Field notes
The site where display ads were never going to work
One of our sites is a news site for a diaspora community: short news items by design, averaging about 170 words each. This is the note on why we stopped trying to put display ads on it, and what we run there instead.
By Mario Bailey · Published
The instinct when a site will not monetize is to work harder at monetizing it: more traffic, better placements, another network. This site taught us to ask a prior question. Some sites are not underperforming at the model. They are on the wrong model, and no amount of execution fixes that.
Format sets the ceiling before optimization begins
A 170-word news item does one job well: it tells a busy reader what happened, fast. That same virtue is the monetization problem. Short-form news means short dwell time and one or two viewports per visit, which is thin per-page ad value no matter where a unit sits. And at that length, the pages would read as thin content in any approval review, an assessment we would honestly agree with if the pages were pretending to be articles. They are not. They are news briefs, and briefs are the right format for this audience.
The lesson we took: the content format sets the monetization ceiling before the first optimization is attempted. Placement work, network choice, and layout tuning all operate under that ceiling. They cannot raise it. If the format only earns a few seconds of attention per page, display ads are priced accordingly, and the honest response is not to fight the pricing. It is to question the model.
The tell that it is a model problem, not an execution problem
This site was not failing at search. In a recent window it averaged around position 12. Traffic was never the problem, which is exactly what made the diagnosis useful: when a site gets found and still cannot pay for itself, execution is not the suspect.
The tell is that your best case still fails. Run the math with every execution variable set to generous: perfect placements, a strong network, traffic doubled. If the number still does not justify the work, you have a model problem, and iterating toward that best case is waste. Execution problems shrink when you optimize. Model problems do not move. The secondary tell is behavioral: if every fix you reach for is "more of the same" (more posts, more pageviews, more units), you are optimizing a ceiling instead of questioning it.
Changing the model without burning the site down
We did not delete the news or force the site to become something its readers never asked for. The news is the audience engine: it is why people come, and it earns the visit. What changed is what the visit is for. The moves, in order:
- Keep publishing news exactly as before. The format was only wrong as the money page. It is right as the front door.
- Add depth where the value lives. Evergreen guides on the questions this audience pays to solve (immigration, money, business), with every news page routing readers to the guides relevant to its topic.
- Put the money at the end of the funnel, not on the pageview. A directory of professionals (attorneys, accountants, insurance) with a simple contact form. For this audience, a single qualified lead is worth far more than display ads on the same visit.
- Prune what only existed for the old model. Thin tag pages were de-indexed until they have enough content behind them, which cut the sitemap from 1,573 URLs to 600 while every page stayed live for readers.
- Write the decision down. Our notes say, in so many words, do not re-attempt display approval on this site. A model change only sticks if the old model is formally retired. Otherwise every slow month re-litigates it.
None of this was dramatic, and that is the point. The site kept publishing through the whole change. The calm version of a pivot is additive: you build the new model alongside the old one, move the money's center of gravity, and retire the old expectation on paper rather than the site in production.
Pivot or persist: the framework The net-positive test
Method: the figures on this page come from our own records and Search Console data, rounded as written. The sites are ours, and we deliberately do not name them. Field notes describe what happened on our network, not a promise of what will happen on yours.