Skip to content
qiyas.cc
Problem Solver
cc · blog · Business 2026-11-10 · 11 min

The most common technical problems in business, and how to fix them

The technical problems small and mid-sized businesses run into most often: the symptom, the cause and the fix. Numbers with sources, advice from my own projects.


A technical problem never arrives at a business labelled “technical problem”. It arrives as a sentence: “a customer says the site won’t open”, “the payment went through twice yesterday”, “the guy who wrote to us on Instagram just disappeared”, “the developer left and nobody knows the password”.

This post is about what sits behind those sentences. For each problem there are three parts: what you see, why it happens, and what to do. The numbers have sources, listed at the end. Most of the advice comes from my own work as a developer in Baku, because I have had to solve these exact problems in my own projects.

The other posts in this series go deep on single topics. This one is the map.

1. The website is slow

What you see. On a phone the site sits on a white screen for a few seconds. Money goes into ads, sales do not come back.

Why. Huge images, a cheap server far away, dozens of scripts on every page, no caching. And very often the site was tested on the owner’s fast office connection and new laptop, not on a customer’s three-year-old phone.

What to do. Measure first. Google judges speed with three numbers: the main content should appear within 2.5 seconds (LCP), the page should respond to a tap within 200 milliseconds (INP), and things should not jump around while it loads (CLS under 0.1). These have to hold for 75 percent of real visits, not for the developer’s machine.

One old number that still holds up: in Google’s data from 2016, more than half of mobile visits were abandoned when the page took longer than three seconds to load. The data is old. People have not become more patient since.

The full version: Why is my website slow, and how to fix it.

2. The domain or the SSL certificate expires

What you see. One morning the site is gone, or the browser shows a red “not secure” warning. Email has stopped too.

Why. Someone bought the domain three years ago on their own card. The card expired, and the reminder emails go to an inbox nobody reads anymore.

Two facts worth knowing:

  • For generic domains like .com, ICANN requires registrars to warn the owner roughly one month and one week before expiry. After a domain is deleted there is usually a 30 day window to restore it. So the warnings do get sent. The question is who receives them.
  • SSL certificates are getting shorter. In 2025 the CA/Browser Forum voted to cut the maximum lifetime of a public certificate from 398 days down to 47, in steps ending in March 2029. Let’s Encrypt moves its default certificates to 64 days in February 2027 and to 45 days in February 2028. At that pace a certificate renewed by hand will be forgotten. It is only a matter of when.

What to do. Turn on auto-renew for the domain and name the person who owns the card and the inbox. Renew SSL automatically, never by hand. On emlakagenti.az, when an agent signs up for their own listings site, the DNS record, the web server config and the SSL certificate are all created without anyone touching them. That is not laziness. Tracking certificates for many sites by hand simply does not work.

3. The domain, the hosting and the accounts belong to someone else

What you see. You have parted ways with whoever built the site. The domain is in their account, the server is on their card, the admin password is on their phone.

Why. At the start it feels convenient: “you set everything up, I’ll run the business.” The cost shows up on the day you split.

What to do. Open the registrar account with the business’s own email and add the developer as a user. Same for the server, the code (GitHub or GitLab), the payment provider’s dashboard and the social accounts. Keep one short document: which account, in whose name, where. Depending on a single developer is a bad business model, even if that developer is me.

4. There are no backups, or there are and nobody has ever restored one

What you see. The server dies or the database gets encrypted, and it turns out the “backup” either does not exist or lives on the same server behind the same password.

Why. Backups are invisible while everything works. So nobody checks them.

Verizon’s 2026 Data Breach Investigations Report found ransomware (attacks that encrypt your data and demand payment) in 48 percent of all breaches, up from 44 percent a year earlier. The same report says small organizations are disproportionately hit by it.

What to do. The US cybersecurity agency CISA puts it plainly: keep offline, encrypted backups of critical data and test restoring them regularly. The common rule of thumb is 3-2-1: three copies, in two different places, one of them completely off-site. The part people skip is the most important one: a few times a year, restore a copy somewhere clean and watch the site or the database actually come up. A backup you have never restored is not a backup. It is a hope.

5. Passwords get stolen and accounts get taken over

What you see. The Instagram account “got hacked”, a stranger got into the admin panel, fake invoices went out to customers from your email.

Why. The same password everywhere, two-factor authentication (MFA) switched off, and one employee’s password already circulating from some other site’s breach.

The 2026 DBIR has a separate section for small and medium businesses, covering 7,152 confirmed breaches. The top way in was exploiting a software vulnerability (26 percent), followed by stolen credentials (13 percent) and phishing (9 percent). A third party was involved in 55 percent of those breaches, and the human element in 45 percent.

What to do.

  • Turn on MFA everywhere, email first. A 2023 Microsoft study found that MFA cuts the risk of an account being compromised by 99.22 percent, and by 98.56 percent even when the password had already leaked. An authenticator app does better than SMS.
  • Use a password manager. A different long password for every account, none of them to remember.
  • Stop forcing password changes. The US NIST standard (SP 800-63B) now says outright that services shall not require periodic password changes. Instead, new passwords should be checked against a list of known breached ones, and a password used on its own should be at least 15 characters long.
  • Check for leaks. Have I Been Pwned’s Pwned Passwords service tells you whether a password has appeared in a breach without you sending the password: only the first five characters of its hash leave your machine.

For admin panels I wrote Django Admin Secure Pack: when someone opens /admin/, a one-time key is sent to Telegram, and without that key nobody gets in. A small thing, but it closes the “anyone who knows the password is inside” hole.

6. Software never gets updated

What you see. The admin panel has been saying “updates available” for months. Nobody wants to touch it, because “last time we updated, the site broke”.

Why. An update looks like work with no visible result and a visible risk.

According to the 2026 DBIR, exploiting vulnerabilities is now the most common way into an organization across all breaches (31 percent). Worse: only 26 percent of the critical vulnerabilities in CISA’s catalogue of actively exploited bugs were fully fixed by organizations, and the median time to a full fix was 43 days.

What to do. Put updates on a schedule instead of leaving them to luck: once a month, on a test copy first. Cut down the plugins, every extra plugin is an extra door. Write down who updates the server, the database and the site itself.

7. People do the computer’s job

What you see. A manager spends half the day copying the same data from one place to another. The same listing is posted on five sites by hand, and again on all five when the price changes.

Why. Each tool was bought on its own and none of them talk. Sometimes they cannot: the main Azerbaijani listing sites, for example, have no public API for listings.

What to do. For one week, write down every repeated task: what, who, how many minutes. Anything repeated more than ten times a day is a candidate for automation.

Some of mine:

8. The data is everywhere and nowhere

What you see. What you know about a customer is spread across a WhatsApp chat, a manager’s notebook, an Excel file and a few people’s memory. When an employee leaves, their customers leave with them.

Why. The business grew, the process stayed the way it was on day one.

What to do. One centre: a CRM, or a simple panel built around how you actually work. Off-the-shelf CRMs often do not fit. I went through this for real estate in detail: CRM for real estate agencies: what you actually need.

As the system grows a different problem starts: the code stops speaking the language of the business and every change gets more expensive. One cure for that is domain-driven design: What is DDD, and how do you apply it?

9. Customers are on WhatsApp, the business is not

What you see. Messages arrive at night and get answered in the morning, or never. The same question comes in fifty times a day.

Why. WhatsApp lives on one person’s personal phone. When they sleep, the business sleeps.

What to do. A bot and an operator panel on the official WhatsApp Business Platform. In my projects WhatsApp is usually the main channel to the customer: Khalif Estate runs a Baku real estate agency end to end over WhatsApp, and Oyun Qurdu has a WhatsApp bot and a live support console. What a bot really does for a business, and what it does not: What a WhatsApp bot does for a business.

10. Payments: the money left, the order did not arrive

What you see. A customer says the card was charged, the system says the order is unpaid. Or the opposite: one order, two charges.

Why. The integration was written for the happy path. It trusted the redirect back from the bank page and never checked the payment provider’s signed notification (the callback). When the connection dropped, it blindly retried.

What to do. Take the payment status only from a callback whose signature you verified, or from a direct status request to the provider. Never auto-retry a request that moves money when you do not know what happened to the first one. I built the Epoint Python SDK around exactly that rule: if a call that moves money times out, the library does not retry it. It raises an error carrying the order id, so you check the status first. Step by step: How to integrate Epoint payments into your site.

11. The first one to notice the site is down is a customer

What you see. You learn that the site has been down since yesterday from a customer’s message.

Why. Nobody is watching. The server does not complain and email does not either.

What to do. Simple monitoring is enough: something that checks the site every minute and writes to Telegram when it fails. Uptime Kuma is a free, open source option you host yourself, with Telegram among dozens of notification channels, and it can watch certificate expiry too. Setting it up is an hour of work. A lost day costs more.

12. One person knows everything

What you see. “Only Elchin knows how that works.” Elchin is on holiday.

Why. Teams call this the bus factor: how many people would have to disappear before the project stops. In a small business it is usually one.

What to do. The account list from point 3, a short runbook (“where the server is, how to restart it, where the backups are”), and the code in a repository, not on one person’s laptop. It does not have to be perfect. It has to let someone else start working when that one person is not around.

Check this week

What How to check
Domain Look at the expiry date and auto-renew. Whose email is the account on?
SSL Renewed automatically, or by someone by hand?
Backups When was the last test restore?
Email and admin panels Is MFA on?
Passwords Any shared passwords? Is there a password manager?
Updates Anything waiting in the panel?
Monitoring If the site goes down, who finds out, and how many minutes later?
Accounts In whose name? Is there a list?
Speed Open the site on a phone, on mobile data, and count the seconds.

One more thing

None of these get fixed by technology alone. Most of them get fixed by looking in time, thinking calmly, and dropping the habit of saying “it still works for now”. I wrote separately about what goes on in your head while you solve a hard technical problem: The psychology of solving technical problems.

If you recognise your business somewhere in this list, have a look at what I do or just write to me.

Sources

related