Algeria AgencyIT & infrastructure
IT: what actually stops a business, and in what order to deal with it
This pillar is bought after the incident, so in the worst conditions. One hour of questions puts the five families back in order.
The five families of spending grouped under the word "IT" — the audit, the infrastructure, support and hardware, the network, security and surveillance — share one thing that is not technical at all: they are nearly always bought after the incident rather than before it.
A disk died, a laptop was stolen, a file vanished, a camera was missing in the wrong place. The decision is therefore taken in the worst possible conditions: under time pressure, with no inventory, from the first person who picked up the phone.
That is why this pillar produces more wasted spending than the other three combined. Not because the hardware is poor, but because nobody had time to ask what actually stops the business — and the answer is nearly always invisible.
This article will sell you nothing. It gives you the question that ranks the five families, the inventory you can compile yourself in an afternoon, and the four things we refuse to bill for even when they are ordered.
What this pillar contains, and why it arrives too late
The audit is the doorway to everything else, and it is the only line that produces no object. It is also the one people skip, because it hands over a document where the others deliver a machine you can put on a desk.
The infrastructure is where your data lives: a server, hosting, storage and its copies. It is invisible, it cannot be shown to a visitor, and that is why it gets budgeted last while deciding everything.
Support and hardware are the most tangible family: the workstations, their repair, their replacement, their purchase. They are the easiest invoices to sign, and their effect on whether you stop trading is the weakest of the five.
The network is the cheapest layer in the pillar: cables, ports, a switch, a line, telephony. It is also the layer producing half of what gets reported as "the server is down", and nobody documents it.
Security and surveillance close the list, and they carry an asymmetry worth stating early: what happens to a small business is almost never a targeted attack. It is a stolen laptop, a shared password, an account nobody closed, an email asking for a bank account number to be changed.
The question that settles it: what stops the business for a day
One question properly ranks the five families, and it is about neither the age of the hardware nor whatever is making the most noise: what, if unavailable tomorrow morning, would stop you invoicing?
Write the list down. Which file, which machine, which access, which line. Forbid yourself general answers of the "the system" kind: you need the names of objects, because objects are what get protected.
The list is nearly always short — three or four entries — and nearly always invisible: a spreadsheet that exists on one workstation, an accounting file, one mailbox, a telephone line your customers have known for ten years.
Then compare that list with what you spent on IT last year. The gap between the two is the subject of this article: the money went towards what could be seen, and the list itself cannot be seen.
That ranking is free, it takes an hour, and it is the only thing that makes the eleven sections below decidable. Without it every proposal looks reasonable, including ours.
The inventory nobody has
Nothing can be decided without knowing what you own, and almost no small business owns that list. It is not hard to compile: it is merely dull, and nothing makes it urgent until the day it is.
Four columns are enough. Which machines and from which year. Which still receive security updates. Where each important set of data sits and who holds a copy. Who holds administrator access, including the people who have left.
It takes an afternoon and it is the audit minus the invoice. What an audit adds is an eye that knows what to look for; what it cannot add is the list itself, because you are the only people who know which file matters.
The commonest finding is the same everywhere: somebody who has left still has access, and a machine everyone believes is backed up is not. Those two discoveries cost nothing at all and they move the order of the spending.
The practical consequence: compile the inventory before asking for any quotation. A supplier who receives it prices your real problem; a supplier who receives no list prices their catalogue, and they are not wrong to.
The hardware is not the spend, the stoppage is
A machine costs what it costs, once. A stoppage costs the day of the people who cannot work, plus whatever was not sold, and it repeats for as long as the cause stays in place.
The calculation is done once: the number of people prevented from working multiplied by the cost of one of their days, plus one day of average turnover. That single number decides the rest of this pillar, and almost nobody has written it down.
And that number has changed, which nobody updates. Over one year the number of online merchants grew 26%, card transactions on the internet 38%, and the amount paid online 179%, while the card base grew only 9%.
The useful reading for you is this: when a business can be paid without anybody travelling, a day of downtime stops costing idle hours and starts costing turnover directly. The price of a stoppage has gone up without the hardware changing at all.
Redo the calculation every year if your sales channels have moved. A stoppage cost estimated in 2022 is a poor reason to refuse a backup in 2026, and it is nonetheless the reason we hear most often.
GIE Monétique, 2025 annual balance sheet
Repair or replace: the rule that avoids both mistakes
There are two mirror mistakes and they cost about the same: replacing what works, and repairing what should have left the fleet two years ago.
Two rules settle most cases. The first is arithmetic: if the part costs more than a third of a replacement, replace it. Below that, repair, and repair is the right answer far more often than a salesperson will say.
The second is not negotiable: a machine that no longer receives security updates is not a working machine. It is a working machine with a permanently open door, and no antivirus closes that door.
Between the two lies the commonest case and the worst handled: the slow machine. A machine that is slow because its disk is full, or because it carries ten years of installations, is not a machine to replace — it is an afternoon of work.
What we refuse here is simple: replacing hardware that works. It is the easiest invoice in this pillar, the least defensible, and the one that leaves untouched the problem we were called about.
You depend on a connection you do not control
Everything else in this pillar assumes a line that works, and that line belongs to somebody else. It is the least discussed dependency and the one you have least purchase on.
The internet market observatory published by ARPCE counts, for the second quarter of 2025, some 59.10 million internet subscriptions in Algeria, of which 88.71% are mobile and 11.29% fixed.
The usual reading of that figure is commercial. For a business it says something else: the spare connectivity available in this country is in a phone. Your fallback is therefore a phone, whether you planned for that or not.
Decide now which of your functions have to survive on a mobile connection — most often taking an order, invoicing, and one mailbox — and check that they do, on a day you choose rather than a day chosen for you.
The corollary is the one people forget: if a function needs the office line in order to exist, it needs a written plan for the days that line does not. "We will see" is a plan, and its price is one day.
- Mobile subscriptions88.71%
- Fixed subscriptions11.29%
ARPCE, internet market observatory, second quarter of 2025
Half of all "server failures" are cable failures
The network is the cheapest layer in this pillar and the one producing the most misdiagnosis. A workstation that can no longer reach the server gets reported as a server failure, and it rarely is one.
What a diagnosis costs when nothing is labelled can be measured: an hour spent unplugging cables to see what falls over, during which other things fall over too. That hour is paid at every incident, not once.
Three free things remove it. Label both ends of every cable. Photograph the cabinet. Draw one page showing what is connected to what. One afternoon, once, and the hour disappears from every later incident.
The honest limit of this section: the difference between professional and consumer equipment is another subject with its own article, and we are not reopening it here. It comes up after the diagram, not before.
The check that costs nothing: ask whoever maintains your network to show you the diagram. If it does not exist, that is the first thing to commission, before any equipment — and it costs less than a single switch.
Cloud is not cheaper, it is differently expensive
Cloud is not a saving, it is another shape of spending: a purchase becomes a subscription, less at the start, more every month, and a bill that follows your growth instead of following your budget.
Two questions decide it, and neither is about price. First: what happens when the connection goes down? Some functions tolerate an hour with nothing; others stop the business, and you already know which thanks to the list in section 2.
Second: can you retrieve your data, in a readable format, without the service that produced it? Ask for an export before signing rather than when leaving, because the answer to that request is the only real trial.
What we refuse: a migration whose exit path has never been tried. A service you cannot leave sets its price as it likes on the day you can no longer leave, and that day always arrives at the wrong moment.
The mixed answer is nearly always the right one and it is rarely proposed: some things hosted, some local, and the boundary drawn by the first question rather than by the fashion of the moment or by the supplier’s business model.
Security: what actually happens to small businesses
What happens is almost never a targeted attack. It is a laptop stolen from a car, a password used across five services, a former employee’s account left open, and a very polite email asking for a supplier’s bank details to be amended.
The three cheapest defences cost nothing: a distinct password per service, a second check on the mailbox, and closing accounts on the day somebody leaves rather than the month after. No product replaces those three.
The fourth is a habit rather than a purchase: any change to payment details is confirmed by calling a number you already had, never the number given in the message. That one rule avoids the heaviest loss on the list.
We will quote no incident figures for Algeria: we know of no dated, checkable series, and the international statistics describe other populations. What we can say is what we see, and we present it as exactly that.
Finally, the check that comes before every product: has your backup ever been restored in front of you? The article on backups explains how to build one; here it is a purchasing question, and a restore never attempted is not a backup, it is a hope with a subscription.
Video surveillance solves one precise problem, and only one
A camera answers the question "what happened", after it happened. That is a real service, it is worth its price in several trades, and it is the only one a camera performs.
It prevents nothing on the list above. No camera stops a shared password, a fraudulent email, or the still-open access of somebody who has left, and those three are the expensive ones.
So cameras do not come first. A business fitted with cameras and no tested restore has protected what can be seen and left its invoicing exposed: that is the configuration we walk into most often, and it is this whole article in one sentence.
There are also obligations attached to filming employees and customers: informing people, how long recordings are kept, who may view them. Find out what applies in your case before installation, not after an incident.
And the same trap as with backups: a recording nobody has ever tried to retrieve is not a recording. Ask today to see footage from three weeks ago, and you will learn what you actually bought.
What to check before signing
The first check is about the accounts: who holds the administrator access and the licences? In your name, with a company email address, or you are renting your own systems without knowing it, and the rent goes up when you want to leave.
The second is about the response time, in writing, and about what starts the clock: the call, the ticket, or the moment somebody reads it. Without that precision, "rapid response" is decoration on a quotation.
The third is a demonstration rather than a promise: a restore performed in front of you, on real data, before signing. A screenshot with a green tick is not a restore, it is a screenshot.
The fourth is about the end of the contract: what becomes of access to your data, in what format, and how long you have to retrieve it. That question is asked at the beginning because it cannot be negotiated at the end.
The fifth is a test, and in this pillar the answer is especially telling: ask what they would refuse to sell you. A supplier who sells cameras to a business with no tested backup will sell you anything, in any order.
What we do, and what we will refuse to do
What we will refuse: selling cameras or a firewall to a business whose backup has never been restored. The order matters more than the product, and the refusal belongs before the quotation, while saying it still costs us something.
We will refuse to replace hardware that works, and to run a migration to hosting whose exit has not been tried. Those two lines are the most profitable in our catalogue, and that is exactly why they are here.
We will refuse a support contract with no written response time and no restore trial. A subscription promising nothing checkable is a subscription to a promise, and we would rather not bank it.
What we do: the list of what stops the business and the inventory compiled before any proposal; the accounts and the licences in your name; a restore performed in front of you; and the cheapest layer checked before the most expensive, even when that shrinks the invoice.
And what you should do without us this week: the four-column inventory, which takes an afternoon, and the cost of one day of stoppage, which takes ten minutes. Those two figures will settle the conversation better than an audit, and they belong to you.
Frequently asked questions
Where do we start when everything looks urgent?
With the list of what would stop you invoicing tomorrow morning, in names of objects rather than generalities. It takes an hour, it is nearly always short, and it ranks the five families of spending on your behalf.
Should we pay for an audit?
Compile the inventory yourself first: machines and their age, updates, where the data sits and who holds a copy, who has administrator access. An audit adds an eye that knows what to look for; it cannot add that list.
Cloud or a local server?
Two questions decide, and neither is about price: what happens when the connection goes down, and can you export your data without the service hosting it. The mixed answer is often right and rarely proposed.
Should we replace a computer that has become slow?
Not if it is slow because its disk is full or because it carries ten years of installations: that is an afternoon of work. Replace if the part costs more than a third of a new one, or if it no longer receives security updates.
Do cameras count as security spending?
They answer "what happened", after the event, and that is useful. They stop neither a shared password, nor a fraudulent email, nor the access of somebody who left — so they do not come before a tested restore.
How do we verify a backup?
By asking for a restore in front of you, on real data, before signing. A green tick in an interface shows that a copy was written, not that it can be read back: those are two different claims.
Where we come in
Two lists written in an afternoon — what would stop you invoicing tomorrow, and what a day of downtime is worth — beat three quotes received without them.
- We order the five families by what an outage costs you in dinars and in hours.
- We name what can wait a year at no risk, and there is always something.
- We refuse to add hardware until a file has been recovered in front of us.
Until you can lay hands on a file deleted last month, buy nothing else from us: the rest protects something you would lose anyway.
Read next
Moving the business: what to order three months ahead, and the weeks you exist at two addresses
A move tests your whole system at once, on a date a landlord set. The long pole is never the van.What a named lead changes, and what it does not
The top tier adds a person rather than a feature. What a single point of contact really alters, what it does not, and how to check the difference.IT audit: start by opening the cupboard
An audit is done standing up, on the premises, opening doors. What is found there decides everything else, and it is written down nowhere.
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.