Ad refresh: what it is and how to implement it without sinking your prices

Refreshing the ads on a page looks like free money: same reader, more impressions. The arithmetic works on paper and fails in practice, because impressions are not paid for by the unit but by what they are worth. Here is when ad refresh is legitimate, what conditions it has to meet, and when it turns expensive.

“Refreshing ads gives you more impressions”: why it tempts

The reasoning convinces because it is simple. If a reader spends four minutes on an article and the sidebar ad does not move, that slot has produced a single impression — one counted ad delivery — in four minutes. Changing it every so often would produce several, without taking up more space.

The arithmetic works on paper and fails in practice: impressions are not charged by the loose unit but according to what they are worth to whoever buys them. Refreshing badly is the fastest way to manufacture impressions worth nothing and drag down the price of the ones that were worth something.

What refreshing a slot actually means

An ad slot is a reserved space on the page where an ad is requested and rendered. Refreshing it means firing a new request for that same space while the page stays open, replacing the ad that was there with another one. The page is not reloaded and the content does not change: the auction for that rectangle is simply repeated. That is what separates it from a new page view, which is the user requesting a different document, with new context.

The difference matters for your metrics: a refresh adds impressions without adding page views, so eCPM — average revenue per thousand impressions — drops mechanically even if total money does not move.

Only what is genuinely on screen gets refreshed

Viewability is the share of served ads that had a real chance of being seen: the most widespread standard counts a display ad as viewable when at least half of its pixels have been on screen for one second. From that comes the first rule, without exceptions: only a slot inside the visible area at that moment gets refreshed, and the timer counts visible time, not elapsed time. A block the reader left three screens above must never be refreshed.

And with the user actually there

Being on screen is necessary but not sufficient. A background tab has its slots technically inside the viewport, and so does a phone left on the table. In neither case is there a person looking: refreshing there is manufacturing phantom impressions.

• The tab has to be in the foreground and focused; if the user moves to another one, the timer pauses.
• There must be recent activity: scrolling, cursor, keystrokes, taps. After a while with no signal at all, the session counts as idle and refreshing stops.
• Set a cap on refreshes per slot and session: without one, a tab forgotten for hours produces hundreds of impressions nobody has seen.

The minimum interval and why it exists

Enough time has to pass between one ad and the next in the same slot for the one being removed to have done its job: appear, be understood, and be counted as viewable by the buyer's measurement system. Before that there is no impression, only a flicker somebody paid for.

That is the criterion; the number is set by each platform in its terms, and its documentation is the first thing to read. The reference most often repeated as a floor is on the order of thirty seconds of visible time, but not all platforms ask for the same thing or measure it the same way. And heavy formats — video, expandables, anything with sound — need more room than a static banner.

What the platforms' policies say

Demand platforms do not ban refresh: they put conditions on it. The line almost all of them draw is the same: the refresh has to respond to something real — the user is still there, the ad has been visible, the minimum time has passed — and it cannot mislead the buyer about what they are buying. Breaking that does not arrive as a warning: it arrives as demand that stops bidding without explaining why and, in the worst case, as a suspended account.

• Refreshing a slot the user is not looking at, or without enough time in between, is treated as invalid traffic: it is not paid, and if it repeats, it is clawed back from what was already paid.
• Many platforms require the refresh to be signalled in the request itself, so the buyer knows they are bidding on a slot that has already shown other ads. Hiding it turns a technical mistake into an account problem.
• A refresh triggered by the user themselves — changing section, moving to the next photo, applying a filter — is the least disputed case: there is an action and a new context behind it.
• In-stream video runs on its own ad breaks; bolting a timer on top of it usually breaks the format's rules.

The hidden cost: impressions worth less

Buyers do not buy blind: they measure what share of what they buy on a domain turns out viewable and how long it stays on screen, and that history feeds their bidding models. A badly configured refresh worsens both at the same time as it multiplies impressions, and the typical result is hard to spot: far more ads served, the same money or less, a lower average CPM, and well-placed slots that start performing worse.

Doubling impressions while revenue holds steady is not a draw: it means each impression is now worth half as much.

The damage is not symmetrical either. Premium demand — the buyers who pay better and require high viewability thresholds — is the first to leave and the last to come back, and fixed-CPM campaigns are exactly the ones that fit worst with inventory full of rotations.

Where it makes sense and where it does not

Refresh depends on one underlying condition: that the user spends far longer on the same page than an ad takes to be consumed. If that does not happen, no configuration will fix it.

• It fits well on: live coverage and minute-by-minute reporting; data portals people consult for a while — sports results, quotes, flights; forums with long threads; long-form reading; and tools where people work without changing view.
• It fits badly on: fast-consumption homepages and listings; short articles; sites whose average time on page does not comfortably exceed the minimum interval; slots placed where almost nobody gets to; and formats with their own cadence, such as interstitials.

Among formats, the natural candidates are fixed bars: a sticky footer or a sticky header is on screen by definition for as long as the session lasts. And for the same reason they expose an interval that is too short straight away: if the ad flickers, the problem is not caught by a report, it is caught by the user.

How to switch it on without getting into trouble

The order of work matters more than the specific settings, because nearly every disaster comes from turning it on everywhere at once and watching only the impression counter.

• Measure average time on page per section first. If it does not comfortably exceed the minimum interval, the conversation ends there.
• Enable it on one or two slots and in a single section, never across the whole site at once.
• Compare what matters before and after: total revenue, average eCPM, viewability per slot and page view RPM, over equivalent periods. The impression counter on its own says nothing.
• Roll it back without drama if total revenue does not go up: more impressions was never the goal.

In a Header Bidding with SmartTag setup, refresh is not a switch but a rule: it decides when the auction for a slot is fired again, with its visibility, activity and interval conditions, and signalled to buyers. ADEQ configures it when the type of session justifies it, and leaves it off everywhere else.

Ad refresh is not a dirty technique or a forbidden shortcut: it is a tool with strict conditions. Where session time justifies it, it adds real revenue. As a way to inflate the counter it produces the opposite effect, and slowly: prices that slide month after month for no apparent reason.