The Ecomm Analyst

Every few weeks someone asks me which attribution tool they should buy, and the honest answer is that the question is backwards. Almost every bad purchase I have watched an operator make started the same way, with a shortlist of platforms assembled before anyone had decided what question the platform was supposed to answer. You end up buying a feature set instead of a decision, and then you spend six months trying to make the feature set relevant.

So here is the sequence I actually run when a brand in the $1M to $20M range asks me to help them pick something.

Write down the decision you cannot make today

Not the metric you want. The decision. There is a real difference. “I want better attribution” is not a decision. “I need to know whether to keep spending $18k a month on TikTok” is a decision. “I need to know if my email revenue number is double-counting people who would have bought anyway” is a decision.

This matters because different decisions need genuinely different measurement approaches, and the tools are not interchangeable. If your question is about channel-level budget allocation at moderate spend, multi-touch attribution gets you most of the way there. If your question is whether a channel is incremental at all, MTA will confidently give you a wrong answer and you want a holdout test instead. I wrote about which approach fits which spend level in more detail, but the short version is that buying an MTA platform to answer an incrementality question is the most common expensive mistake in this category.

Check whether your data is clean enough to be worth measuring

This is the step people skip, and it is the step that determines whether the tool will work at all. Attribution platforms do not fix broken tracking. They inherit it, and then they present it back to you with more confidence and better charts, which is arguably worse than having no tool.

Before I look at a single vendor I check three things. Whether UTMs are consistent across every paid channel, including the ones the agency set up two years ago and nobody has touched since. Whether the Shopify checkout is passing order data reliably, including for subscription and draft orders. Whether there is any server-side tracking in place or whether everything still runs through a browser pixel that Safari has been quietly degrading for years.

If those three are a mess, fix them first. A brand I worked with last year was about to sign a four-figure monthly contract to solve what turned out to be a UTM template problem in one Google Ads account. The UTM audit took an afternoon and saved them the contract.

Match the pricing model to your shape, not just your budget

This is where the category gets genuinely annoying, because vendors price on different axes and the comparison is not apples to apples.

• Pageview-based pricing rewards high average order value. If you sell a $400 product and get modest traffic, you win here.
• GMV or revenue-based pricing means your bill grows as you grow, whether or not you are getting more value. Triple Whale works this way.
• Ad-spend-based pricing, which is how Rockerbox structures contracts, ties cost to media budget rather than store size.
• Per-seat or per-source pricing shows up mostly in reporting tools rather than attribution platforms, and it punishes agencies hardest.

Run the math at your projected size eighteen months out, not today. I have seen operators pick a GMV-priced tool at $2M in revenue and get genuinely angry at $6M when the same product costs them three times as much.

Also, notice how much of this category refuses to publish pricing at all. Polar Analytics, Rockerbox, Hyros, and INCRMNTAL all route you to a demo before you can see a number. That is a legitimate business model, but it is also a signal about who they think their buyer is. If a vendor will not tell you the price without a sales call, you are probably not their target customer at $3M in revenue.

Ask how the model works, and be suspicious if you cannot get an answer

Every platform in this space uses some blend of deterministic matching, probabilistic modeling, and outright guessing. That is fine. What is not fine is a vendor who cannot explain which is which.

The question I ask on every demo is simple. When your dashboard says this Meta campaign drove $40,000, how much of that is a matched order and how much is modeled? If the answer is vague, or if the rep pivots to talking about accuracy percentages, that tells you something. You are going to have to defend these numbers to a founder or a board eventually, and “the tool said so” does not survive that conversation.

Weight setup time heavily

The tool you actually configure correctly beats the more sophisticated tool you half-implement. This sounds obvious and it gets ignored constantly, because feature comparison spreadsheets do not have a column for “will anyone here actually finish setting this up.”

If you do not have a data person, be honest about that. Platforms built around warehouse integration and custom modeling assume someone owns them. Lighter tools like ThoughtMetric, which sponsors this blog, sit at the other end and are designed to be running in an afternoon rather than a quarter. Neither approach is better in the abstract. The right one depends on whether you have the internal capacity to feed the heavier option.

Run a parallel period before you switch anything off

Give it thirty days minimum with the new tool running alongside whatever you have now. Two things happen in that window. You find the tracking gaps that the demo environment hid, and you build the intuition to know when a number looks wrong, which is the actual skill.

Expect the numbers to disagree with your ad platforms. They always do, and that disagreement is structural rather than a bug. What you are looking for is whether the new tool disagrees in a way you can explain.

If you want the current shortlist rather than the process, I keep a running comparison of the best e-commerce attribution tools with pricing checked as of this month.