Estimating an online store's traffic: method and margin of error

Let's start with what no sales page will tell you: no tool knows the real traffic of a site it does not own. They all estimate it, from user panels, internet providers and statistical models.

A study comparing those estimates against real Google Analytics data shared by site owners puts the average margin of error between 30 and 50%. On individual measurements, Similarweb lands around 37% off and Semrush around 49%.

That is not a reason to do without them. It is a reason to know what to do with them. A figure accurate to within 40% is more than enough to tell a store getting three hundred visits a month from one getting a hundred thousand. It is not enough to claim a competitor does 112,000 visits rather than 90,000.

Why nobody can give you the real number

Only a site's owner sees its traffic, because only their server and their analytics count the visits. Everything else is reconstructed from the outside.

Estimators lean on three sources: panels of users who installed an extension or an app, data bought from internet providers and carriers, and models extrapolating from known search positions. Each brings its own bias, and the result is a reconstruction, not a measurement.

Hence one simple rule for everything that follows: treat these numbers as an order of magnitude, never as accounting data.

What the estimates are actually worth

The margin of error, quantified

Hold on to 30 to 50% average deviation. Concretely, an estimate of 50,000 monthly visits means "probably between 30,000 and 80,000". That is a range, and it should be read as one.

That is enough to rank, compare and rule out. It is not enough to build a forecast, nor to justify an investment on volume alone.

Why small sites are the least reliable

Accuracy rises with size. Above a million monthly sessions the gap narrows. Below a few thousand visits it explodes, for a mechanical reason: the panel underpinning the estimate contains only a handful of that site's visitors, sometimes none at all.

That is the most common trap in ecommerce research, because most of the stores you will analyse are small. An estimate of 800 visits a month could just as easily mean 200 or 4,000. At that level, stop reading the number and read the other signals.

What stays true despite the error

Good news: the biases are largely constant from one site to the next. A tool that overestimates does so because of its method, so it overestimates by roughly the same amount everywhere.

The practical consequence: comparisons and trends are far more reliable than absolute values. If the same tool gives 40,000 to one store and 8,000 to another, the one-to-five ratio is solid, even though both figures are wrong.

The four signals worth more than a number

The source breakdown

This is the most useful piece of the whole analysis, and the most neglected. Knowing where the traffic comes from tells you how the store works, what it spends and how attackable it is.

The twelve-month trend

A line rising steadily beats a high but flat volume. It points to a mechanism that works, whereas an isolated spike usually signals a one-off push, a discount or a viral video with no follow-through.

Trends are reliable because the tool's bias stays the same month after month.

The countries of origin

A high-traffic store where 80% of visitors come from a country it does not ship to is not making the revenue its volume suggests. Conversely, a small store concentrated on a single market can be far more profitable than it looks.

Advertising pressure

The number of active ad platforms and the age of the installed pixels tell you whether the traffic is bought and for how long it has been. Paid traffic sustained over months proves the economics hold, since nobody pays at a loss for long.

Reading the source breakdown

Mostly paid traffic

The store buys its visitors. It is vulnerable: the day it stops, the traffic disappears. But if it has held for several months, the margin per order absorbs the acquisition cost. That is an attackable model, provided you have an advertising budget.

Mostly organic traffic

The store has built a durable search position. This is the hardest profile to attack head-on, because that asset takes years. On the other hand it reveals the queries that pay, and those queries you can target another way.

Mostly social traffic

Dependence on an audience or an algorithm. The volume can be considerable and the conversion rate low: people watch a video, they do not necessarily buy. Be wary of impressive numbers on this channel, they rarely translate into matching revenue.

High direct traffic

Two opposite readings. Either the brand is established and people type its name, which is the best possible signal. Or the tool files there whatever it cannot attribute, which happens often on small sites. Cross-check against the search volume for the store's name before concluding.

Cross-check with clues that do not lie

Some things are not estimated: they are observed. These are your anchors when the traffic figure looks doubtful.

The number of customer reviews left over a given period is a floor on real orders. The pace of new products indicates activity. Repeated stock-outs on the same variant point to what sells. The presence of advertising pixels reveals where the budget goes. And the age of the domain places the store's history in the right order of magnitude.

When those clues agree with the traffic estimate, your confidence rises. When they contradict it, believe the clues.

From traffic to revenue: the maths and its traps

The basic calculation is simple: visits × conversion rate × average order value. A store at 100,000 visits, 1.8% conversion and a $55 basket would do $99,000 a month.

Except you are multiplying three estimates. Traffic is uncertain to within 40%, the conversion rate is unknown and ranges from 0.5% to 4% depending on the sector and the traffic source, the average basket is guessed from displayed prices. The errors do not cancel out, they compound: the result can be off by a factor of three.

Use it for what it is worth, which is to answer "is this market worth tens of thousands a month or millions?". And remember that this figure is turnover, not profit: after product cost, advertising and returns, less than 10% often remains.

Compare rather than measure

That is the practical conclusion of everything above. An isolated number teaches you nothing actionable; ten stores placed side by side with the same tool teach you a great deal.

The gaps between them are the information, and they survive the shared bias. You see who dominates, who is slipping, who buys their traffic and who earns it. Above all you spot the anomaly: the small store whose organic traffic is climbing while the others buy, the one that has been selling for years without anyone copying it.

Ten stores compared beat one store measured precisely.

The mistakes that cost the most

Taking the number at face value. It is a range. Treating it as a measurement leads to decisions built on nothing.

Mixing tools. Comparing one tool's estimate to another's is meaningless, their biases differ. One tool for the whole comparison.

Ignoring sources and looking only at volume. A hundred thousand social visits and a hundred thousand search visits do not produce the same revenue, nor the same risk.

Concluding from a single month. Seasonality distorts everything. Look at twelve months before deciding anything.

Giving up because it is not exact. The alternative is not perfect data, it is no data at all. An order of magnitude you own beats a hunch.

Frequently asked questions

Can you know a competitor's exact traffic?

No, and be wary of any tool claiming otherwise. Only the site owner has the real measurement. Everything else is an estimate, with the margin of error described above.

What site size makes the estimate usable?

From a few tens of thousands of monthly visits, the order of magnitude becomes usable. Below a few thousand, stop relying on the number and switch to reviews, catalogue and advertising pixels.

Is it legal to analyse a competitor's traffic?

Yes. These estimates rest on aggregated data and on public information available to any visitor. You are not accessing anything private: you are observing what the store shows everyone.

Traffic is only one signal among several. Catalogue, best sellers, installed apps and tech stack are read the same way, and we detailed them in our guide to analysing a competitor's Shopify store.

Where to start this week

Three actions to turn this reading into decisions.

First pick ten stores in your market, using one single tool for all of them, and record for each the estimated traffic, the source breakdown and the twelve-month trend. Then add the verifiable clues: number of recent reviews, catalogue size, ad platforms detected. Finally rank those stores not by volume, but by the gap between what they receive and what they appear to get out of it. That gap is where the positions are.

That survey takes an afternoon by hand. Evidio does it in one click from your browser, on any Shopify store: estimated traffic and sources, advertising pixels and active platforms, installed apps, best sellers and tech stack, with an export so your ten stores line up in the same table. The extension is free, and it is precisely the comparison, not the measurement, that it lets you do quickly.