrics that drive your outcomes are re-checked on their own cadence, and an adverse move is projected forward before it lands on revenue.
Every warning names the move it implies — pause the ad set, shift the budget. You approve it, or let it run on auto-approve; picoask carries it out and watches what happens.
Meta reports the conversions Meta believes it caused. Google reports the same order again. Shopify and Amazon report what was actually banked. picoask syncs all of them into one warehouse, on one calendar, so "what did this campaign actually make" becomes a question with a single answer instead of three conflicting ones.
Orders, line items, refunds, discount codes and customers come across from Shopify; marketplace orders and returns from Amazon, Myntra and Nykaa, on a shared marketplace model so every channel reads side by side; campaign, ad set and ad performance from Meta and Google; sessions and traffic sources from GA4; settlements and fees from your payment providers; delivery and return status from your shipping platform. Budgets and targets that only live in a spreadsheet come too.
Most tools collapse a refund into a column on the order, which dates the money to the day the order was placed — so "how much did we refund in June" lands in the wrong month and quietly stops being answerable. picoask stores a refund as its own fact with its own date, and keeps the order booked where it belongs.
The same care goes into the rest: a restocked return with no money moved is recorded as goods returned, not revenue lost; discount codes carry the amount actually given away, not the configured percentage; and money is stored in the currency it settled in, never silently converted at import.
Ask a question and picoask plans an investigation: it writes down the explanations that could account for what you're seeing, fetches exactly the data each one needs, tests them, and tells you which held up — and which it ruled out.
SPRING20 was applied to 1,840 orders in March, of which 71% came from returning customers who bought at full price in February.
| Code | Orders | Given away | Returning |
|---|---|---|---|
| SPRING20 | 1,840 | $41,220 | 71% |
| WELCOME10 | 612 | $7,480 | 4% |
| BUNDLE15 | 203 | $3,110 | 38% |
Revenue tells you a month too late. picoask builds a graph of what drives what in your store — CPM into traffic, traffic into orders, return rate into net revenue — and puts the drivers under watch, each re-checked on the cadence its own grain deserves.
Every series is de-seasonalized against its own stored profile before anything is scored, so your Black Friday spike isn't an alert and a quiet Sunday isn't a crisis. An incomplete current period is trimmed, and a feed that's lagging suppresses the check entirely rather than reporting a cliff that is really just missing data.
When a driver moves the wrong way, its measured elasticity is used to project the move onto the outcome — and concurrent driver moves are combined into one net number, so an offsetting pair doesn't generate two contradictory alarms.
Bad early signs get stopped. An ad set burning budget with no attributed orders past day three is exactly the pattern Metric Watch detects — so the pause arrives as a one-click action, not a sentence you have to go and act on yourself.
Budget moves incrementally, never in one jump. Reallocation shifts a small slice of daily spend from a worse-performing campaign to a better one, re-measures on the next cycle, and keeps going only while the evidence holds — reverting automatically if the result gets worse. Deliberately unglamorous: one large reallocation on one day of data is how automated tools lose people money.
You decide how much it decides. By default every action waits for a human yes, and logs who approved it and what it changed. Once you trust a kind of move, switch it to Auto-Approve Mode and picoask runs it on its own — still logged, still reversible.
Some campaigns aren't meant to convert. Brand, a launch, a partner commitment — mark a campaign or ad group Do not touch and picoask keeps reporting on it but never pauses it or moves its budget, however its numbers look.
Each answer arrives with the query it ran, the rows it read, and the test it used — so you can check it rather than take it on faith. See how it works →
Shopify takes a read-only custom app you create yourself — no App Store listing, no review. Ad platforms, analytics, payments and shipping connect the same way. Everything picoask asks for is a read scope. Shopify setup guide →
picoask reads the schema, works out which metrics matter and how they relate, and generates a starting dashboard — so the first question you ask already lands on a modelled warehouse rather than raw tables.
Start asking. Tick the outcome metrics you care about and Metric Watch takes over from there — re-checking their drivers on cadence and telling you when something is about to move.
Only what you let it. Analysis runs on read-only access to your store, ad accounts and payments. Actions — pausing an ad set, shifting budget — wait for your approval each time unless you've switched that kind of move to Auto-Approve Mode, and anything you mark Do not touch is never changed. Every action is logged and reversible.
All of it, as long as your custom app grants the read_all_orders scope. Without it Shopify's API exposes only the last 60 days. picoask won't quietly import two months and present it as your history — the initial import stops and tells you, and you can grant the scope or accept the shorter window deliberately.
In a warehouse provisioned for you, which you keep. You can also point picoask read-only at a database you already run — PostgreSQL, MySQL, Redshift, Snowflake or BigQuery — and have it analysed alongside the connector data.
No. Connecting the sources is the whole setup — picoask derives the schema, the metric relationships and the first dashboard itself. From there it's plain language.
WooCommerce is live on the same commerce model, and the ad, analytics, payment and shipping connectors are platform-independent. Everything on this page applies — see the full integration list.
picoask is in private beta. Tell us what you sell and we'll get you a seat.