I keep seeing big names in ecommerce state this like it's settled: don't worry about attribution until you hit a certain size, just watch your MER. But this advice has one gap in it: you can't go back later and rebuild the history you didn't keep.

To be clear, the case for it is good. Channel tracking broke after iOS 14.5, the platforms often inflate their own numbers, and with MER being total revenue divided by total marketing spend, there's no attribution logic in it.

That means no arguing about whether Meta or Klaviyo gets credit for a sale. It comes out of your bank account instead of somebody's tracking pixel, which also means nobody can game it.

Brat has written about this at length before: both about incrementality testing and about what happens when one metric starts running your business, so I won't repeat his arguments here.

The companion advice to all this is to stop over-optimizing and look at the big picture. Also fair.

But put the two together and you end up watching one number that moves without any way to find out why. And when you can't find out why, you start changing things at random. Which is the habit the second piece of advice was trying to prevent.

Less than it used to be.

When we ask brands why they haven't set up attribution yet, the answer is usually some version of "I keep meaning to." The argument about whether attribution is even worth it at their size rarely comes up at all.

You know your tracking is thin. You've known for months. But this week a supplier is late and the one SKU that actually sells is nearly out of stock. Setting up attribution has never once been the most urgent thing on that list, so it slides from one week to the next.

That's fair. And honestly, if you're short on time and short on cash, we'd tell you not to add a tool you don't need. Because realistically, one number you actually look at beats five you don't have time to analyze.

But things have shifted.

Time. Attribution setup used to mean a pixel someone had to install correctly and UTM discipline across every channel enforced by an actual human. That was basically a hire.

With most tools on the market right now, all it takes is connecting your store and your ad accounts and walking away.

For instance, in Bratrax there's no pixel to install at all if you're running on Shopify. For your ad accounts, we offer to apply our tracking templates for you. Under ten minutes, and the data starts flowing.

Money. This one has changed under everyone's feet. Attribution used to cost a few hundred to a couple of thousand a month, usually scaling with your revenue, so it got more expensive exactly as you grew. For a brand doing its first few hundred orders a month, that wasn't worth the money.

The good news: there are way more options now than there ever were, including free tiers and genuinely affordable plans that didn't exist a few years ago.

We're trying to be one of them (a dollar for the first month, then $99 flat, whatever your total sales volume). So these days you can actually compare a few options before committing to anything, and do so cheaply.

MER is one number made of two totals. It tells you whether you're efficient overall. The problem? It won't tell you which part of what you're doing is actually working.

You can hold a steady MER of 3.0 for six months while your new-customer share slides and repeat buyers carry the number for you. Two channels can have identical CAC and bring in customers whose value a year later differs by half. MER sees one figure, and both channels are lumped together inside it.

The question you'll eventually need answered is which channel brings customers who stay. Answering it needs per-customer history, grouped by when people showed up and where they came from - a cohort analysis. But when it comes to cohorts, you can't build one backwards, since knowing what a customer you acquired in March is worth by the following March requires that somebody wrote down what happened when the customer made their first purchase.

Everything else in your business is recoverable. An ad bombs, you switch it off, you lost an afternoon. You misjudge a channel for two months, you fix it in a week. Being small means you get to be wrong constantly and survive, which is a real advantage that people undervalue.

But history is the one exception. It either gets recorded or it doesn't.

And sooner or later, you will want that history to help you make a decision. As you grow, you'll have the budget to double down on one channel. You'll be looking at two that cost about the same per customer, with no way to tell which one brought people who stuck around. Keeping logs of what happened is the only way to avoid making that decision blind.

Now back to the over-optimizing advice. When your sale count is small, the KPIs can shift quickly, and dramatically.

Say a campaign brought you 10 sales last week and 6 this week. That reads like a 40% collapse. It might be nothing at all, just four people who happened to buy the following Tuesday instead. But it looks alarming, so you switch the campaign off. A brand doing 500 sales a week would never have noticed those four.

That's the actual trap. With a small sample size you can't tell a real change from a normal wobble, so you are more likely to react to randomness and call it optimization.

Once you have a year or two of history, you can look at that same dip and check how often a week like that has happened before. If it's happened plenty of times and always came back, you leave it alone.

So more history makes you steadier. It won't make you disciplined on its own, but it does show you whether a dip is worth acting on.

Recording and optimizing are two different jobs that ended up under one word. Optimizing asks something of you every week, while recording asks nothing at all, since it doesn't point at a channel or hand you a green arrow for cutting something you should have left alone.

You can connect it, let it run, and not look - just let it collect data for when you actually need it.

We've seen a handful of cases now where a brand's own sources disagree about the same order. Shopify, Google Ads and Meta each apply their own rules, so one purchase can be paid traffic in one system, unattributed in another, and direct in a third. Technically, none of them are lying, just applying different logic. Problem is they don't agree.

What actually helps in those situations is looking at the timeline for that specific order: every action and touchpoint that got attributed to it, in the order it happened. Once you can see the sequence, the disagreement usually explains itself.

We encountered a case like this recently. A shopper had bought from the same brand twice in about fifteen minutes, and the owner was trying to work out what happened behind the scenes.

Strictly speaking, that second order counted as Direct. The shopper had come back on their own and gone straight to the site. But it happened within an hour of them viewing an ad, so should it really be Direct? I'd credit the ad. Someone else might not, and they'd have a case.

That's a judgment call about your business rather than a fact sitting in your data, and the timeline is what allows the store owner to make that judgment deliberately instead of taking whatever a platform decided for them.

MER wouldn't have surfaced any of it, since MER can't tell you anything about one specific order.

If you're early in your business and every sale is one you paid for (because hardly anyone is searching for you yet and few people come back on their own), MER really is enough at this stage. Adding analytics tools to your decision-making process is a distraction from finding your next big angle.

But if you have repeat buyers, branded search, an email list doing real work, and people coming back on their own, MER stopped being enough a while ago. Unless you've had tracking installed and let it diligently collect data, the record you'd need to see which channel produces customers who stay is the one nobody was keeping.

One caveat worth noting, at the risk of sounding like a broken record: perfect attribution doesn't exist, whatever anyone's marketing says. Attribution isn't causation either, so none of it can tell you why someone bought. Our philosophy at Bratrax is to tell you what your customer did and in what order, so you can make better bets on what will make a difference for your business. And when we can't match a sale to anything, we say so instead of guessing.

Sure, we are biased, but our advice is: if you aren’t already doing it, start recording now. Don't change how you spend based on it until you have enough sales for the numbers to mean something. But don't deny yourself the opportunity to have a comprehensive log ready and waiting to help you make truly informed decisions down the line.

You don't have to take our word for any of this. We keep a demo workspace open at bratrax.com/try-demo. It's synthetic data, modeled on how real stores actually behave, and an email gets you in within about a minute.

Open any order and look at what led the customer to it. Then decide whether you agree with what we credited. That's the whole argument in one screen: whether you can see enough to disagree with the number.

So which are you, the type who keeps it to one number on purpose and tunes out the rest, or the type who wants the full context before deciding anything?

— Yuliya

P.S. Bratrax is the attribution tool we're building for DTC brands — revenue reconciled back to the actual order, no "trust us" math. We open enrollment in waves — join the waitlist and you'll get an email when the next one opens.