TLDR: Almost every analytics request starts with a metric. The project only really begins when the metric fails to answer the question behind it, and that question is rarely a neat choice between X and Y. It is usually about where to invest and what to prioritise. Start from decisions like those and the metrics that matter fall out. Start from the metric and you build a dashboard nobody opens twice.
Click to discuss this article with your favourite AI
Almost every analytics request starts the same way. Someone asks for a metric. Conversion rate, average order value, bounce rate, sessions by channel. On its own that is a perfectly reasonable ask, and if a single number is genuinely all someone needs, they get it and we are done.
The project starts later, at the moment someone looks at that metric and realises it does not answer the question they actually had. The number moved, and they still do not know what to do about it. That gap, between the metric on the screen and the decision sitting behind it, is the whole job. It is also why building straight from a list of requested metrics gives you a beautiful dashboard that nobody opens after the first week. A metric on its own does not tell you what to do. It goes up or down and leaves you to supply the meaning, and most of the time nobody can.
The question is not what metric you want, it is what you would do differently
So I do not start with metrics. I start by asking the people who will use the data what decisions they are trying to make and cannot.
These are rarely a clean choice between a specific option A and option B. Most people cannot tell you in advance whether the answer is X or Y, and they do not need to. What they can tell you is what matters to their business and where they are trying to decide to put their effort and money. Which kind of merchandising actually works. Whether to back site search or product placement. Whether to shift advertising budget. Whether the next round of work belongs at the top of the funnel or the bottom. Those are the real questions, and every one of them is a decision about where to spend, not a metric to admire.
So ask a room full of stakeholders what metrics they want and you get every term they have ever heard. Ask them instead what they would do differently if they could see something they cannot see today, and the conversation shifts completely. Now you are talking about where to invest, and a direction like that points at exactly the data that would inform it and nothing else.
The clearest example I have of this came from a single question about average order value.
“I need AOV” is not a request, it is a symptom
An ecommerce lead I worked with wanted average order value on the dashboard. Fair enough, it is a standard number. So I asked why it mattered, what they would actually do with it. And they could not really say. It was a number they felt they were supposed to watch.
That is not a criticism of them. It is the normal state of things. Average order value is a headline metric, and headline metrics get asked for precisely because they are familiar, not because they tell you what to do.
So we pulled it apart. Average order value is only two things multiplied together: how many items are in a typical order, and how much each item costs. For this retailer that was an order of about sixty-three dollars, made of roughly two items at around thirty-one dollars each.
Now look at the two levers. Price per item was close to fixed. The catalogue was mostly books, where the price on the cover is the price, and the business was already at the premium end of its market. They could not discount their way to a higher number without hurting margin, and they could not raise prices the market had already set. So one of the two levers behind average order value was effectively bolted down.
Which left basket size. The only realistic way this business could move its average order value was to get a second item into more orders. And that is a completely different brief. “Show me AOV” is a chart. “How do we get more people to buy two items instead of one” is a decision that merchandising, bundling, related product placement and free shipping thresholds can all act on. Same starting metric, but only one version tells anyone what to do on Monday.
Two ways in, depending on who you are talking to
Not everyone can have that conversation on the spot, and that is fine. In practice I run this in two tiers.
With the two or three people in the business who are genuinely fluent with data, usually a head of ecommerce and whoever owns the marketing spend, I go deep, one on one. They can articulate a decision if you give them room, and their answers shape the whole framework.
With everyone else I use a more structured format, a workshop or a short survey with prompts, because an open “what do you need” question just produces the wish list again. Instead of “what metrics do you want,” I ask “if you could see one thing you cannot see today, what would you stop arguing about,” and people tell you what is actually stopping them from acting.
The point of both is the same. Move every person from naming a number to naming a decision.
What you end up with
The output of this stage is not a dashboard and not a metric list. It is a set of decisions, each tied to the specific data that would inform it, ordered by which ones the business most needs to make and currently cannot.
Some of those you will be able to answer with what you already collect. Some you will not, and that gap is the most valuable thing to come out of the exercise, because it tells you what the rest of the work is for. Every tag you fix and every metric you define after this has a reason to exist. It answers a real question that a real person is waiting on.
That is the difference between analytics and a measurement foundation. One gives you numbers. The other gives you a reason to look at them.
There is a catch hiding inside “which of these can we answer,” though. Before you can check whether the data is right, you have to know what your site actually lets a customer do, every path, every button, every action. That map comes first, and it is next.
Recognise this as a problem for your business?
If you have a dashboard full of numbers and still cannot answer the questions that would change what you do, that is the gap I work in. I help ecommerce businesses turn vague reporting into decisions they can act on.