Filters vs segments: the key difference
A counter filter works at the moment data is collected. A visit that matches it is never stored, and you cannot bring it back later. A segment works differently: it is applied inside a report, the data stays intact, and you simply look at one slice of it.
That leads to a practical rule. Use filters only for things you will never need to analyse: your office IP address, a test domain, service pages. Anything that depends on the question you are asking, such as Tashkent only, mobile only or non-ad traffic, belongs in a segment that you can edit or delete at no cost.
Which Yandex Metrica filters to set up first
Right after installing a counter, a few actions are enough. They take little time, and your data is clean from the first day instead of from the day someone notices odd numbers.
- exclude the office's external IP address, if it is static;
- turn on the setting that ignores your own visits for the owner and anyone with counter access;
- count visits only on the production domain;
- remove the admin area and service pages from statistics;
- keep the standard robot filtering and avoid tightening it without a reason.
Excluding by IP: where it goes wrong
An IP filter is only as good as the address behind it. Many providers assign dynamic addresses, so after some time the filter may point at strangers while your own staff reappear in reports. Ask your provider whether the office address is truly fixed before you rely on it.
Never add mobile internet addresses: one address can be shared by many subscribers, and you would cut out real customers along with your own phone. An overly wide range or a single mistyped digit does the same damage. Also remember that an employee opening the site from home is not covered by an office IP filter.
After saving, test it. Open the site from the office and check the real-time report. Your visit should not appear.
Managers, sales staff and test orders
The biggest distortion comes from people who use the site all day. Sales managers place orders on behalf of customers and fill in forms while talking on the phone. Their visits look unusually active with a high conversion rate, and they skew your picture of traffic sources.
If your team does not share one network, ask them to work in a browser signed in to a Yandex account that has access to the counter, so the setting that ignores your own visits applies. Also agree on a single marker for tests, for example a special word in the name field or a dedicated link parameter. Test payments through Click or Payme should be marked too, otherwise your goals will show purchases that never happened.
Test copies of the site and foreign domains
A common situation after development: the site moves to the production domain, but the staging server stays online with the same counter code. Its visits mix with real ones. Either remove the code from the test version or restrict counting to the main domain.
When you list the allowed addresses, include the versions with and without www and any subdomains. Otherwise the filter can cut off part of your live traffic.
How to avoid losing data to a filter
After adding a filter, watch total visits for a few days. A sharp drop means the rule is cutting too much. It also helps to compare the same period with another source, such as your ad account.
Keep a short log: which filter, why, when and who added it. A year later, when someone asks why data fell, that note saves hours. Mark the date the filter went live in annotations so you know where the break sits when comparing periods.
Clean counters are the base of end-to-end analytics
A clean counter underpins everything else: judging Meta and Google ads, calculating cost per lead, connecting to a CRM. If internal traffic reaches your goals, every payback calculation will look better than reality. At markdev we begin end-to-end analytics projects with exactly these basic checks. If you want to be sure your Metrica counts only customers, describe your task and we will review the counter and suggest what to fix first.
FAQ
Can I recover visits removed by a filter?
No. A filter acts during collection, so excluded visits are not saved. Test every rule in real-time reports and apply filters only to traffic you will never need.
Which is better, an IP filter or the setting that ignores my own visits?
They complement each other. An IP filter removes everyone in the office regardless of account, while the setting works for anyone signed in to a Yandex account with counter access, even outside the office.
How do I exclude staff who work from home?
An IP filter will not help. Have them sign in to a Yandex account with counter access, or add a marker to internal links and remove those visits with a segment.
Do I need to filter robots manually?
Usually the standard filtering is enough. Strict counter-level rules are risky because you can lose real visitors and cannot restore them.
Let’s look at your task for free
Leave your contact — in a 15-minute call we’ll show where leads get lost and what to fix first.