Ad-servers - Part II

In this article we’ll take a close look at how ad-servers actually serve the ads that the user sees.

The first step, our imaginary company Weightloss (WL) has to place its ads into its ad-server. The ad-sever then creates tags. The tags give the following details –

Ad size

Where the ad can be located on the web page

The tags are sent to the publisher alongwith information on WL’s campaign like –

Number of times a day it has to appear also called frequency cap. An ad can be 1/24 frequency capped, meaning it appears once in 24 hours on a particular PC, or 2/24 meaning it appears twice in 24 hours etc

Demographics of the targeting i.e. should the ad be targeted to men/women of a certain age group, who have specific interests ( interested in golf, gardening..)

Geography to which it is targeted – Campaigns could be targeted to the UK, US, Germany….

Specific time during the day when the ad should appear

All of WL’s campaign instructions are now loaded onto our imaginary newspaper the England Journal ‘s(EJ) ad-server. The ads themselves reside on WL’s ad-servers. The EJ similarly receives tags from other advertisers as well.

What happens next?

Suppose a user enters the website URL of The England Journal onto their PC, the EJ website is displayed on his/her PC. What is loaded on the user’s PC is content… A logical question that arises is – How does an ad get loaded onto a user’s browser? The user’s browser follows instructions it receives from The EJ, which in turn directs to get an ad from the EJ ad-server.

The EJ server in turn instructs the WL ad-server to serve an ad. The WL ad-server then has to figure out which ad to show. A cookie helps the WL ad-server in this process. When the EJ ad-server sent the user’s PC content, it also sent it a cookie. A cookie is a small piece of software that travels between the EJ ad-server and the WL ad-server. A cookie identifies a browser. A cookie controls the number of times the user sees an ad, has user information like age, sex , preferences etc.

The EJ ad-server selects the ad and serves it on the user’s PC. The ad selected is based on user information provided by the cookie. When WL’s ad-server sends an ad to the User’s PC, it also sends it a cookie which allows it to identify the user’s browser. This allows WL’s ad-server to track number of times an ad is shown on the user’s PC.

The process of serving an ad, though happens in a few seconds, is a fairly involved process. This process is completely managed by the different ad-servers in the network without any human intervention whatsoever. Amazing isn’t it?

Ad-serving Part 2

As a continuation of my earlier article on basics of ad-serving, I will be dwelling a bit in depth on ad-serving and how it works.

As a first step let’s look at how ad-servers target ads that the user sees on his PC. In a typical online ad display, there necessarily are two ad-servers ( if not more)- the publisher ad-server and the advertiser ad-server.

The publisher ad-sever decides which advertiser’s ad to show. The advertiser ad-server decide which ad to show from among it’s stock of ads.

Going back to our imaginary newspaper of yesterday, the England Journal (EJ). How does the EJ’s ad-server decide which ads a user Roger is going to see? The ads Roger’s browser has show and the ads he has clicked on play a role in which ads Roger sees. Roger’s demographic information – age, sex etc – will also play a role in what he sees. To understand how this can happen, let’s assume that Roger has registered himself on another website with him name, age, sex, kind of work he does etc… Many a publisher shares this demographic information with other ad-servers so that they can demographically target ads.

For example, therefore, a website (publisher) might tell an ad-server that Roger is a male between 35 and 40. And that fact that Roger often visits home and decor websites identifies him as a person who is interested in garden, furnishing etc. Roger might therefore, find that a number of home decor ads start to appear on his website. Roger’s computer IP ( Internet Protocol ) is used by ad-servers as well. An IP address helps identify around what area a PC is located.

Is ad-serving an intrusion into one’s privacy? Do they have your personal details like name sex, email id etc? Not quite. Ad-severs target primarily to an IP address – they do not have any personal details like name, phone number emai lid etc. They primarily target people’s interests, demographics and where people are located.

What is Ad-serving technology and how does it work?

Have you ever wondered how ad-serving technology works? Have you wondered how an ad is served? And how it is tailored to make it as relevant to users as possible?

Why do many companies use 3^rd party servers?

Let me run through some of these aspects by using the example of a non-existent newspaper – The England Journal (EJ). Unlike the ad on the actual newspaper where the ad remains constant, the ad on the online version of the newspaper keeps changing – first it is an ad for the Marriot chain of hotels and then you have an ad for Britain as a tourist attraction.

Which ad appears on a page is determined by a number of factors, like who is the advertiser and what is the content on the page. Is there a clash between the content on the page and ad ? Is there a synergy?

Suppose the a weightloss company – Weightloss UK- wants to advertise on the EJ. The EJ will have to build ad-serving technology to determine when it display Weightloos’ ads and to whom and Weightloss will have to determine which ads were show and when they were shown.

That’s why publishers like the EJ and advertisers outsource this to 3^rd party companies call Ad-serving companies. These ad-serving companies serve ads on their behalf. Outsourcing ad-serving allows publishers to focus on their core competencies – stick to their knitting as it were. Advertisers are also allowed to stick to their core competencies- writing ads that hit their audience between their eyes as it were.

Ad-serving technology serves ads on their behalf. It also helps the advertiser and publisher alike to better manage their campaigns.

It helps the publisher helps see and measure performance of different advertisers on their website. It helps the publisher manage his inventory and also generate reports as and when required. The advertiser, on the other hand, can track performance of ads on different sites, allowing him to generate reports as well.

In the second part of this report I’ll dwell on popular ad-serving technologies available in the marketplace.

Semantic Targeting

In an earlier article I had dwelt on types of targeting –behavioural, contextual, re-targeting etc.

Targeting as an area in online advertising is continuously evolving. Reasons are obvious. A campaign is as effective as it is sharply targeted. If a method of targeting is evolved which guarantees reach to a specific individual, one can imagine how popular it would be!

A step forward in the targeting evolution is ‘semantic’ targeting(ST). ST is a much more refined form of contextual targeting. In contextual targeting ads are served in context to what the visitor to a website is doing. Typically, a contextual system looks out for keywords and display ads to the webpage that the user is visiting. Contextual advertising , therefore, is predicated around keywords.

Behavioural and contextual advertising, though popular have their own pitfalls.

Behavioural targeting is facilitated by placing a cookie on the users PC. The cookie is the spy that tracks the users viewing behaviour. It clearly intrudes on the user’s privacy Consumer groups and law makers have expressed these concerns already.

Contextual on the other hand goes entirely by ‘words’. But contextual targeting goes by the word itself, not its meaning which could lead to misplacement of ads. For example the word orange could be a fruit, could the mobile services provider, could the colour ….

ST goes a step further. The Merriam Webster dictionary defines Semantics as ‘the study of meanings’.

ST looks at all the words and gets an overall meaning. Since most words have multiple meanings, without understanding the context of why the words occur , it is quite possible that an ad is wrongly assigned where the context doesn’t exist.

ST also identifies the sentiment of a webpage by analyzing the language on the page. If the page is not positive about a particular topic, the ST system may not allow the ad to displayed on the page.

ST also protects the value and integrity of the brand. It would, therefore, prevent the ad from occurring on sites like Adult/Erotica, Nudity, Alcoholism … thereby preventing potential damage to the brand.

ST is the ‘new kid on the block’. It is evolving, and with time it would become increasingly effective allowing sharper campaign focus.

Wanted - An advertiser - network Partnership

In my earlier articles, I have dwelt extensively on campaign performance and what needs to be done to maintain / increase campaign performance.

Any online campaign should ideally work in a partnership between advertiser and network/publisher. A relationship based on the principle ‘I’ve paid you the money, you deliver’ will get you that far but will fail in due course.

Campaigns succeed or fail for the following reasons –

Creative Latency

Website latency

Lack of interest in the product/service

Poor creative

Wrongly positioned creative

Inappropriate creative

Recently, I had a call from a client who wishes to run a campaign based on pop-unders only. She explained to me that they had tried different creatives and they found pop-unders to be most successful. I had a very clear brief, therefore, to run a pop-unders based campaign only. Brilliant! I was impressed at the clarity on part of the client. I felt like I was entering into a relationship that would be man to man – in terms of the way discussions would proceed post campaign kick-off.

An important element here is for the advertiser to understand display advertising in detail, in terms of factors listed above. If the customer is focused only on ROIs and nothing else, the advertiser/network comes under enormous pressure because on the one hand he has to keep the customer happy and on the other hand he/she has to contend with issues that go with an online initiative.

Such understanding on part of the customer should not only be on account of the technical aspects of a campaign but the ‘commercial’ aspects as well. For example, if a client comes to me asking for IASH inventory (UK) for US $ 0.25, he is asking for the moon! IASH inventory is simply not available at those rates. Likewise, using ROI as an excuse, if a customer offers ridiculously low CPM rates, the campaign is a failure ab initio. It is like accelerating the car with the hand-brake on!

It is not as if all customers lack empathy however. I was running a campaign for a client where CTRs were fairly healthy, till about 2 weeks ago, when they dropped off suddenly. The client did initially throw a fit on declining CTRs, stopped the campaign and asked for a refund of unutilized funds. But after a detailed analysis, I explained to them that with the football world cup this was par for the course! I suggested that we pause the campaign, allow the world cup fever to die down after which we can re-start the campaign. My customer saw logic in my argument and readily agreed. What an understanding customer!

May their tribe increase!

Future online Advertising trends

Today, I am going to gaze into the crystal ball and forecast what I feel are likely to be online advertising trends over the next year or so. Having been in this business for a while, I have had many a moment of cogitation when I have gazed into the future. Here are a few thoughts –

Mobile advertising

Mobile advertising will definitely show a clear and positive growth. Biggies like Google acquiring the likes of AdMob are a clear shape of things to come. The iPhone, Blackberry, Android etc are devices that lend themselves to mobile advertising

Focus on ‘niches’

My suspicion is that advertisers are move to small ‘niche’ publishers. Niche publishers will have their own captive audience. A relevant campaign on such niche websites will definitely get many ‘eyeballs’.

Social Media Advertising

As mentioned in of my earlier articles, social media advertising is definitely going to be increasingly relevant going forward. With its inherent informality, it is going to be a more used medium to get a message across.

Advertising Exchanges

Though looked on with suspicion by the advertiser and network owner alike, it is an ‘all to gain nothing to lose’ medium.

Remnant Inventory

Remnant Inventory is ‘remnant’ because of its price point. Publishers are getting increasingly canny. My guess is that the publisher community will start demanding more for their un-sold inventory.

Measuring Campaign performance

Another clear trend I see is better campaign performance metrics emerging. The advertiser and publisher would equally like to better measure impact of campaigns.

These are a few trends I could thing ok. Would love to hear from you on any other trends you foresee.

Relative merits of RON, ROC and ROS targeting

There are different types of targeting that we can adopt in online campaigns –

Run of Network ( RON) where you run the campaign on the entire network

Run of Channel ) ROC) – where the campaign is run in a specific channel , examples being, healthcare, entertainment, technology etc.

Run of Site ( ROS) – where the campaign appears on specific sites only…

It is obviously that costs escalate as you move from RON to ROC to ROS… RON is cheapest because it is across the entire network and to that extent the efficacy of the campaign is dilute

ROC is dearer because it involves running the campaign on websites within a specific channel only

ROS is the costliest because of the fact that ad impressions appear on specific sites only

As an advertiser, I’d have a question – which is the best method of targeting I need to employ? Let’s try to find some answers.

RON – is employed when a product/service being advertised has general relevance across a community of users. And the over-riding objective of the campaign is ‘brand building’ … For example, advertising a new fizzy drink from Pepsi… Could be relevant across a channels. An ROC/ROS campaign in such a scenario simply doesn’t make sense

ROC – is employed when the product service being advertised is relevant only to a specific community. For example, launch of lab testing equipment. Such a campaign would be effective within the Science and technology channel. There is simply no point advertising such product RON as many impressions would be ‘burnt’ inventory

ROS – is employed when you wish to target users of a specific property. For example, I am aware of www.btymplace.com being a hotly marketing property. This property was marketing on its visitor profile –

250k Westminster residents

800k daily Westminster commuters

2 million monthly leisure visitors

500k business travellers

Anyone who was looking to reach out to this demographic would, therefore, find btmyplace an ideal inventory to place their banners on. But such a campaign needs to be sharply focused in terms of creatives and ROIs calculated to a high degree of accuracy. Simply because ROS rates can be anywhere from US $ 10 CPM upwards…

Link Building

As the owner of a website, one is always looking at ways and means of getting website ratings up. Among the various ways of achieving this is ‘link building(LB)’.

Link Building is about creating inbound links to one’s own websites. This can be done by having links on a reciprocal basis i.e. website ‘A’ links to website ’B’ and vice-versa. LB is considered one of the most effective ways of increasing website popularity. The number of sites which link to a particular website is known as link popularity. This helps improve the search engine ranking of a website.

Benefits of Link Building

Increase traffic – By its very nature, LB increases traffic to a website. However clever linking building i.e links with relevant websites has been found to increase sales. For example, the website of an automotive could be linked to the website of an auto racing event

Rating of Website – When links are chosen carefully, this results in high quality incoming links which will ensure the website’s rankings improve considerably. Again , improvement of website rakings will ensure it being thrown up on the first page of a search, if not the occupying the topmost position!

Raising Profile – LB raises the profile of a website i.e. it increases visibility, awareness and credibility of a website. In the example quoted earlier, many visitors to the auto website will link thru’ to the auto racing site. This would raise the profile of this site and very rapidly increase awareness. And a link to a Mercedes website will improve its credibility as well.

Search Engines and LB - LB results is a much wider search engine exposure simply because links would have more keywords, which would result in the website performing well on searches

Search Engine Indexing - LB helps the website getting indexed by search engines resulting in the website performing well on google or yahoo searches

LB is , today, considered a worthwhile investment. In fact, LB as spawned a mini-economy of companies selling links or offering to build links as well.

Campaign discrepancies and how to avoid them

A regular problem that confronts a network is one of ‘discrepancy’ – discrepancy in number of clicks / impressions as recorded by the network and the advertiser. Discrepancies can range from 5 % to over 50 %.. Yes, believe it or not, high discrepancy campaigns are not unheard of.

An obvious question is, why do discrepancies occur? Let’s examine some reasons –

Wrong Targeting

Targeting criteria need to be clear at the outset. Often, a US targeted campaign ends up being targeted inaccurately. Impressions are delivered outside US inventory. This obviously results in a difference in the impressions count between the network and advertiser. This seems rather a silly reason for a discrepancy to occur; but occur it does. And quite often at that!

Multiple Ad-servers

A single ad impression often ends up running thru’ 2 -3 ad servers – the advertiser’s ad-server, the network’s ad-server and the publisher’s ad-sever. When piece of communication passes thru’ more than one person, message distortion is par for the course. Likewise, an impression going thru’ multiple ad-servers is recipe for discrepancy

Internet Problems

An impression may be served on the advertiser’s ad-server. But between the ad being served and being seen on the user actually seeing it, there could be a broadband issue at the user end. This would then result in an impression being counted by the advertiser but not by the network or publisher.

Bandwidth

Bandwidths are not consistent across the ad delivery chain – advertiser to network to user. Even with advancement of broadband technology, there are regions in the world were low bandwidth are a reality. Bandwidth difference cause latencies leading to a discrepancy.

Incomplete delivery of Ad

Often, an ad is partially loaded on the user system before he/she clicks away. This will result in a count at the advertiser end but no count at the publisher end.

Ad-server idiosyncrasies

Different ad-servers use different logic ( algorithms) to count an impression leading to discrepancies.

It is clear that discrepancies will occur. How do we handle such discrepancies?

Firstly, we need to recognize that discrepancies are a reality. It would be wise, therefore, to have a ‘discrepancy’ clause in the IO. Some IOs for example, state that a 5 % discrepancy is acceptable. Discrepancy percentage would be based on kind of campaign, geographies covered ( whether there are bandwidth problems in the target markets) etc.

Secondly, it is advisable to keep close tabs on a campaign in its initial days. Information on number of impressions/clicks should be exchanged between advertiser and network on a daily basis. Discrepancies should be discussed, analyzed and a way forward agreed.

Thirdly, the advertiser should share information on any incorrect targeting that is happening. For example, therefore, if a US targeted campaign appears in the UK, such information should be brought to the notice of the network immediately enabling them to take remedial action.

Case study of a successful CPA campaign

There is a perpetual conundrum of the client wishing to run a CPA campaign and the network running a CPM/CPC campaign. Both for obvious reasons.

I have , in my earlier articles, spoken of using a combination of CPM and google adwords to meet CPA objectives.

Today I am going to discuss a case study where this combination was effectively used to deliver CPA results month on month.

The client was a online merchant – buying and selling electronic items and used mobile telephones. The client brief to the network was that they wished to run a cost per action campaign – where the payout would happen when the visitor to the website would buy an item on the website.

The network, after considerable cogitation, decided to run a CPM + google adwords combination campaign. They went back to the client with a request for banners as well as a list of google adwords. The network, in their briefing to the client, explained to them how they proposed to run the campaign. They got a customer buy in to the idea and explained the collaborative model. The campaign required that the client share daily ‘sales’ information with the network.

Once the campaign was live, the network and the client worked very closely. The client shared ‘sales’ information on a daily basis. This helped the network correlate number of impressions/clicks with sales. The network optimized the campaign increasing number of clicks/impressions. After a week of the campaign going live, between the network and the client a figure of number of impressions a day and number of clicks was arrived at. This figure was based on client expectation of number of sales as well as Network ROIs. Obviously number of impressions/clicks could be continuously increased. Such an increase would get more sales, but compromised Network ROIs. The number of impressions/clicks per day was a win-win for both network and client – from the network point of view, it assured good ROIs , while from the customer point of view as well acceptable ROIs were achieved.