Inspect Tool – WHY you should ABSOLUTELY use it

Inspect Tool - WHY you should ABSOLUTELY use it

Are you AS crazy about numbers and analyzing them AS I AM? Then, you 100%! need to READ THIS!

So I’ve recently encountered that GREAT Jon Loomer’s article on the subject and got fascinated about the new ways you can analyze your audience’s performance!

I’ve decided to try this analysis method on my own campaigns and discovered that there are 4 main blocks that Inspect Tool provides data on:

– All stats – main KPIs that influence ad set’s performance are presented here

– Auction competition – how competitive this pocket is at a specific day (+your CPA)

– Audience saturation – looks at the frequency and % of first-time impressions

– Auction overlap – check if you are not competing against yourself on the auction!

Basically, you can break down ANY ad set (unfortunately not campaign) and see how it has been performing HISTORICALLY and WHY it stopped working well.

Let’s review each of them based on a specific ad set:

1% LLA buyers from a CBO campaign

Was performing well at the beginning

Experienced sudden change in performance

Performance got restored later due to several changes introduced (will discuss it later)

1. All Stats

This one is particularly cool as it comprises main stats from the other 3 graphs. So, you can turn on/off different parameters to see/hide them on the graph

– For example, below, you can turn on “Amount spent” and “Purchases” to see how EFFECTIVELY you adjust your budget daily. 

Ideally, you want both graphs to grow together – you increase the budget-># of sales goes up.

If it’s not like that, you need to go down and analyze probable reasons.

Preliminary conclusion here: Feb 19, we’ve had a drop in performance – need to understand why this happened.

– Another effective comparison is “Amount spent” and “Reach” – the more you spend the more people you wanna reach. 

So, if you see that you don’t increase the budget yet your “Reach” goes down, then you have either oversaturated the pocket or your ads are not that good anymore.

2nd conclusion here: same date Feb 19, same amount spent – much less people reached – CPM has gone up hence less sales. Why? Let’s break it down further.

You can use other 2x or even 3-4x comparisons to dig deeper and get to the root of the problem (few examples – “First time impressions” and “CPA”; “Auction overlap” and “CPA”).

2. Auction competition

So we’ve got the problem – on Feb 19, our REACH went down and this caused a drop in the number of sales. Let’s see if this was due to Auction competition.

On the graph above, we see that on Feb 19, audience competition was EXTREMELY high, which obviously influenced the results. Possible reasons for that:

– We’ve increased the budget and started delivering ads to a pocket that is highly-desired by others;

– Our ads got negative feedback and started performing worse, hence poorer traffic

3rd conclusion: probably, need to update ads or scale the campaign down.

3. Audience saturation

Anyone telling you that you can have an EVERGREEN audience doesn’t know what they are doing!

You eventually tire off any audience – so you should ALWAYS try to find new winners to get substitutes ones your current winners burn down.

As you can see on the first graph and images above, however big your audience is, the quality of traffic inside is not the same – there are better and worse pockets.

Yet, Facebook always tries to reach those BETTER pockets that will most likely bring you the conversion. Thus, not all the traffic from 2 mil users will be served your ads.

That’s why, with each day, your frequency (number of time a user has seen your ad for the given period) will go UP, even if you use an audience with 2 mil users inside or NO TARGETING at all!

This is a natural process – you need to just get EVERYTHING out of that pocket, as long as it’s profitable.

But let’s get back to our case – on Feb 19, we’ve encountered a problem.

– Is it due to increased frequency? – No, the frequency slope has not changed a lot. Plus, you are good, as long as frequency on cold traffic < 2.

– What about first-time impression ratio? – YES! On Feb 19, the slope went down for the first time – this influenced the COST of our traffic.

4th conclusion: the audience reached the point where there was not enough traffic to serve ads to EFFECTIVELY (the slope went down thereafter).

How can we deal with this problem?

– Lower the ad spend

– Serve them new ads

So, basically we’ve confirmed our fears from the “Auction competition” part.

4. Auction overlap

Let’s just check our auction overlap graph. It’s indeed a very important factor – we’ve got ENOUGH competition on the market, so we DON’T want to compete against OURSELVES!

But what does the graph say about February 19th?

– There was a slight increase in the audience overlap on Feb 19

– BUT there were higher peaks during successful days before and after

– So it’s not about the overlap

5th conclusion: there is no significant auction overlap, so the main reason is either 1) poorer performance of ads or 2) too high budget for this pocket

Let’s do both – decrease the budget and update ads (on the 1st graph you can see that we’ve introduced 2 new ads on Feb 20 and reduced the daily budget)

Let’s check or results on the “normal scale”:

Feb 19 (oversaturated pocket)

Feb 20-21 (learning process restarted; budget reduced)

Feb 22-23 (new ads optimized)

Final conclusion: 

So, you can see that our CPC, CTR, and CPM have NOT improved, BUT our results HAVE!

Why? Because of the auction competition EXACTLY 

– With the same ads, with higher budget

– We could not get enough traffic from that audience to serve ads to effectively,

– With new ads, on lower budgets, we could “revive” that audience, hence lower the auction competition.

This is just a small example based on 1! ad set and 14 days of performance!
BUT when it comes to optimization, EVERY SMALL change might make a BIG difference!

Every month I choose a few entrepreneurs to personally help them scale their Facebook ads and businesses.

Watch the free training and apply here: