±±¾©Èü³µpk10

Though it's one of the most measurable, effective and cost-effective vehicles for advertising, the internet has not always been so good at promoting itself. If you talk to marketers who are not too clued up on what the technology can offer, you'll generally hear the opinion that hardly anyone clicks on an online ad these days, so how can it be an effective medium? But, of course, there's much more to the story.

According to Danny Meadows-Klue, chief executive of the Interactive Advertising Bureau (IAB): "Internet advertising has given rise to some exceptionally sophisticated media buying and advertising creative. The technology powering these systems is breathtaking in what it can do."

One of the most impressive aspects of the technology is that clients can be given real-time reports on how many people are looking at their ads and what their reactions have been. However, this transparency can sometimes cause problems as there can be discrepancies at times in the data, which can lead to disagreements between the media owners and advertisers on how many ads have been served or clicked through.

This kind of data discrepancy can lead to problems since most ad campaigns are paid for on a cost-per-thousand impressions (CPM) or cost-per-click (CPC) basis. There's nothing more likely to sour a relationship than a row over the bill.

So how is it possible for the client to come up with one set of figures regarding how many people have viewed an ad or clicked through to its web site, while the media owner records another set? And what is currently being done to prevent discrepancies and reassure advertisers that, by committing their resources to online advertising, they're getting one of the best deals around and not being cheated in any way?

The IAB is working to help the effective promotion of online marketing, and has been developing industry standards for measurement and accountability to reassure newcomers to the medium. The organisation has just updated its guidelines on dealing with data discrepancies and is investigating the reasons for this. Its findings can be found in Smarter Advertising Statistics, available on the IAB web site (www.iabuk.net) or by emailing secretariat@iabuk.net.

First of all, it's important to know what happens when an ad is delivered to a browser. When a user logs on to a site, the server of the media owner site receives a request for that page file. If there are any ads being shown on that page, this request will also ask for the relevant ad. Usually, the ad space is managed by another server - the advertising server - which decides which ad to send, according to various criteria. This criteria is based on what it knows about the user, such as demographics or content preferences, the time of day or the weighting of a campaign against others, and so on. The ad server will either send the ad directly itself or link to a third server - the advertising agency's server - to collect the ad and send it to the user.

In theory, this complex process should take no more than a split second, from the moment that the user clicks the mouse to the ad appearing on the requested page alongside all the other content. Thus, the ad has now been served to the user and the ad server of the media owner and, if relevant, the ad agency will have logged the ad as having been successfully served to the user. The stats are updated accordingly.

However, sometimes the situation is not so straightforward and this is where discrepancies can creep in. It is essential that what is being measured in a campaign report by the media owner's server and the agency's server is the same thing.

In the past, discrepancies have been caused by one server - usually the media owner's - measuring the number of ads requested by users while the other server measures only the number that have been successfully served.

In this situation, if a user decides not to let a page load fully and clicks elsewhere, the servers may have been requested to serve the ad but they won't get the chance to do so. This can happen if the network is running slowly. If the client and media owner aren't measuring the same thing, there will be an immediate discrepancy.

Will Green, traffic manager at Rivals Digital Media, knows this problem only too well. "Our ad server, Adtech, counts the number of ad requests that are made to the system after it has responded to the user's browser with the ad to be delivered. This corresponds exactly to the guidelines of the IAB."

But he recognises that other ad servers act differently, leaving some room for discrepancy. "The problem is that ad servers use different counting methods and there are no common counting grounds or set standards." The solution, he suggests, is for greater consistency across the industry, but he believes that "all players in the market are seriously interested in finding a solution to these discrepancies".

According to Green, Adtech has been working with these ad servers for several months now to try to find a common solution to this issue. "Adtech is currently running beta tests and the preliminary results are looking very promising," he adds.

Another point of discrepancy lies in the common practice of 'caching' images and other larger files to speed up the transmission of pages across the network. In this way, a page which is viewed often by a user will usually be stored by their own browser or ISP servers to avoid having to make a request to the main media owner's servers each and every time they look at the page. So, although a user may look at a particular page several times a day and see the 'cached' ad which is on it, there is no way for the server of either the media owner or the agency to count this in their reporting because no request will have been made to serve the ad. This is becoming less of a problem as most advertising servers are now equipped with special 'cache-busting' software. But it is important for this software to be regularly checked and kept up to date to avoid any discrepancies with the client.

Martin Filz, European vice-president of web analytics firm RedSheriff, agrees that caching is becoming less of an issue today than it was in the past. "Caching has varying degrees of importance, depending on the popularity and size of the web site," he points out. "A typical rule of thumb is that the more popular the web site is, the higher the likelihood that it will be cached by ISPs or proxy servers.

"Today, with bandwidth at such a very low price, we are unlikely to see caching of much more than 20 per cent of content. Just three years ago, the issue of caching was much greater and we would often see caching at around 50 per cent of content," Filz continues.

Discrepancies have also arisen in the past when clients and media owners have not specifically agreed on their definition of 'invalid' clicks.

These 'invalid' requests come from spider and robot visitors which are cataloguing the web site, usually for a search engine. Since most search engines such as Google continually trawl the internet, cataloguing and updating, so as to be able to provide comprehensive and accurate search results to their users, it is likely that any web site will receive a reasonably regular number of these 'invalid' clicks.

For now, there is no single industry-wide list of IP addresses that can be used to define 'invalid' clicks, so each ad server will have different criteria for measuring them. The IAB is currently working on formulating such a centralised list, which will hopefully close this loophole in reporting for good.

Perhaps the thorniest problem in online ad campaign reporting is caused by users making use of the 'back' button. This is a part of every user's experience of surfing the internet. They may click a link accidentally and have to click back or the they might wish to do a u-turn and retrace their steps back to a previous web site. Whatever the reason for clicking on the 'back' button, if it happens before a page has loaded fully, it will cause a headache for ad server stats. Although the ad may have been requested by one server, the request may not have reached the next one in the chain.

Furthermore, on a particularly slow network day, an ad may have been served by the server at the end of the chain but not reached the previous ad server, which will only register a request and not a served ad.

There is not much that either the media owner, ad agency or the client can do about the whims of surfers on the internet. Perhaps one of the most common reasons for the 'back' button being clicked is when a page takes too long to load. So, of course, it makes sense for all online content providers to ensure that their pages and files are as engaging and as content-light as possible, so that pages can be served both quickly and efficiently to users. But beyond this general rule of good housekeeping, there's not much else that can be done to stop someone retracing their tracks.

This discrepancy can also affect the number of clickthroughs that are counted. This is particularly pertinent to pay-per-click search engines, which offer deals to advertisers based on the number of people who click through to the advertiser's site. If a client has agreed to a pay-per-clickthrough deal, the discrepancy caused by the search engine recording a user clicking the link to the advertisers' site but not getting there because they changed their mind or the system was too slow, will be a big problem. The client will be unwilling to pay, although the search engine will have recorded a successful clickthrough.

Tony Samios, director of advertising services at Espotting Media, a major pay-per-click player, believes the criticism levelled at search sites for their charging policies are unfair. One reason he cites for discrepancies occurring is an unusual one. "Most click-tracking software counts one IP address per day, ignoring the rest of the traffic from that address," he says. "Proxy gateways and internet cafes will often connect through the same IP address and these must be accounted for."

But the integrity of the industry is vital for the success of Espotting and all other pay-per-click search providers. "Protecting our advertisers from illegitimate clicks is of utmost importance for Espotting," continues Samios. "The integrity of our marketplace is of primary concern, hence the fact that we have invested substantial resources in fraud-protection technology."

RedSheriff's Filz feels the importance of data discrepancy cannot be overstated. "Until advertisers feel confident in the figures, they won't be investing large ad spends online."

His sentiments are echoed by Tom Bowman, director of revenue strategy at MSN Europe, Middle East & Africa and chair of the IAB's market statistics working group. "Trust is the big one. We need to understand what the data is really telling us and develop trust between parties. This is the only right basis for doing business," he says.

The IAB's contribution to the debate, Smarter Advertising Statistics, has some pragmatic advice on helping companies ensure that discrepancy issues are dealt with quickly: "If discrepancies between the reports from each ad server are going to appear, then these will generally be seen from the very first days of the campaign. If the differences are not spotted until the campaign has been live for a week or more, then resolving the problem becomes significantly more difficult.

"Monitoring campaigns in the first few days will reveal discrepancies and then you can begin trouble-shooting. Both the media owners and the agencies should check the performance of a campaign within 48 hours after it has been launched."

The long-term solutions to the issue would appear to be the ongoing development of industry-wide standards and a higher level of trust between customers and service providers. However, Tim Brown, managing director of the IAB and head of the team which put together its report on data discrepancies, is quick to point out that a certain amount of perspective is needed when looking at this thorny issue.

"The internet is far more accountable and transparent than other mediums, which rely on sample basis and methodologies that can be unreliable sometimes," says Brown. "Stakeholders need to get around the table to have an understanding of the issues, continue to create standards that will reduce discrepancies and make them more consistent. Most importantly, they need to educate the market that this medium is the most accountable and transparent one around."

RedSheriff's Filz is more forthright in his own suggestions: "The UK has suffered by not having a single internet body," he says. "In countries which do have such bodies, such as the KIA in Sweden, FAKDIS in Denmark and the CIM in Belgium, the situation and confidence of advertisers is much clearer. It is time that the UK internet industry united behind a common set of methodologies. After all, TV has BARB and radio has RAJAR."

Perhaps ABC Electronic has the answer. Managing director Richard Foan sees hope for some light at the end of the tunnel. "We have to accept that differences will arise at the different points of the delivery process where measurement occurs," he points out. "But we can use industry standards to minimise these differences, independent of verification of claims, so that all parties have increased confidence in the totals."

In an industry first, ABC Electronic has just completed a campaign audit for Real Media's Open Advertiser platform. If this kind of independent auditing became an industry-wide practice, the data discrepancy issue could be put to rest.