Home > >

---

Day 32-35: b)Blacklisting and Whitelisting - Examples (5)

---

If the previous lesson was confusing, no fear. Below is a practical example to illustrate some of the tips and more.

To keep things simple, we'll assume that all stats shown in the example are:

1)Stats for a single day (to eliminate the need to average/extrapolate), and

2)Stats from the "winner" offer+lander, to free up all 3 filtering levels in Voluum for us to drill down to various variables.

The example will be a PopAds campaign. Similar optimization strategies can be applied to any other traffic source. The important thing to keep in mind is Tip 1, i.e. we can only optimize what can be included/excluded at the traffic source.



***********************************


Drilling Down to Various Variables:

Overall Campaign Stats:

Again, let's just assume these were collected over 24 hours for simplicity.





Placement Stats:

Major Placements EDP + Minor Placements EDP = $17.64

This means that theoretically, if we don't touch anything except but placements, we can eventually hope to approach this profit level.

But let's see what else we can whitelist/blacklist to effect bigger changes to get green faster!


Brands > Models Pair:

Actually, instead of Models, for PopAds we can look at DEVICENAME which is passed back by PopAds, so we can be certain that tracker stats for this variable will correspond with targeting options at PopAds.

Sorting Brands by decreasing Profits - not much green:



If there was enough green, and if we were wanting to optimize fast by blacklisting/whitelisting in "big chunks", then we'd have the option of whitelisting the green immediately to "lock into" those profits. But that's not the case here.

So we'd need to look at the bigger Brands to see which ones are hopeful/promising that we can target and/or which ones are hopeless that we can cut, to effect significant increases in the campaign's overall profits/ROI, and then rely on cutting placements to get us the rest of the way to green.

So next: Sort Brands > DEVICENAME by decreasing Traffic.



(Make sure that in Voluum, "Tree" is selected, not "Table".)



Biggest Brands (don't waste time with smaller brands - we just want a rough estimate of profits): Samsung ROI = -19.53%; LG = -8.30% ROI; Lenovo = 11.54% ROI; Motorola = -43.72% ROI not entirely hopeless.

Let's expand DEVICENAME for these 4 biggest brands to look for promising segments. First for Samsung:



Samsung devices - bigger segments mostly either profitable or at least hopeful. ("Bigger segments" is subjective - I go by gut feeling mostly. Looking at small segments is a waste of time really. Try to avoid over-analyzing/over-optimizing).

Note: "Generic" is very negative, but can't be blacklisted in PopAds. We'd just need to whitelist other brands instead.

Motorola:



Motorola devices - bigger segments mostly hopeful. (Potentially can exclude Motorola/Moto G 3rd Gen which is very negative - won't matter very much as it's a smaller segment.)

LG:



LG devices - bigger segments mostly profitable/hopeful.

Lenovo:



Lenovo devices - bigger segments mostly profitable/hopeful.

Next: We want to see how much profits we can get from green placements, if we are to whitelist these 4 brands. Here's what we do:

1)Drill to Brands > Placements, export the stats as CSV/XLSX.

To export CSV in Voluum: Export > Download CSV.



To export CSV in Binom: Export > CSV.



To export CSV in Funnelflux Pro: Click on the spreadsheet icon at the bottom of the screen.



2)Download my spreadsheet which you can use to combine stats:

https://www.dropbox.com/scl/fi/vmye7...oosoy4ahywbt14

3)Open your stats CSV/XLSX, sort by Brands:



4)Starting with Samsung, paste the corresponding placement+revenue+cost stats into my spreadsheet > DATA SET 1:



Search for "Samsung", then click+drag to highlight the placement+revenue+cost cells in the top row, a combination of SHIFT+PGDN and SHIFT+DOWN to highlight all the way to the last row for Samsung, CTRL+C to copy, then go to my spreadsheet and CTRL+V to paste.

5)Repeat step 2 for the other 3 brands. Then click on the "Matched Data - ADDED" tab in my spreadsheet to see the "Sum of Green Placements":





So, if we were to whitelist these 4 brands, then keep cutting placements, the estimated daily profits would be $18.43. (The profits we'll actually reach in the end may be smaller, due to the fact that traffic costs are usually not very accurate when they're passed to the tracker.)

The spreadsheet also tells us that the resulting ROI would be -22.52% ROI - which would be a nice jump from the original -35%.

For these 4 Brands, we could then drill further down into Models, to identify really-negative ones to cut in order to potentially increase ROI further. Note though, that there are so many Models to target, that it would be impractical to add the sets of placement data for each and every Model together using my spreadsheet as illustrated above. So, unless the Models that need to be blacklisted jointly have a big effect on the estimated daily profits, the $18.43 should be a good estimate.

What if a lot of Models need to be blacklisted and/or they make up a big part of the traffic? In that case we could "adjust" the $18.43 by subtracting the placement revenue+cost of those models, from the overall placements data. For example when we drill down to Model for the Samsung Brand, we see:



The 2 marked segments only account for a small percentage of the 4 Brands so I wouldn't bother doing an adjustment to our $18.43 estimate. But let's assume these segments DO comprise of a larger percentage of the total traffic just so I could illustrate. We could then do the following:

1)Export placement data (drilling down directly to placements in first level) to CSV. Copy and paste data into DATA SET 1 in my spreadsheet.

2)Drill into Models > Placements, export as CSV and open. Copy placement data for Samsung Galaxy Tab 3 Lite into DATA SET 2 in my spreadsheet.

3)From the same CSV, copy placement data for Samsung Galaxy J1 2016 into DATE SET 3 in my spreadsheet.

4)Click on the "Matched Data - SUBTRACTED" tab to see the adjusted EDP.

IMPORTANT NOTE: If the adjusted EDP is smaller than the original $18.43 EDP, then it means that by cutting those models, we'd be limiting potential profits. However, it is NOT our goal to merely maximize the EDP - it's the FINAL PROFITS we're wanting to maximize - so decreasing LOSSES is another priority as well! So in this type of scenario you'd just need to weigh things and make a decision, e.g. if by cutting some segments, the EDP will shrink a little, but the overall campaign ROI will result in a significant jump, then the cutting would be worth it.

(Note that we could and should have done what we did in the "TIPS" lesson: Drill down to Device Types > Brands > Models (or for PopAds, FORMFACTORNAME > Brands > DEVICENAMES). Will do this later if the need arises.


Browsers > Browser Versions Pair:

Sorting Browsers by descending Profits -> no green segments.

Sorting Browsers by descending Visits:



Basically the only hopeful browsers are Chrome Mobile and Mobile Samsung Browser.

So we do a daily profit estimate by drilling to Browser > Placements, exporting CSV, and pasting data for these 2 browsers into my spreadsheet:



EDP = $18.70; ROI = -25.22%.

If we drill down further into Browser Versions for these 2 browsers:





(I wouldn't even drill down into the Samsung Browser - because it's already not a big segment, and its versions are even smaller.)

We see that we can blacklist some of the Chrome Mobile browser versions as marked in the first screenshot to move the campaign closer to green (again, if they had been bigger segments, we could have drilled down into Chrome Mobile > Placements to make sure EDP isn't affected).

Note: PopAds has a BROWERNAME token that you can use. PopAds has so many tracking tokens available that I chose to track other tokens instead, but it's an option you can choose.


OS > OS Versions Pair:





Vast majority of traffic is from Android, and when we drill to OS Versions for Android there's not much to cut. Not really an opportunity for optimization here.


Connection Type > Mobile Carrier Pair:

For PopAds, Connection Type > ISPNAME would be more appropriate.

For the most part, we can look at all these segments as just 2 segments - "Mobile" which is carrier traffic, and all the rest being mostly Wifi traffic.





Vastly majority of traffic is wifi, and when we drill down to ISPNAME there's not really anything to cut. So again, not really any opportunity for optimization here.


A word about Connection speeds targeting at PopAds:

The "Connection type" variables tracked by Voluum and the "Connect speeds" tracked by Popads are somewhat different. I thought I would provide a cross-reference chart in case you need it.








Device types:



Vastly majority of traffic is from mobile phone, but the rest can be cut. Not worth doing an EDP based on green placements for just Mobile phone - because it would be similar to the $17.64 EDP done just for placements as first level drill-down (calculated at the beginning of this example).

As for ROI, we can see that it's -30% for the Mobile phone segment.

Again, since this is a PopAds camp, I could have used FORMFACTORNAME data, but Voluum's categories are similar to PopAds' for this variable so I used the token slot to track another token instead.


QUALITY:



Looks like it may be good to target Quality 2+ websites:



Just to make sure, we can drill down into Quality 1 > Placement stats, and sort by decreasing profits to see green placements:



There ARE a couple of green placements that would get cut by excluding Quality 1 sites, to decrease the placement EDP to $17.64 - $2.25 = $15.39. But the cut would result in quite a jump in ROI by saving us close to $20/day in losses. Well worth cutting I'd say!


CATEGORYID:



Can exclude the 2 categories marked in the screenshot.

Can even drill down further into placement stats. CATEGORY ID > Placements makes another good variable pair - as far as I can tell, each placement belongs to one category ID only.

When drilling down into the 2 bad categories, there are almost no green placements, so the cutting shouldn't affect the placement EDP, although that would get us to green faster by effecting a jump in ROI by saving us close to $17/day in losses.



*****

Creating a Plan of Attack:

So we've identified some optimization opportunities and done some calculations.

Next: How exactly should we implement the optimization?

Let's first summarize our findings above:

We COULD just whitelist/blacklist all of these segments as listed.

However, know that we've only been looking at the effects of EACH variable or variable pair, and the effects of whitelisting/blacklisting on placement stats.

But we're not sure what the ACCUMULATIVE effects of applying ALL these optimizations would be.

For example, we know that by targeting Chrome Mobile and Mobile Samsung Browser only, we'd be able to push the campaign's profits and ROI in the right direction. But how do we know that when we ALSO restrict targeting to the 4 Brands listed above, that we won't hurt profits/ROI? Remember that many variables are related, so that whitelisting/blacklisting segments in one variable can affect another variable.

What we can do is to pick a variable that we feel can give us the best optimization, and THEN check other variables against it to make sure they don't have negative effects on that variable.

Looking at our example campaign, it seems that optimizing by Brands will give us the best EDP+ROI combination. So let's start there.

Next we decide whether ALSO whitelisting the 2 browsers would be a good idea. Checking the effects of whitelisting the 2 Browsers in addition to the 4 Brands:



Whitelisting Chrome Mobile browser and Mobile Samsung Browser is good for the Samsung Brand - those are the only 2 promising segments.



For Motorola, Chrome Mobile is the only hopeful segment - so whitelisting the 2 browsers can't hurt.

We would do the same for LG and Lenovo - I did and saw similar results - that whitelisting the 2 browsers would only increase the ROI and not hurt it.

Next, we check whether it would be a good idea to whitelist Mobile Phones, against the 4 Brands:



Only Samsung has any significant Tablet traffic, and its ROI is actually not hopeless! So we could actually reconsider about just whitelisting smart phones. We may decide to include tablets as well, to see if cutting more placements would make that green.

Moving onto Quality - what would the effects of cutting Quality 1 sites have on the 4 Brands?









So, excluding Quality 1 sites (rather, targeting 2+ sites) will have a positive effect on all 4 Brands, because the Quality 1 subsegments are all in loss. However, notice that the ROIs in all cases are not hopeless, so we may decide NOT to exclude Quality 1 sites, but instead rely on cutting placements to see if we could get to green.

Lastly - checking the exclusion of the 2 Categories against the 4 Brands:



Hopeless.



Not hopeless, but may be difficult.

LG > CategoryID and Lenovo > CategoryID result in very small segments - not worth checking.

Overall effects of cutting the 2 categories are more good than bad.

Therefore, the final optimization we decide to implement, may look like this:



***************




Amy

---

Thankyou @vortex for update.
Eagerly waiting for this 2021 version to complete so that I can re-read this thread without interruptions.

---

Noted! Only a couple more posts to go - will wrap this up in a week or two!

Thank you for your support!


Amy


Sent from my iPhone using STM Forums

---

Hi Amy,
Could you please share the latest link to the Excel Sheet used for the calculations. I checked all the dropbox links given in this article and the suggested ones but none of them has the 'Matched Data - ADDED' tab, 'Sum of Green Placements' and 'ROI%' fields as you have mentioned in the article.

Thanks

---

Apologies @tanmay_b! I've pasted the wrong link into the original post - don't know how that happened. Here's the correct link:

https://www.dropbox.com/scl/fi/vmye7...oosoy4ahywbt14

Again - sorry for the mixup! And thanks so much for asking!



Amy

---

Home > >