Website ROI: how to calculate it honestly and what gets left out

A website is neither a cost nor an asset until somebody counts it

The site does not pay for itself is said far more often than it is calculated. Usually it rests on a feeling rather than arithmetic: money was spent, enquiries are few, the conclusion seems obvious.

The trouble is that without arithmetic it is equally easy to close a channel that works and to feed one that does not for years. Both happen.

Working out a website's return is not hard — the hard part is honestly assembling both sides: what exactly counts as income and what must not be left out of the costs.

Below: a formula that fits on a napkin, three ways of fooling yourself in the calculation, and the minimum you need in place for there to be anything to count.

What counts as income

The first fork: a site almost never brings money directly unless it is a shop taking payment on the spot.

For a shop it is simple: income is revenue from orders placed on the site, less returns.

For services income runs through a chain: an enquiry from the site, a conversation, a deal. Which means being able to tell a deal that came from the site apart from one that came by referral or from advertising. Without that the calculation becomes an argument.

And do not forget indirect income. An enquiry that arrived by phone after somebody looked at prices on the site is also the site. The only way to catch it is asking how did you hear about us, and it has to be asked every time.

The practical minimum: tag the source of every deal at the moment it appears. Sources cannot be reconstructed afterwards.

What counts as cost

This is where understating happens most, and afterwards the calculation looks prettier than it is.

Development is only the beginning. Add to it: domain and hosting, certificate renewals, updates and support, and the changes made through the year.

Then acquisition. Advertising, search marketing, text, photography. If the site lives on advertising alone, the advertising spend is the main cost line, not the build.

And the thing always forgotten: your own time. The hours you or your staff spend adding content, answering enquiries and approving things are money too, and over distance they add up to more than the build cost.

Count over a period rather than once. The build is divided across the span the site will last without a rebuild — usually two to three years.

The napkin formula

It comes down to four numbers over one period — take a month or a quarter.

How many enquiries came from the site. How many of those became deals. The average deal size. And what the site cost over that period, all in.

Then the arithmetic: enquiries multiplied by the deal rate and by the average size gives income. Income minus cost, divided by cost, is the return.

An illustration with round numbers: forty enquiries a month, one in five becomes a deal, average size thirty thousand. Income is two hundred and forty thousand. Costs for the month: eighty thousand. The return is two hundred per cent.

If there is nothing to count with, that means the things in the last section are not set up. That is the calculation's first finding.

Cost per enquiry and cost per customer

Two measures that say more than the headline return figure.

Cost per enquiry is spend over the period divided by the number of enquiries. It shows how expensive each approach is and lets you compare channels against each other.

Cost per customer is spend divided by the number of deals. Always higher than cost per enquiry, and the gap between them shows how well you handle the enquiries you get.

That gap is most often where the real headroom sits. We covered this separately: the time from enquiry to first reply affects the deal rate more than anything on the site itself. Somebody who wrote to three suppliers picks whoever answered first.

And compare cost per customer with the average deal. If it exceeds a third of the deal, the channel is running on goodwill and any rise in advertising prices closes it.

Why an average deal size is not enough

An average is a convenient number and a dangerous one, because it hides the spread.

If you have services at five thousand and at five hundred, the average describes no real deal at all. A calculation built on it produces a figure corresponding to nothing.

It is better to count by group: cheap deals separately, expensive ones separately. Often it turns out the site brings many small enquiries and no large ones, while the return rests on a couple of deals a quarter.

That changes the conclusion. A channel producing one large deal a quarter looks like a failure by enquiry count and excellent by money — and it cannot be judged by the same yardstick as a stream of small ones.

In practice: calculate separately across two or three service groups. If the spread is within about fifty per cent, an average is usable.

How long a customer lasts

A measure that turns the calculation around for anyone whose work is not one-off.

If customers come back, the income attributable to the site is not the first deal but every deal for as long as they remain yours.

The difference is a multiple. A customer bringing thirty thousand once and a customer bringing thirty thousand three years running are different economics, even though they arrived the same way.

The consequence for the calculation: payback is measured not on the first deal but on expected income across a realistic horizon. A year is a sensible boundary if you do not yet have longer histories.

And the consequence for decisions: if customers return, you can invest in acquisition more aggressively than a first-deal calculation permits.

Realistic payback periods

Reference points, so you neither expect a result before it is possible nor persist beyond reason.

A site living on advertising shows its economics immediately: enquiries arrive from day one, and within a month it is clear whether the numbers work.

A site built for search traffic works differently. For the first months it brings almost nothing: the search engine has to find the pages, assess them and decide they beat what is already above. A realistic horizon is six months to a year.

Hence a common mistake: build a site for search, look at the result after two months and close the project. That is precisely the moment when a result cannot yet exist.

And the opposite mistake: persisting into a third year, explaining the absence of enquiries by search being slow. If in a year not one commercial page has reached the visible part of the results, the problem is not the timescale.

Three ways of fooling yourself

All three occur regularly and all three produce a pleasant number.

The first: counting the build and not your own time. A site you spend six hours a week on costs more than the agency's invoice states.

The second: attributing deals that came by referral. Somebody heard about you from a friend, visited the site to check, and called. The site played a part, but it did not bring the customer, and recording the whole deal against it is self-deception.

The third: counting from the best month. One good quarter does not describe a channel; count over a year, or at least half a year, weak months included.

An honesty test: show the calculation to somebody with an interest in the opposite conclusion and ask them to find the holes in it.

What to do when it does not pay back: A site built for search realistically pays back in six…

Close it is not the only conclusion and usually not the first.

Start by looking at where the break is: too few enquiries, or too few deals from the enquiries. Those are different problems with different costs to solve.

Too few enquiries — work on which queries find you and how well the pages match them. That is slow, but it is where the main growth sits.

Too few deals from enquiries — look at response time, at the quality of the conversation, and at whether the site brings the right people at all. Often the enquiries turn out to come from people who cannot buy.

And count more than money. A site saving ten hours a month on answering the same questions pays for itself even if direct enquiries are few.

What to set up so there is something to count

The minimum. Without it any calculation is an argument about figures that do not exist.

A source tag on every enquiry: which page and which query the person arrived on.

One place where all approaches are visible: from the form, by phone, from messengers. Different channels in different places do not add up.

An outcome marker on every enquiry: did it become a deal, and for how much.

A record of the time to first reply. That single number explains differences in conversion better than any other.

And a rough account of your own time — hours a week on the site. Without it, costs are always understated.