How to Test a New Ad Network on Your Website Without Losing Revenue
Every publisher eventually wonders whether a different ad network would pay more. The problem is that testing a new ad network feels risky: you have working ad slots, predictable income, and no appetite for a month of lower earnings while you find out. In this guide we walk through how to test a new ad network safely, so you can get a clear answer about performance without gambling your current revenue on it.
Why Publishers Avoid Testing (and Why That Costs Money)
Most publishers stay with their first network far longer than they should. The reasons are understandable. Switching code is tedious, reporting is unfamiliar, and the fear of a sudden revenue dip is real. The result is that many sites run the same setup for years without ever knowing whether they are leaving money on the table.
Ad network performance is not static. Demand shifts by vertical, geography, and device. A network that was the best fit for your traffic two years ago may not be the best fit today. The only way to know is to compare, and the only way to compare responsibly is to test a small, controlled slice of inventory rather than flipping everything over at once.
Step 1: Establish Your Baseline First
Before you add a single new ad tag, document how your current setup performs. You need a reference point, or the test results will mean nothing.
Pull at least two to four weeks of data from your existing network and record these metrics per ad slot and per device type:
- eCPM (effective cost per thousand impressions) — your revenue per 1,000 impressions, regardless of pricing model
- Fill rate — the percentage of ad requests that returned a paid ad
- Impressions and page views — so you can tell revenue changes from traffic changes
- Revenue per 1,000 page views (RPM) — the number that matters most, since it combines eCPM and fill rate
If you are not sure how to read these, our posts on eCPM for publishers and fill rate explain each one in detail. Write your baseline down. Memory is unreliable when you are looking at a dashboard that disagrees with it.
Step 2: Choose What to Test and Where
The goal is to isolate the new network so a bad result cannot hurt more than a small percentage of your income. There are three common ways to do this.
Option A: Test a single ad slot
Pick one placement, ideally not your top earner, and route it to the new network while everything else stays put. This is the simplest method and works well for banner inventory. The weakness is that slot position heavily influences performance, so you are comparing the new network on one position against your old network on different positions. Use your baseline for that specific slot to keep the comparison fair.
Option B: Test a new ad format
If your current network only runs display banners, adding a format it does not offer is the lowest-risk test of all. Nothing existing changes. For example, adding a popunder or an IM floater alongside existing banners gives you a clean read on the new network's demand without touching current revenue. Just keep an eye on ad density so the extra format does not degrade the user experience.
Option C: Split traffic
Send a portion of your page views, say 10 to 20 percent, to the new network's tags and the rest to the old one. This can be done with a simple server-side rule, a tag manager, or by testing on a subset of pages or a secondary domain. Traffic splitting gives the fairest head-to-head comparison because both networks see similar users on similar positions. It requires a bit more setup, but it is the method we recommend when you want a decisive answer.
Whichever option you choose, cap the test at a share of inventory you could afford to see underperform for a few weeks. For most publishers that is somewhere between 10 and 25 percent.
Step 3: Use Fallback Ads to Protect Unsold Impressions
A new network may not fill 100 percent of your requests on day one, especially while its system is learning your traffic. Unfilled impressions earn nothing, and that lost revenue will unfairly drag down the new network's results.
The fix is to configure a fallback. Most networks let you specify what to show when they have no bid: a house ad, a second network's tag, or an affiliate link. Set this up before the test starts so unsold inventory is still monetized. Our guide on setting up fallback ads for unsold traffic covers the mechanics.
Also check your floor prices. If you set an aggressive minimum CPM on the new network, you may suppress fill and misjudge its demand. Start with the floor at or slightly below what your current network pays for the same slot, then adjust once you see real bid data. See our post on publisher floor prices for how to find the right level.
Step 4: Give the Test Enough Time
The most common testing mistake is calling a winner too soon. Ad network results fluctuate day to day and week to week for reasons that have nothing to do with the network. Weekend traffic behaves differently from weekday traffic. Advertiser budgets reset at the start of the month. A single large campaign can start or stop and move your eCPM by a wide margin.
Plan for a minimum of two full weeks, and ideally three to four. Within that window:
- Do not change anything else on the site that could affect ad performance, such as layout changes, new content categories, or traffic sources.
- Compare full weeks to full weeks so weekday and weekend mixes match.
- Expect a learning period. RTB platforms take a few days to understand your inventory and bring in the right demand. Early results usually improve as the system optimizes.
If you want a deeper framework for deciding when you have enough data, our post on statistical significance for media buyers applies just as well to publisher tests.
Step 5: Compare the Right Numbers
When the test window closes, resist the urge to compare raw eCPM alone. A network with a higher eCPM but a lower fill rate may earn you less in total. Line up these figures for the test inventory against the baseline you recorded in Step 1:
| Metric | What it tells you | |--------|-------------------| | RPM (revenue per 1,000 page views) | The true bottom line for the slot or traffic segment | | eCPM | How much advertisers value each filled impression | | Fill rate | How much of your inventory the network can actually sell | | Impression discrepancy | Whether the network's count matches your own analytics | | Payment terms | How quickly earnings reach you and the minimum payout |
Impression discrepancy deserves special attention. Every network counts slightly differently, and a gap of a few percent is normal. A large gap is a red flag that needs explaining before you scale. Our post on ad impression discrepancies explains what is normal and what is not.
Beyond the numbers, consider the operational side. Did ads load quickly, or did the new tags slow your pages? Did any inappropriate creatives appear? Was support responsive when you had a question? A network that pays marginally more but hurts your page speed or user trust is not a win. See page speed and ad latency for how load times translate into revenue.
Step 6: Scale Gradually, Not All at Once
If the new network wins, do not switch everything over the following morning. Expand in stages. Move another slot or another 20 to 30 percent of traffic, wait a week, confirm results hold, and repeat. Performance that looked strong on 15 percent of your inventory sometimes softens at 100 percent because demand for your traffic is not unlimited.
Many publishers also discover that the best outcome is not one network or the other, but both. Different networks have different advertiser pools, and running them together, whether through fallback chains or by splitting formats, often earns more than either alone.
If the new network loses, you have still gained something valuable. You now know your current setup is competitive, and you have a documented process to repeat the next time you want to check.
Common Mistakes to Avoid
- Testing during a seasonal peak or trough. Comparing December results on one network against January results on another tells you nothing. Keep the test inside one comparable period.
- Forgetting ads.txt. Add the new network's required lines to your ads.txt file before launch. Missing entries can block demand and make a good network look bad. Our ads.txt and sellers.json guide covers this.
- Ignoring traffic quality signals. A new network will assess your traffic from scratch. If your site has a higher share of low-quality or bot traffic than you realized, that shows up as low eCPM. Our post on traffic quality scoring explains how networks rate inventory.
- Changing the test midway. Moving slots, adjusting floors, or swapping formats halfway through resets the clock. Set the test up, then leave it alone.
Conclusion
Testing a new ad network does not have to mean risking your income. Record a baseline, isolate a small slice of inventory, protect unsold impressions with a fallback, run the test for several full weeks, and compare RPM rather than headline eCPM. Follow those steps and you will have a confident answer about whether a change is worth making, with your existing revenue intact the entire time.
If you are ready to run that test, Squren is a straightforward network to add alongside your current setup. We support popunders, banners, IM floaters, mobile redirects, and interstitials, with bi-weekly payouts via PayPal, Paxum, Payoneer, or bank wire, and a 24/7 support team to help you get your test tags live. Join Squren as a publisher and start your comparison today.