Algeria AgencyDigital marketing
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.
The article on digital marketing classes conversion as the least sold of the five spends because it is the least billable, and as nearly always the first thing to do. This page picks up from there and starts with the question of the unit.
A conversion rate is a fraction. People nearly always look at the numerator — the number of actions — and forget that the denominator moves too, often more. A campaign bringing less qualified traffic lowers the rate without anything having got worse.
There is something more awkward: in most of the Algerian businesses we work with, what the page produces is not a purchase but a message. So the rate measured on the site stops before the two places where money is actually lost.
The first is the hour after that message. The second is the doorstep, at delivery. This article treats both as conversion steps, because they are, and because nobody bills for them.
A rate is a fraction, and both halves move
Start with a reflex that avoids half of all misreadings: when a rate changes, look at the denominator first. A rate that rises because traffic fell is not an improvement, it is a loss of volume.
The commonest case is this one: a campaign stops, traffic halves, only the visitors who were already searching your name remain, and the rate doubles. The report announces a success, and turnover has gone down.
The symmetrical case happens when a broad campaign opens: traffic triples, the rate collapses, and the agency explains that the site needs rebuilding. Nothing needs rebuilding; the mix changed.
The methodological consequence is simple and rarely applied: a rate only compares to itself, at constant source. You measure separately the visitors who came from a search on your name, those from advertising, and those from a sent link, because those are three populations with nothing in common.
And a rate never compares to a sector average. None of the ones in circulation covers this market, your traffic or your definition of a conversion, and the article on content marketing explains why we do not publish that kind of figure.
Here, a conversion is a message
On most Algerian sites there is no buy button, or there is one and nobody uses it. What the page produces is a conversation: a message on a messaging app, a call, sometimes a form.
That is not a fault to correct. It is this market’s real buying behaviour, and the article on advertising on the country’s largest network describes it for its own channel: a campaign there produces messages rather than orders, because the person never leaves the app.
The consequence is that the definition of a conversion has to move. Counting messages received is useful for judging the page; it is not useful for judging money, because a message is not a sale and many will never become one.
So two measures are needed rather than one. The first stops at the page: how many visitors wrote in. The second goes to the end: how many of those messages became an accepted, paid order.
The distance between the two is the most useful piece of information in this whole pillar, and almost nobody calculates it. We regularly see it divide the figure announced by a dashboard by three or four.
The unit is the accepted order
The right unit is neither the visit, nor the message, nor even the order: it is the order accepted at delivery and paid for. Everything before that is a step, and a step is not a result.
That precision is not theoretical here, and the payments market shows it. Over one year the number of online merchants grew 26%, card transactions on the internet 38%, and the amount paid online 179% — while the base of cards in circulation grew only 9%.
The reading that matters for this page is the last column rather than the first: the payment instrument is growing far more slowly than the usage. So a majority share of sales still completes somewhere other than the page, and most often at the door, in cash.
And a sale that completes at the door may not complete. The parcel is refused, the person is not there, the price was misunderstood, the item does not match the photograph — and each of those is a conversion cancelled after being counted.
That is why we refuse to present a rate that stops at the registered order. An honest report in this market counts accepted orders, and it is always less flattering than the one the tool produces on its own.
GIE Monétique, 2025 annual balance sheet
Three corrections that beat any test
Before any optimisation, three missing pieces of information cost more than anything a test could win. They are corrected in a day and they require no tool.
The price. The article on content marketing explains the mechanism and it holds here: a page with no price multiplies messages and collapses the closing rate, because people outside the budget write in too. A range is enough to sort before the conversation.
The phone number, written so it can be dialled with one tap on a phone. It is a ten-minute technical correction and we find it missing on most sites we audit — the number is there, as an image or as text that cannot be tapped.
Delivery: which wilayas, how long, at what price, and what happens if the customer is not in. Those four lines remove a whole category of messages and part of the refusals at the door.
Those three corrections share something that explains their return: they do not try to persuade more people. They take out of the queue the people who were not going to buy, which leaves your team time to handle the ones who were.
The hour after the message
The first serious loss is not on the page, it is in the response delay. A message received at seven in the evening and handled the next morning arrives after the decision, because the person wrote to three other businesses in the same minute.
It is a conversion step and it has to be measured as one: how many messages, answered within what time, and what proportion of replies happened within the hour. Those three numbers are counted by hand over two weeks.
The correction is almost never a tool. It is an organisational decision: who watches, at what hours, and until when in the evening. The article on reputation on the largest network shows that the delay is even displayed publicly on that surface.
A second correction costs even less: preparing the answers to the four recurring questions — price, availability, delivery, does it exist in something else. Written in advance they are pasted in ten seconds and they halve handling time.
And a third, counter-intuitive one: reply even when the answer is no. "We are out of that size, it comes back Thursday" closes a conversation cleanly and keeps the person. Silence keeps them at somebody else’s.
You probably do not have enough traffic
Here is the limit this page sets and the page on testing develops: to compare two versions reliably takes a volume most Algerian sites do not have.
The order of magnitude is easy to feel without mathematics. If you get a hundred visitors a week and three write in, a version producing four messages has proved nothing at all — the gap sits inside ordinary chance.
The practical threshold we apply is a few hundred conversions per variant before concluding anything. Below that, what you observe is noise, and deciding from noise amounts to changing your site at random while believing you are improving it.
That does not mean there is nothing to do. It means the method changes: at low traffic you do not test, you correct what is obviously missing or broken, and you judge over months rather than weeks.
It is also what makes this service hard to sell honestly, and why it is so often replaced by a rebuild. A rebuild can be billed; "add your price and reply within the hour" cannot.
The form: every field has a price
Where there is a form, it is nearly always too long, and the length has a measurable cost: every additional field turns away a share of the people who would have filled it in.
The sorting rule is to ask only for what is needed to give the next answer. For a quotation that is often two fields: what it concerns and how to reach you. The rest is asked in the conversation that follows.
Three fields are particularly costly and particularly common. The email address when the person communicates by messaging app; the title, which adds nothing; and the "company" field, which silently excludes individuals.
On a phone, an eight-field form requires zooming, scrolling, switching the keyboard and losing your text if the page reloads. That is not hesitation, it is work, and a share of people stop halfway.
The best correction is often to offer the form *and* the channel people already use. A button that opens a conversation, beside the form, catches the people who were not going to fill in eight fields and were going to buy.
The price hidden behind a private message
The practice of writing "price by private message" deserves a section, because it is widespread, it has an apparent justification, and it is expensive.
The justification is real: not giving your prices to competitors. It does not hold, for the reason the article on content marketing gives — your competitors already know your prices, it is your customers who do not.
The cost is twofold. The first is volume: you receive messages from people outside the budget, who occupy whoever answers and delay replies to everybody else. The second is trust: a hidden price is read as a price that varies by who is asking.
There is a legitimate exception, and it is not the one usually invoked: genuinely bespoke work, where the price depends on the measurement, the site, the quantity. In that case publish the calculation method and three real worked examples, which produces the same sorting effect.
The difference in outcome we observe is nearly always the same: fewer messages, more orders, and a team that stops spending its days qualifying conversations that will lead nowhere.
Delivery is a conversion step
The second serious loss happens at the door, and it appears in no site measurement tool. A refused parcel cancels a conversion already counted, and it costs two journeys plus an item that comes back handled.
The causes are known and nearly all of them are upstream. The total price not understood because the charges were not written; the size misjudged for want of a scale indication, which the content-creation article treats in detail; the object different from the photograph; and the deadline missed, which leaves time to change one’s mind.
One organisational cause is added: nobody told the person which day to expect it. A message the day before with the slot and the exact amount to have ready removes a measurable share of refusals, and it costs the time to write it.
The measure to keep is simple: the refusal rate, weekly, by area and by product family. Gaps between areas are often a sign of a carrier rather than a page problem, and that shows up within four weeks.
That is why we treat this subject as conversion work and not logistics. The editorial work on the page — writing the total price, showing the scale, announcing the real lead time — is what brings down a figure the site does not measure.
Counting by hand when the tool lies
Measurement tools tell an incomplete story, for reasons that are not mysterious: blockers, refused consent, messaging apps that open links in an internal browser, and conversions that finish off the site.
The consequence is that a dashboard undercounts visits and overcounts conversions, each for different reasons. The page on analytics goes into that; here, the only thing to hold on to is that an important decision should not be settled on those figures alone.
The manual tally is the counterweight, it costs little, and it is surprisingly convincing. Over four weeks: messages received, quotations sent, orders, orders accepted. Four numbers a week, written by hand.
Add the question asked of every new enquiry — how did you hear about us — as the article on public relations recommends. The two records together give a picture neither gives alone.
That tally has a second virtue: it forces a definition of what an accepted order is, which is precisely the decision most businesses have never taken explicitly.
What we will not count as a conversion
This section exists because the list of what counts as a conversion is where a report becomes dishonest without any figure being false.
We will not count a click on a button. A click is not a completed intention, and counting clicks on "call" rather than calls actually made inflates the result by a factor no report explains.
We will not count a visit to the contact page, nor a time on page, nor a scroll depth. Those are behavioural indicators useful for understanding a page; they are not results and they do not get presented as such.
We will not count a registered but undelivered order. It is this article’s main line and it costs us: the figure we present is lower than the competition’s, and it has to be explained every time.
Nor will we count the same person twice. A visitor who writes on two channels the same day is one conversion, not two, and manual de-duplication is feasible at small volume — it is one of the few advantages of low traffic.
What we do, and what we will refuse to do
What we will refuse: presenting a conversion rate that stops at the registered order. In this market the order refused at the door is a real share of the total, and ignoring it produces a flattering, false report.
We will refuse to run a comparison test on a site that does not receive the volume needed to conclude. The page on testing says where the threshold sits; below it, a test sold is a random decision billed as a method.
We will also refuse to propose a rebuild when the three corrections in section 3 have not been made. A rebuild bills well, it is pleasant to produce, and it does not repair a site whose price is written nowhere.
What we do: the written definition of what counts as a conversion, before any measurement; the three corrections first; the four-week manual tally; response delay measured as a conversion step; and the refusal rate tracked by area.
And what you can do this week without us: take your last twenty orders and count the ones that were accepted and paid for. Divide by the number of messages received in the same period. That figure is your real rate, and it is probably the first time you have seen it.
Frequently asked questions
What is a good conversion rate?
The question has no usable answer: no sector average covers this market, your traffic or your definition of a conversion. A rate only compares to itself, at a constant traffic source.
Our rate doubled — is that good news?
Look at the denominator first. If traffic halved because a campaign stopped, the rate rises and turnover falls. A rate that changes while the mix of sources stays put is the only one that tells you anything.
Should the price be shown or given by message?
Show it, at least as a range. A hidden price multiplies messages from people outside the budget, occupies whoever answers, and reads as a price that varies by who is asking. For bespoke work, publish the method and three worked examples.
Why are our parcels refused?
Nearly always for four upstream reasons: charges not written, size misjudged for want of a scale, item different from the photograph, deadline missed. Add a message the day before with the slot and the exact amount: it removes a measurable share.
Can we run A/B tests on our site?
Only above a few hundred conversions per variant. Below that the observed gap sits inside ordinary chance, and deciding from there amounts to changing the site at random while believing you are improving it.
Are our statistics reliable?
They undercount visits and overcount conversions, for different reasons. Keep a four-week manual tally alongside — messages, quotations, orders, accepted orders — and compare.
Where we come in
The share of orders accepted and the time to a first reply locate the loss. They do not say which of the three to take first at your company.
- We put in writing which event you are counting, before anything is measured.
- We follow the order through to payment, and then to the customer’s door.
- We take the three corrections before the question of a redesign is opened.
If more than nine orders in ten are accepted and paid, the page is not your subject, and we will tell you so within the hour.
Read next
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.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.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.
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.