Algeria AgencyConversion
Traffic is not the problem; the page is
Acquisition cost and conversion rate multiply. A point gained on the page is worth the same as an equivalent discount on the click, costs less, and does not stop when the invoice does.
There is a recognisable moment in the life of a business that advertises: the campaigns work, traffic rises, the reports are good, and revenue does not move. That is the moment the problem changed places without anybody announcing it.
The cause is nearly always the same. Acquisition brings people to a page, and that page converts a fraction of them. Those two numbers do not add: they multiply. Doubling traffic on a page that converts one visitor in a hundred gives exactly the same result as keeping the traffic and converting two in a hundred — except that the first is paid for every month and the second once.
That is what the second tier does, and it is also why it is harder to sell than the first. Nobody wakes up wanting to improve a conversion rate. People want traffic, because traffic is visible, whereas a conversion rate is an abstract number nothing makes spectacular.
This article explains where visitors are lost, in what order to plug the leaks, what search does that advertising does not, and when a test has proven nothing at all. All of it is verifiable without us, and some of it can be applied in an afternoon.
A leaking bucket is always filled from the top
When enquiries are short, the universal reflex is to buy more visitors. It is the most expensive reflex there is, because it treats the symptom by the one means whose price recurs.
The comparison that describes the situation correctly is the leaking bucket. Pouring more water does increase what stays at the bottom, in some proportion, and the hole stays exactly the same size. The amount lost grows at the same rate as the amount poured.
What makes this hard to hear is that it does not show up in advertising reports. Those measure the water poured: impressions, clicks, cost per click. The hole is measured on your site, in another tool, and often nobody looks at it because it is nobody’s job.
The practical consequence is an order of operations, not an opinion. Before raising a budget you need to know what proportion of current visitors does what you want them to do. If that number is unknown, raising the budget is betting on a hypothesis it would have been enough to measure.
Two numbers that multiply, never add
The cost of a customer is the cost of a visitor divided by the proportion of visitors who become customers. It is a division, which means the second term carries as much weight as the first, and nobody treats it that way.
One example makes it concrete. At twenty dinars a visitor and one customer per hundred visitors, a customer costs two thousand dinars. Take the page from one per cent to two, without touching the budget, and the same customer costs one thousand. Get the same drop by negotiating the visitor cost and you have to halve it — which, on an auction platform, is not something one negotiates.
The difference between the two routes does not stop there. The gain made on the page is still there next month, and the month after, including on traffic you do not pay for — visitors from a search, a shared link, a recommendation. The gain made on the cost per click disappears the day the campaign stops.
That asymmetry is what justifies the tier’s ordering: the page first, the increase after. A business that reverses the two pays more, for longer, for a result it does not own.
What search does that advertising does not
Advertising and search are sold as two ways of getting the same thing. They do not get the same thing, and confusing them produces either a disappointment or a pointless expense.
Advertising is a tap. It opens in a day, it yields volume proportional to budget, and it closes as fast as it opened. It is the right tool when volume is needed now, when an offer is dated, or when a hypothesis has to be tested in two weeks rather than two quarters.
Search is an asset. It is built slowly, it does not answer a budget increase with an immediate traffic increase, and it keeps producing when you stop funding it. It is the right tool for the questions your customers ask all year and type in the same words.
The two together have a property neither has alone: they inform each other. The queries that convert in advertising point at the pages worth writing, and the pages that rank point at the queries it has become pointless to pay for. That is the only real reason to buy them together rather than in sequence.
Content is not a flow, it is a stock
Most content calendars are built like television schedules: the grid must be filled, so material is produced. That is flow reasoning, and it produces volume that evaporates.
Flow content lives three days in a feed and disappears. Stock content answers a question somebody types into a search engine, and it goes on answering it for years. Both have their use, but only one accumulates, and a business producing only flow starts from zero every week.
The question that sorts the two is simple and comes before writing: is anyone searching for this? If the answer is no, the text may still deserve to exist — to speak to existing customers, to show a finished job, to announce a date. But it should not be expected to bring visitors, because that is not what it does.
The honesty test for stock content is this: is it useful to somebody who will never buy anything from you? If so, it will be found, cited and shared. If not, it is a brochure, and a brochure does not rank.
A page that converts is not a beautiful page
Site redesigns most often fail for a reason that sounds like a compliment: the new version is more beautiful. It is airier, more modern, closer to what is being done, and it converts less.
What decides on a page can be counted on one hand: what you sell, to whom, at what price, with what proof, and what to do next. A page that answers all five within one screen height beats a magnificent page where the five answers are spread over four sections and two clicks.
Price is the element most often removed, and that is almost always a mistake. "We would rather discuss it" is a sentence that protects the seller and costs the buyer a call they do not want yet. A range with its conditions filters better than a form, and it filters for free.
Proof is the second missing element. An assertion about your seriousness is worth nothing; a photograph of a real job, a company name, a date, a checkable figure are worth the rest of the page. It is also the part we cannot manufacture for you, and the one that most often stalls a redesign.
The form is a toll gate, and every field has a price
Every field in a form turns away a share of the people who started it. This is not a theory: it is the mechanics of effort demanded before anything is received in return.
We put no number on that loss, and that is deliberate. The only published scales on the subject come from the vendors of form software, whose product is sold on the promise of exactly that measured improvement. Citing their figure would treat a sales argument as an independent measurement — the same refusal as section seven of the Starter article.
What can be checked without a scale is the sorting rule: every field must serve either replying to you or not wasting time. A phone number serves. A company name serves. "How did you hear about us" serves only you, and asking it before answering the person makes the customer pay for your reporting.
The short version is a test you can run this evening: fill in your own form, standing, with one thumb, on your own mobile connection. The number of fields at which you give up is the number of fields to delete.
Speed is part of the conversion rate
A slow page does not lose unhappy visitors: it loses visitors who will never know what it contained. They close before it renders, and they appear in your advertising statistics as paid clicks.
The dominant cause is weight. The Web Almanac’s global crawls give the order of magnitude of a median page, and most of that weight is unresized images, fonts loaded in several weights, and third-party scripts added one at a time with nobody ever removing them.
That weight is not judged in the abstract but against the network carrying it. In Algeria that network is mobile for almost every subscription, which means variable latency, throughput that changes with location, and a data bill the visitor pays. A three-megabyte page tests beautifully on office fibre and fails in a street.
The good news is that the most effective fixes are the least glamorous: resize images to the size they display at, remove scripts nobody looks at, load the typeface in a single weight. None of the three requires a redesign, and the three together are often enough to bring a page back within bounds.
HTTP Archive, Web Almanac 2025 — global July crawls
"All the platforms" does not mean all at once
The second tier opens advertising to the full set of platforms, and it is worth saying immediately what that means, because the misunderstanding is predictable.
It does not mean broadcasting everywhere simultaneously. The threshold rule described in the entry tier’s article does not vanish because the budget grew: it moves. With two channels instead of one, the question becomes whether the second clears its own threshold, and the answer is often no for another quarter.
What it does mean is that the choice is no longer forced. At this level a channel can be opened because the first has plateaued, because an audience is not reachable there, or because a season justifies it — and it can be closed again without dismantling what was working.
The sequence we follow is nearly always the same: one profitable channel, then the page fixed, then a second channel funded with the money the page released. It is slower than the version where everything opens in month one, and it is the only version where you know, at the end, which change produced the result.
The test that proved nothing
Comparing two versions of a page is the right instinct, and it is also where the greatest number of false conclusions are drawn with the greatest confidence.
The problem is volume. A test sets two rates against each other, and a rate measured on a small number of visitors varies enormously by chance. At two hundred visitors per version, the observed gap between two identical pages can reach several points — which means a test that size produces a winner even when there is nothing to win.
We publish no numeric threshold because there is not one: the volume required depends on the starting rate and on the size of the difference you are trying to detect. A site converting at ten per cent detects a quarter-sized improvement quickly; a site at one per cent seeking the same relative improvement needs far more traffic for the same certainty.
The consequence is uncomfortable and worth stating: most Algerian SME sites do not have the traffic to test what they would like to test. For them the right method is not testing but correcting the obvious errors, which are numerous, free to fix, and require no statistical proof at all.
The month a channel gets stopped
Stopping is a harder decision than starting, because it looks like an admission. It is nevertheless the only thing that distinguishes a managed budget from a renewed one.
The signal is not a bad month. It is a channel whose cost per enquiry rises while volume stalls, over two or three consecutive months, after the obvious corrections have been made. That combination means the useful audience is exhausted and you are now buying the people who are left.
What to avoid is the emotional stop — the one that follows a mediocre week, before the data is legible. It often costs more than the channel does, because it resets the accumulated learning and starts month one again.
The discipline is to write down, before launching, what would trigger a stop. A threshold decided cold, knowing what a customer is worth, is a decision; the same threshold discussed mid-flight becomes a negotiation between the wish to continue and the fear of having lost.
When the page improves, everything else gets cheaper
The most useful property of a conversion gain is that it is not confined to the channel it was won on. It applies to everything arriving at the page, whatever the source.
Traffic from a search benefits, traffic from a shared link benefits, traffic from a recommendation benefits, and none of those three was paid for. Improving the page is therefore the only acquisition spend whose return also applies to free traffic.
There is a second effect, less obvious and more important over time: a better conversion rate widens the set of channels that are profitable. A platform too expensive at one per cent becomes viable at two, which reopens options rightly discarded six months earlier.
This is why we treat the page before raising budgets, including when it delays the visible part of the work by a month. The reverse order works too and costs more — a legitimate choice, provided somebody made it knowingly.
What we do, and what we refuse to do here
At this level our work is to measure where visitors stop, fix pages in order of decreasing cost, write the content that answers questions actually being typed, open a second channel once the first is measured, and deliver a monthly report that ends in a decision.
What we refuse fits in a sentence: we do not take a conversion engagement on a page we do not consider worth keeping. When the right answer is to rebuild it, we say so, rather than billing six months of corrections on a structure we know is doomed.
We also refuse to announce a conversion gain in advance. The figure depends on your starting point, your sector and your offer, and an agency promising a percentage before seeing your pages is promising another client’s result. We say what we are going to fix and in what order, not what it will produce.
Finally, we do not test what cannot be tested. On a low-traffic site we correct the obvious errors and say so plainly, rather than installing a testing apparatus that would produce random winners and impressive reports. Billing for a method the volume renders inoperable is the politest way to sell nothing.
Frequently asked questions
Does improving conversion mean rebuilding the whole site?
Rarely. The first gains almost always come from a published price, an added proof, a shortened form and resized images. A rebuild is justified when the structure itself prevents those corrections.
How long before a correction shows an effect?
The effect is immediate; measuring it is not. You need enough visitors to distinguish a real change from chance, which ranges from days on a busy site to weeks on a quiet one.
Does search eventually replace advertising?
It reduces it on the queries you rank for; it does not replace it. The two do not capture the same moments: one waits to be searched for, the other goes and finds people.
Can you work on a site we did not build?
Yes, provided we have access and the structure allows changes. We say at the audit stage whether the site is modifiable or will have to be replaced, and we would rather say it before taking the engagement.
Who writes the content, you or us?
We write, you correct the facts. The details that give a text its value — a real lead time, a trade constraint, a frequent objection — come from you and cannot be invented.
What if the conversion rate does not move?
We say so with the numbers, and we look elsewhere: the offer itself, the price, the campaign’s targeting. A rate that does not move after several sound corrections is almost always an offer problem, not a page problem.
Where we come in
A page that loses eight visitors in ten loses them again when you pay to bring them: the same money, spent twice.
- We order the fixes by what each one costs you, the dearest first.
- We ask for access to your statistics rather than a rebuild budget.
- We tell you when a test has too little traffic to mean anything.
On a hosted shop whose pages you cannot edit we will decline: billing for recommendations nobody can apply is not a service.
Read next
When it works: what breaks next, and in what order
Growth breaks a business before it breaks a campaign. The constraint does not disappear, it moves — and never to where you are watching.Conversion rate: the unit is the accepted order
In this market the event on the page is a message, not a purchase. The two biggest losses come afterwards, and neither one shows on the site.Why they do not buy, and how to find out
Almost every optimisation programme starts by changing things. The diagnosis costs a week and makes the rest useful.
Let us talk about your project
A free audit, no commitment: we look at your online presence and tell you what is holding it back.