Home > The Newbie Zone > Questions and Answers

---

What to check when CTR drop significantly (13)

---

I have an offer which converts quite well in the first day but then it's CTR and CR drop significantly in the following days as can be seen in the figure below

This offer running with CPA bid model


The same offer running with CPM bid model


What are reasons for this shaft drop of CTR ? Any technical issue I should check for with my lander ?

---

What's the traffic type? Which network? What vertical?



Amy

---

@vortex
I'm running a sweepstake IVR offer of Traffic Company network. POP traffic. I'm using my own lander

---

I would say that's pretty typical for pop in general. Performance tends to be better initially - I suspect traffic quality is just better in beginning in general. And then performance would drop and plateau into "true" or average performance. Then gradually drop as saturation sets in (ad fatigue + offer lead database saturation and maybe other factors).

Sometimes by setting up new campaigns that are exact duplicates of the original campaign will again get you that initial burst of good quality traffic - but really, without automation, it can turn into a lot of work.

And to combat the gradual decrease in performance, optimize the campaign - by cutting placements etc.

And then of course there's the usual volatility in performance that is also very typical of pop traffic - due to variance in competition levels (number of advertisers/campaigns, changes in bids), visitor habits, etc. What you can do is pause the campaign on days when performance is bad, resume the next day and repeat (i.e. pause if unprofitable by mid-morning or noon, leave running if profitable).

I haven't run Traffic Company's IVR offers in a while. @twinaxe is running them as we speak - I bet he could give more specific advice.



Amy

---

@vortex: What is exactly good quality traffic ? Which one of the two below scenarios are correct
- 1st: is this equivalent to high converting zones. Does this mean that at the beginning, the traffic network will open up some very high converting zones to give us leads. Then in the following days it will stop sending traffic from those particular zones and switch to average zones ?
- 2nd: or is this equivalent to high converting users in each zones. That means we still see the same set of zones everyday, but at the first day traffic will come from high converting users in each zone. Is traffic networks capable of pinpoint to this high level of granularity., differentiating the high converting viewers and lower ones
- 3rd: or any other scenarios I haven't thought of

---

Also don´t only check the daily performance or so.

Go deeper, check the placements that send traffic.

For CPA campaigns it´s normal to receive best traffic in the beginning, when the campaign converts you will also receiver less converting traffic so that the algorithm can check what´s converting good at what bid.

Then aftewr few days when the campaign keeps converting the performance should become more stable.

With these offers the performance can change pretty fast so what converts extremely good today can be a total fail 2 days later.

When you find a good offer/LP/geo/carrier combination run it as long as it lasts and when it stops converting stop it.

Doesn´t have necessarily be an issue of your campaign, can also be that there are just no good numbers for that combination available anymore.

There are geos/carrier for IVR that are more stable than others but there you can still go from 100% ROI to -100% ROI from one day to the other and then from -100% ROI back to 100% ROI 2 days later.

---

I've seen both.

Compare your placements stats before and after. Look for both cases, i.e. any differences in the distribution of traffic from various placements, and also the CTR+CR of individual placements.

For the 2nd scenario you laid out: Yes some networks are able to track high-converting users (as some advertisers post conversions back to the traffic source) and/or users that tend to click more. The other factor is browser session depth. A user can be subject to multiple popups while browsing a site, and your ad's performance can vary depending on how early/late your ad is shown to that user - and on most platforms this isn't something we can control, or even shown stats for.

It can be difficult to come to any definitely conclusions because there can be many placements, and the "initial burst of good traffic" period is usually not long enough to give enough data. In other words, aside from the biggest few placements, most other placements may have no conversions or too few conversions - not enough to show "true" performance.



Amy

---

@twinaxe: What do you mean when you say "there are just no good numbers'?

---

@twinaxe: When running CPA campaign especially in first few days when it's still learning, do you temporarily stop the camp when the perform drops significantly similar to above example and restart the camp next day. I often do this when running CPM camp to avoid the bad day but I wonder if stopping cpa camp temporarily during the first few days affect it's learning curve.

---

This is where you need to perform split testing. First, create a similar campaign and check whether its indicators will be better than those of your current one. If they're worse - start with the creatives, then change the headlines and the creative itself

---

When running something on the CPA model, I don't touch it during the learning phase as it would "confuse" the algo. It definitely depends on a particular network and how their learning phase is configured, but in some cases it might reset the whole test or even cause the algo to stop sending traffic to such campaign completely.

But to be fair, I have seen cases where it helped actually. I believe that particular network simply had a certain amount of impressions reserved for the test, so when I paused the campaign (because it performed poorly) and then restarted it, the algo just continued to display the ads until it reached the limit. And most likely, the second part of impressions were of higher quality so the performance of the test improved and ended up with a better result.

So once again, you will need to test it with the network you're using to see what happens in your case.

---

Thanks @matuloo for this valuable advice. I go ahead to check with my network (which is Propeller Ads in this case) and they say their algorithm will not mess up if the stop period is less than 3 days.

---

Ok, if that's the case, it should be safe to pause the camp for a bit and it might even help just as I mentioned in the reply you quoted. Let us know how it went

---

Home > The Newbie Zone > Questions and Answers