Eight questions every attribution vendor should be able to answer

If you can’t answer that question with confidence, you have an attribution problem, not just a creative problem.

For eight years I’ve run paid acquisition at Inceptly across every major attribution tool — Triple Whale, Northbeam, Hyros, Polar, Rockerbox. Our job has been to make whatever was already installed do its work, and to figure out which numbers we could actually trust.

The questions below are what we hand operators when they ask which tool to buy, or whether to switch. Not feature comparisons. Not vendor pitch decks. Eight questions you can run on whatever tool is in your stack right now.

1. How do your numbers reconcile to my Shopify orders and ad-platform reports?

Most attribution tools’ revenue figures don’t match the source of truth. Pull 10 orders from yesterday and compare line by line — the vendor’s reported revenue against your Shopify admin, their reported spend against the native ad platform. Day-one discrepancies become every-day discrepancies. If their numbers don’t tie out at $50K/month in spend, they will absolutely fall apart at $200K.

The vendor that can’t reconcile is selling you a story, not a measurement.

2. How do you define a customer, an order, and revenue?

These should be written-down definitions. When a subscription renews — one customer with two orders, or two customers? When someone refunds — does the order vanish or stay flagged? When the same person buys from two stores — two customers or one? When a payment plan executes its third installment — is that a new order?

Different tools answer differently. You discover the mismatch three months in, when your CFO’s spreadsheet stops matching your dashboard and neither of you can figure out which number is right.

3. Pick one order from last week — ideally one where the customer touched multiple devices. Walk me through every channel that got credit, the stitching logic, and the math.

This is the demo killer.

The decent tools handle the first part — Triple Whale and Hyros will pull up the order and show you the touchpoint list. That’s table stakes now. The real test is the next question: how did you decide which touchpoint gets credit, and how did you stitch a phone session to a desktop checkout?

That’s where the math has to be reproducible, not just visible. If the answer drifts into “our proprietary attribution algorithm” or “advanced multi-touch modeling,” the credit logic is a black box — you can see the clicks but you can’t reproduce the output.

Pick a cross-device order. Watch how they handle the stitching and the credit math, not just the click list. That single demo tells you more than any feature comparison sheet.

4. What happens to attribution when a customer refunds, returns, or cancels a subscription?

One of the most under-discussed attribution problems. Does the order disappear from reports? Stay flagged but still counted? Does it adjust your CPA and ROAS retroactively, or only going forward? What happens at the campaign level versus the blended-metrics level?

You’ll process thousands of refunds. The vendor’s policy directly affects your reported margins. The honest vendors will tell you the specific trade-off they made — usually something like “campaign ROAS uses gross, blended metrics use net.” The opaque ones will hand-wave and say “we handle it.” When you hear that, push for specifics. The mechanism either exists and they can describe it, or it doesn’t.

5. For brands in my niche, what percentage of revenue typically lands in your “Direct” or “Unattributed” bucket — and what specifically do you do to shrink it?

For most tools, the bucket sits in the 15-30% range. Sometimes more. That’s up to a third of your revenue you can’t explain — and you’re paying $500-$1,500/month for the dashboard that hides it.

The vendor that knows their customers will be able to quote a range. The vendor that doesn’t will hand-wave or say “it varies.” Same with the methodology: ask for the specific technical mechanism — server-side pixel, first-party tracking, post-purchase signal recovery. “Industry-standard tracking” is not an answer. If they can’t quote a typical number and a real method, the bucket will be whatever it ends up being.

6. What happens when our business changes — new product type, new payment flow, multi-store, subscription mechanics?

Most attribution tools ship a fixed template at signup, and your business bends to fit it. Ask the vendor to walk you through a specific change a customer made last quarter — a new object, a new property, a new revenue stream — and how long it took to implement.

If “we’ll add it to the roadmap” is the answer, the template is frozen. Your business will keep evolving. The tool won’t.

7. Give me the full pricing picture — starting price, how it scales, contract terms, price increase history.

Most vendors quote a starting price and stay vague on the rest.

The full picture: when my revenue doubles, what happens to your bill? Annual contract or month-to-month? Are existing customers grandfathered when you raise prices, or do they pay the new rate? When was your last price increase?

A vendor who’s raised prices three times in three years without grandfathering is telling you what your bill will look like in two years — regardless of what they’re charging now. Most VC-backed vendors raise prices to hit board-mandated revenue targets, not because their cost to serve you went up. Your renewal is somebody else’s quarterly slide.

8. If I leave — what data is mine, in what format, and how do I get it out?

Can you export the raw dataset to your own warehouse — BigQuery, Snowflake, S3? In what format? Today, or on the next tier, or via a support ticket?

Walled data is the most reliable indicator of a business model that depends on you not leaving. If the vendor can’t or won’t let you take everything when you go, the platform is the product, and your data is the leash.

Why this list matters now

If you can’t trust your attribution layer, you can’t tell whether a creative is fatiguing or just being mis-tracked. You can’t tell whether scaling spend is profitable or whether the numbers will fall apart at the next tier. You can’t tell whether a new offer is a real winner or a leaky funnel.

Every question on this list is a way to force opacity into the light.

A week ago, we launched Bratrax Lite.

It’s the attribution tool I built after eight years of running other people’s stacks at Inceptly. The one I wished I’d had all along.

It’s built for Shopify DTC. If you’re not on Shopify yet, we’re not your tool — we’ll get there. If you are, here’s how Bratrax Lite answers every question on the list above:

• Numbers reconcile to Shopify orders and ad-platform reports. Tested for weeks pre-launch against live client data. The Shopify order record is ground truth — we don’t invent a parallel revenue figure that drifts from yours.

• Documented entity definitions, customizable to your business. A customer, an order, revenue — written down, queryable inside the tool, and adjustable through a conversation with the in-app AI.

• Decomposable attribution math for any order. Ask the AI to walk you through the channels, weights, and formula on any specific order from last week. Single-touch defaults so the math is additive and verifiable.

• Honest refund handling. Campaign ROAS uses gross, blended metrics use net. The trade-off is explicit, the AI tells you which is which, and you can switch the calculation if you want refund-adjusted numbers.

• First-party server-side pixel. Pulls the Direct bucket to 3-5% on our best clients so far — vs. the 15-30% industry typical. The Shopify order record helps recover the source when the pixel misses.

• Your model flexes when your business does. New product type, new payment flow, multi-store, subscription mechanics — you adjust the data model through the AI, not by filing a support ticket and waiting for the roadmap.

• Flat $79/month for life as a founding member. No GMV scaling. No annual contracts. No demo gate. Existing customers don’t pay new-customer rates — ever.

• Your data stays yours. Lives in your own ClickHouse instance, exportable to BigQuery, S3, CSV — anytime, no support ticket required. Plus an MCP endpoint, so you can point your own Claude at it using your own Anthropic API key.

We’re engineers trying to build a product, not run a sales operation. The tool reflects how we work — solve a real problem, document the trade-offs, name the limits. The list above isn’t a pitch; it’s the same criteria we’d run on any vendor, including us.

Run the questions on whatever you’re paying for now. Run them on us. If we don’t measure up on every line, you’ll know. That’s the point of the list.

We’re inviting founding members in batches. The first 100 lock in $79/month for life. After the founding cohort fills, pricing moves to $99/month for everyone after — and the founding rate is gone for good.

If you want the full feature breakdown first: bratrax.com. If you want a founding spot directly: bratrax.com/signup.

— Brat

💌 Like this newsletter? Let's continue the conversation

Get in touch with us by responding to this email or tagging us on LinkedIn or Instagram, and sharing your thoughts. Your feedback helps us keep our newsletter relevant and interesting.