When managing your own Prebid wrapper, optimising Timeouts can be a delicate subject because they depend on various settings.
The timeout setup on Prebid allows more time for bidders to respond to auctions and therefore optimises timeout rates. However, this is certainly not the solution to the problem.
Increasing the timeout setup is beneficial for bidders, since they have more time to bid. But it can slow down your website pages and decrease user experience. It also causes ads to be served later on the page, so it can lower ad viewability and in the end the quality of your inventory.
In this article, we will explain how to optimise bidders’ timeouts in order to increase their advertising performance, and achieve the balance between high yield and optimal user experience.
1) What are Prebid timeouts?
When a Prebid auction takes place for a user, on a specific placement and a specific web page, several header bidders will be queried through bid requests (most often one bid request per bidder).
As most often several header bidders are queried, you must define a delay in your Prebid setup which is called the timeout setup (most often comprised between 500ms and 4000ms) during which all bidders will have to submit their bids. After this period of time, we consider the auction as terminated, whether the bidders have answered or not.
When you send a bid request to a header bidder, it can only lead to 3 possibilities:
- The header bidder submits a bid response, i.e. he has a potential buyer for the site
- The header bidder indicates that he does not wish to participate in the auction, he submits a “no bid”.
- The header bidder did not have time to answer within the time limit. The header bidder is in timeout;
So out of 100 bid requests sent to a header bidder, if the latter responds 30% of the time with a bid, 50% of the time he replies by saying that he has no bid, and that 20% of the time he does not respond at all within the time limit, then we say that the bidder has a 20% timeout rate.
If we look at the ratio, it gives the following figures:
|Header Bidder Values||Values|
|Bid rate (bid responses / bid requests)||50%|
|No bid rate (no-bid responses/ bid requests)||30%|
|Timeout rate (timeouts / bid requests)||20%|
Note that: Bid rate + No-Bid rate + Timeout rate always make 100%, since a bid request necessarily leads to one of the three scenarios.
2) What are their impacts on advertising revenues?
The impact of timeouts in terms of revenue is simple to measure.
If a bidder has a 20% timeout rate, this simply means that 20% of the time it didn’t have enough time to submit its answer when a bid request was sent, whether it is a bid response or a no-bid response.
So, a timeout of 20% reduces a bidder’s bid rate by 20%, i.e. statistically you lose 20% of your bid responses.
Using the example above, in the case where our timeout rate is equal to 0% this would give:
|Header Bidder||Scenario 20% timeout||Scenario 0% timeout|
|Bid rate||50% (50/100)||62.5% (62.5/100)|
|No bid rate||30% (30/100)||37,5% (37,5/100)|
|Timeout rate||20% (20/100)||0% (0/100)|
So, we gain +12.5 bid rate points (which is a very good uplift).
3) Where do timeouts come from?
Timeouts can come from different reasons:
- A bidder with a weak infrastructure in Mexico, for example, will have longer response times for Mexican trafﬁc. So he will have a harder time responding within the time limit than another bidder,
- Bad CMP configuration can generate Timeouts,
- An ad unit connected to too many bidders can slow down your page and prevent bidders from responding within the time limit,
- Bad network quality (on mobile for example),
- Bad configuration for a particular bidder can cause other bidders to time-out.
All in all, there is no miracle solution to troubleshoot timeouts, you have to measure, test new setups and iterate until you reach the required level of optimization.
4) How to measure timeouts?
As we have seen above, Timeouts can come from different reasons.
To be able to find their root-cause, measuring timeouts by bidder only is usually not enough. It is necessary to be able to cross-check this data by device, by country, by ad unit, by site, etc., in order to clearly identify the exact origin of these.
Here is a simple example where the overall timeout rate of a bidder is high, but also extremely different from one device to another.
The immediate decision may be to increase the timeout setup only on mobile, since the timeout rate on this device is 2.5x higher.
When it comes to timeouts measurement, using a powerful Prebid analytics connector is key, as you need to keep in mind that the more data granularity you have, the easier it will be to optimize your timeout rates.
5) How to optimize them?
Once you have set-up a proper Timeout measurement system, a 4-step process can be applied to quickly troubleshoot timeouts:
- Identify the contexts where timeouts are highest, i.e, specific devices, sites, ad units, bidders, site sections, etc.,
- Make a hypothesis on one of the contexts,
- Test a new setup,
- Observe the impact of the change right on the next hour.
Timeouts are a vast subject among publishers. They must be constantly monitored because many factors can impact them: changes on your website, connecting a new bidder, changes in your CMP, changes of Prebid version, etc.
To optimise them there is no magic formula. But you can setup a process of continuous measurement and a test & learn approach to guarantee optimal performance on the subject.