When a media buyer says the traffic is bad, they can mean three different things: bots instead of people, the wrong people, or the right people but too few of them for the price. Traffic quality answers all three at once: what share of visitors are real, how new and targeted they are, and whether they turn into money. To check traffic quality means to break it down into these layers rather than look at a single number.
The short definition: high quality traffic is a stream of real, non-repeating visitors who match the targeting and convert consistently. A traffic quality check starts with filtering out bots, continues with checking uniqueness and ends with comparing conversion and approval rates by slice.
What traffic quality is and why one number is not enough
It helps to picture traffic as a funnel of filters:
- Is it a real person? A scanner, a platform bot, a spy service scraper, an auto-clicker — these are all clicks you paid for that will never buy.
- Is it a new person? The same visitor coming ten times is not ten potential buyers.
- Is it the right person? Real and new, but from another country, on a device the offer is not built for, or arriving by accident from a banner in a mobile game.
- Do they pay? Lead conversion, approval, purchase — what the whole thing was for.
Each layer breaks for its own reasons, and a metric that describes one layer well says nothing about another. A high share of real visitors does not guarantee sales, and a good conversion rate on a small volume may be luck. That is why measuring traffic quality always takes a set of metrics.
Traffic quality metrics: a table
| Metric | What it shows | Warning sign |
|---|---|---|
| Share of real visitors | How many visits were made by people, not programs | A noticeable share of hosting, proxies, visits without JavaScript |
| Uniqueness | How many of those who reached the offer are new people | Many repeats from the same addresses and devices |
| Targeting match | Whether GEO, language and device match the campaign settings | IP country says one thing, browser language and time zone another |
| Conversion rate (CR) | Share of target actions among those who reached the offer | A sharp drop with the same offer and creative |
| Approval rate | Share of leads confirmed by the network | Leads exist, confirmations barely do |
| Consistency | How stable the volume is by day and hour | Spikes and dips unrelated to bid or budget |
| EPC | Revenue per passed click | Falling while click cost stays stable |
The formulas for the money metrics — CR, EPC, approval rate, ROI — are covered in the article on affiliate marketing metrics. What matters more here is how they add up to an assessment.
Share of real visitors
The main metric of the first layer. It is calculated from network and technical signals: an address from a hosting or cloud network, VPNs and proxies, no ISP, a browser without JavaScript or with traces of automation. A detailed breakdown of such signals is in the article signs of bot traffic, and how network detection works is covered in the piece on VPNs, proxies and data centers.
A subtlety people forget: every ad platform's traffic includes its own reviews — moderators, preview robots, ad verification services. This is not a defect of the source; the platform generates them itself, and they have nothing to do with the quality of the audience you bought. If you count them together with click fraud and junk placements, the source's score comes out unfairly low on review days and inflated on the rest.
Uniqueness
The second layer. Repeat visits can be normal: a person came back to think it over, reloaded the page, opened the link on another device. But when the same addresses reach the offer again and again, it is either fake traffic or a narrow audience you have already burned out. Either way, buying more of it is expensive.
Conversion and approval
The third and fourth layers are only visible through money. Real, unique traffic may not convert if it is untargeted: incentivized clicks in push networks, accidental taps in apps, clicks for a bonus. Comparing with the previous period helps here: the same funnel on the same source should produce a comparable conversion rate. A sharp drop with the same offer and creative is a sign the source started mixing in different traffic. How to tie leads to clicks so that conversion is counted at all is covered in the article on postbacks.
Do not forget about approval. A lead the network later rejects or flags as fraud is not a conversion. If scripts are submitting such leads directly through the form, the measures from the article bots submitting leads will help. Statuses are covered in detail in conversion statuses.
Consistency
Real traffic breathes: more during the day, less at night, different on weekends than on weekdays. Both a perfectly flat line around the clock and unexplained spikes are suspicious — for example, clicks multiplying within an hour with no bid change. Consistency is a weak metric on its own but a good hint about where to look.
The trap: the offer pass rate is not a traffic quality metric
If there is a filter in front of the offer, a tempting number appears — the pass rate: how many visits reached the offer. It seems logical: few passed — the filter works and the traffic is bad; many passed — the filter is leaky. Both ideas are wrong.
The pass rate describes not the quality but the composition of a particular source's traffic. With paid traffic where the platform passes a verified click ID (fbclid, gclid, ttclid and so on), 80% of visits may reach the offer — that is normal, and there is no need to tighten the filter for a prettier number. Traffic made up of direct visits, scanners and reviews gives a low pass rate simply because it contains few people.
What is truly alarming is a very low pass rate on a source that definitely has a lot of people. If only a few percent of visits from a paid campaign reach the offer, the rules are most likely cutting real people along with the bots: an overly strict GEO filter, blocking visits without a referrer on a platform that does not pass one, a harsh IPv6 filter. The reason log helps you figure it out — see the article why a click went to the White Page.
Tip. Compare the pass rate only with itself: the same flow, the same source, last week. A sharp drop with unchanged settings is a signal. The absolute value is not.
Other traps when measuring traffic quality
- Small volume. With a couple of hundred clicks any metric swings. Two leads instead of four is not a conversion rate drop by half; it is noise.
- Mixing sources. The account average hides a junk placement behind a good one. Assess by slice: source, placement, GEO, device, provider.
- Comparing different weekdays. Monday against Saturday is a comparison of different audiences.
- Revenue by conversion date. If leads are confirmed with a delay, recent days always look worse. Attribute revenue to the click date and allow for the hold.
- Clean technical metrics with zero sales. Real but untargeted traffic passes every check. Money always has the last word.
How to check traffic quality: a step-by-step guide
- Tag the link. Without campaign, placement and creative labels you cannot find where the junk sits. How to do it is covered in the article on UTM parameters and macros.
- Run a test on a small budget and wait for at least a few hundred clicks per slice.
- Calculate the share of real visitors. How many visits came from hosting, through VPNs and proxies, without JavaScript, with traces of automation. This part of the job is usually handled by ad fraud protection.
- Check uniqueness. How many repeats from the same addresses; are there addresses with dozens of visits per day.
- Verify targeting. Do country, language, time zone and device match what you bought.
- Slice by placement and provider. Junk almost always concentrates in a few places.
- Wait for conversions and approvals. Compare CR and EPC by slice and with the same previous period.
- Switch off the worst slices in the ad platform or with a filter — and check that conversions did not drop along with the junk.
If the process uncovers masses of repeat clicks from the same addresses, that is a click fraud problem with its own set of measures.
How ArtisanClo shows traffic quality
In ArtisanClo the traffic quality assessment is collected in the Traffic quality score — a score from 0 to 100 on the dashboard and in the Statistics section. It rates the traffic itself, not whether the filter is working. Here is what goes into it:
- The share of real visitors, not counting platform reviews — weighs the most. Platform moderators and robots are left out: they say nothing about the quality of the audience you bought.
- The share of new people among those who reached the offer — uniqueness.
- Conversion compared with the previous period of the same length — if there are enough leads to compare.
- How even the traffic is day to day.
The weak spot — the component dragging the score down — is named under the score. Below 200 clicks, instead of a score it honestly says «Not enough data yet»: on a small sample the score would be random. A high pass rate is not penalized — there is no ceiling on it.
Next to it are insights — up to four hints such as «Platform reviewers were turned away», «Almost nothing is reaching the offer», «Offer delivery dropped sharply» or «Many repeat visits». If fewer than 5% of visits reach the offer, the dashboard shows a hint and a notification warns that the rules are most likely cutting real people.
Where else to look:
- Quality by flow on the dashboard — problem flows at the top.
- Statistics — the shares of the three «Where they were cut» steps (network and request, browser check, check never came back), block reasons, traffic mix with CR and EPC, an hourly heatmap and weekday rhythm, slices by GEO, device, OS, browser and provider.
- Click log — every visit with the reason for the decision; the click card shows the network (VPN, proxy, data center), ad data, checks and conversions.
- «Tracker» mode — if you need an assessment rather than filtering (with the PHP file, Keitaro filter or Binom gate connection): every visitor goes to the offer, bots are only flagged, and the report shows what part of the paid traffic was bots.
For conversions to count in the score, you need conversion tracking set up — a postback from the network or a snippet for your own site's form. More on the ArtisanClo features page.
Bottom line
Traffic quality is not one metric but four layers: are these real people, are they new, are they targeted, and do they pay. Measure it by slice, on enough volume and with platform reviews set aside. Do not mistake the pass rate for a quality score: a high pass rate for paid traffic with a click ID is normal, and a very low one is what should worry you. And remember that technical metrics weed out bots, but the final word in any traffic quality check belongs to conversions and approvals. Where to start filtering is covered in the article on how to filter bot traffic.
