EPC & EPM On TD

[Lee_Owen]

Tweet

Could someone explain what these mean on the TD site please.

On some programs the EPM is nought but on others there's a value, now I take this to mean, that per 1000 possible impressions of an ad the average network wide is the value stated for that graphic. So why on one merchant would that be nought throughout? yet and moving on to EPC, it shows an amount not calcuable to the amount of 100 x the individual lead?

Which makes me ponder EPC.

So with EPC, when it's per 100 clicks, how do they get a value which is wholly greater than the actual revenues possible. Is it more a case of that the Average Paid EPC, has nothing to do with 100 clicks and rather the average monthly revenue for that program network wide? 'cause it can't be reflective of 100 clicks.

Plus what exacty is 90th EPC? I ask these questions because hopefully I can spend several hours going through all the programs seeing what works well and what doesn't and concentrate my efforts and getting a piece of the pie.

Examples:

Look at Samsung EPC and EPM, then JackpotJoy EPC and EPM - more confusing because it's a shared rvenue program as well. Then look at Jamster EPM and EPC. It's got me confuddled. If EPC doesn't mean EPC, why do they call it EPC when I always thought it was reflective of 100 clicks.

I can be dopey....

[drivetowin]

If you're dopey then I reckon I'm sleepy (and I reckon Renegade would make a good choice for grumpy - eh Joe!? )

Good questions, would be good if Nick from TD could explain what these mean - I know Andy tried to explain the 90th thing earlier but an actual worked example would help me understand it.

[Rich]

Hmm, well spotted.
Some example figures to stop others having to look...
Screen Select pay £10/click. Average paid EPC £12354.04 or £123.54/click.

I expect it will be something to do with TD not counting all clicks. I've got quite a few programs generating sales that never report any clicks. If a big PPC users clicks weren't being counted it could cause the results.

As for 90th EPC, this was the original response

The 90th percentile is the smallest number that is greater than 90 of the numbers in a given set. The smallest EPC of the top 10%

Assuming they mean top 10% of EPCs, rather than top 10% of earners, then 10% of affiliates have an EPC equal to or higher than the figure show, 90% have a lower one - so you're doing well if you get anywhere near that figure.

Taking Screen Select again, they have a 90th EPC of 246.55. So you know that the top 10% convertors a doing a sale for every 4 clicks or better. But it doesn't tell me much about what performance I could expect - It would be more useful to me to show the commission rates on the search results screens than the EPCs

[everydayholiday]

One possibility is they just started calculating the EPC yesterday so the stats dont mean much yet. Also on individual basis you can get very high EPC , more than your revenue. e.g. if you get 10 clicks and one sale with a £8 commission your earning per 100 clicks are £80 even though you never earned more than £8.

Another possibility is they implemented the system of TD France over the whole network, where all sales are counted and a small fraction of the clicks only, so you always get enormous conversion rates and EPC

Stef

[TD Nick]

Hi Everyone...

The performance metrics are indicators of how well a program converts traffic to commission.

EPC: Earnings per 100 click – displayed when searching for programs
EPM: Earnings per 1000/ impressions – displayed when searching for graphical elements

The EPC values are displayed times 100. The EPM metric is only calculated if there are more than 100 000 impressions registered to a graphical element.

All metrics are based on data from the previous calendar month.

How are the different metrics calculated?

The average is calculated from the publisher base that got paid for that program. The 90th percentile is calculated from all publishers in that program.

The average is the sum of the earnings divided with the unique visitors, whereas the percentile is a rank, i.e. the actual value for the publisher that ranks 90th out of 100.

The values are calculated for 100 clicks and 1000 impressions.

What influences a program’s performance metric?
The main influences are the conversion, the tariff structure and the tariff levels. Other influences are traffic volumes in the program, what sectors and also in some cases what type of publisher that they are working with.

How should they be interpreted?

• They are calculated from the basis of the entire network and should not be taken literally for a specific publisher.

• The relative performance of programs makes sense the more similar programs are to one another. Nor is it likely that you will reach a high EPC if your target group is not in line with that of the program.

• What if the average is higher than the 90th percentile? Then it is reasonable to assume that the top 10% of the sites are earning substantial amounts of the total program earnings.

• What if the 90th percentile is higher than the average? Then it is reasonable to assume that the top 10% of sites that earn from the program play a proportionally lesser part of the total.

• Most of all it is important to understand that the values are only indicative!

• If the paid average is at a reasonable level and the 90th substantially larger than the paid average, it is reasonable to expect that (if the program category applies to your target group) it is would not be too difficult to earn well from the program

• How should zero values be interpreted? If the 90th is zero but the paid average has a value, this can mean that the program has many sites and therefore the 90th falls short of being an “active site”. It could also mean that relatively few sites generate revenue within the program.

• If the paid average has a zero value, but the 90th has a larger than zero value. Paid average means that no publisher has been paid for the program, so this means that it’s probably a new program, a poor converter or that there is something wrong with the tariff structure. If the 90th still has a value it means that some sites still generate revenue within the program, but that they are new to the program.

• EPM and graphical element search; some graphical elements display extreme value. These can in most circumstances be derived from the fact that there are relatively few clicks but some significant sales attributed to the graphical element in question. Even more than for programs, do not take these values as facts. But evaluate them as one piece of information among other.

• EPM and graphical element search; as there are fewer users per graphical elements there tend to be more “0” values on the 90th percentile calculation, than is the case for programs.


Why do we publish it?
We believe in transparency as a company. Transparency will minimize the “waste” of traffic within the network and allow for everybody to compete on a level playing field,.

Why it is not published for all programs?
Some programs are new to the network and therefore we don’t have enough data to support a calculation. Some advertisers cannot give out information that can be considered financial.

I hope the above information has helped.

Many Thanks

Andy Andreou

[everydayholiday]

Hi Andy,

this makes sense in theory, but there s something wrong with the figures. I dont think on average affiliates really make £2400 per 100 clicks on PO Ferries Belgium just to name one example.

Stef

[Lee_Owen]

Hi Andy,

this makes sense in theory, but there s something wrong with the figures. I dont think on average affiliates really make £2400 per 100 clicks on PO Ferries Belgium just to name one example.

Stef

Thanks for the info update, I'll try and get my head around it later and see if it's handy or not. And EH, some EPCs won't make sense due to a lot of clicks not being counted on some programs, so average EPC may not be any help and 90th EPC could be helpful but as TD say it's far from accurate and on the dot and should be a sign post rather than a place to drive.

I'll have a look at the theories for sorting stats later on, cup of coffee time.