Selling an advertising campaign, either to a client or to yourown colleagues, can be nerve-wracking. You’ve explained thesegmentation strategy, outlined the creative executions andtalked about the cost of acquisition, when someone raises theirhand and asks, “But how will we know if this campaign works?”
Measurement isn’t the most glamorous part of a campaign, but isessential in ensuring that spending is justified. Before theadvent of the Internet, marketers largely paid lip service tothe discipline of measurement; it was hard to do, often wildlyinaccurate, and conclusions from which to improve futurecampaigns were difficult to draw. But the emergence of onlineand interactive marketing has changed all that; these new fieldsare leading a new, measurement-based approach tomarketing.
Online advertising’s fortunes have, over recent years, very muchfollowed those of the Internet itself; initially the darling ofthe advertising industry, a profound scepticism has settled inamongst media planners and buyers as rates have plummeted andusers have become increasingly frustrated by intrusiveadvertisements (several ISPs offer ‘pop-up killers’ as part oftheir packages nowadays). But the saving grace of onlineadvertising is its measurability.
The immediate, comparatively accurate, and easily accessiblemeasurability of online advertising is fundamentally changingthe way that marketers approach the discipline, allowing them totake a much more iterative, almost experimental approach:rolling out a run of advertising, measuring the delivery andresponse in near-real time, and tweaking placements andcreatives during the run of the campaign to maximise itseffectiveness.
However, until recently, measurement of response to onlineadvertising stopped, as it were, ‘at the door’: whilst measuringthe initial click-through response to an ad was possible,analysis of what the user did once they’d arrived at the sitethe ad was directing them to was not. This is where webanalytics comes in.
Simple analysis of an online ad campaign might tell you that ahundred people responded to an ad by clicking on it (a so-called‘clickthrough’). Web analytics can enhance this data byrevealing that, of these hundred people, ten went on to buysomething, and of those, one purchaser spent more than a hundredpounds. With information on how much the campaign cost, and theprofit margins on purchases made, you can reach a return oninvestment figure for the campaign.
Web analytics is the discipline of measuring and analysing thebehaviour patterns of visitors interacting with web sites. It ispossible to extract a rich set of data about the paths visitorstake through a site, which pages they view, and which actions(such as making a purchase) they carry out. Web analytics canalso reveal how visitors were directed (‘referred’) to a website, which is of real value when analysing the effectiveness ofonline marketing.
The field of web analytics is a broad one, encompassing a widevariety of approaches. Broadly, though, the field can be brokendown into three types of solution:
- Delivery system-specific measurement Online ad servers come with their own measurement systems, which can reveal how many times a particular ad was viewed, and how many people clicked through. Some advanced ad servers can also provide ‘post-click’ or ‘post-impression’ analysis, where the usage of key pages from the target site (the site to which users are directed by the advertising) is tracked, and linked to the ad delivery and initial response data. This can provide a good solution for analysing online advertising in isolation; but it can’t give any detail on how, say, an e-mail campaign compares with a banner campaign.
- Target site behaviour analysis Online behaviour analysis tools can deliver a great deal of information about people’s behaviour on the target site after they’ve responded to online advertising. Powerful behaviour analysis tools can segment visitors based upon the marketing they’ve responded to, and deliver detailed information about the effectiveness of individual elements of a campaign that might span online and offline media (although in the latter case, only those components of an offline campaign which are aimed at driving traffic to a web site). Another benefit of this kind of analysis is that it allows the behaviour of ad respondents to be compared with that of visitors who haven’t responded to advertising; something that delivery system-specific measurement can’t do.
- Multi-site analysis services Several services are available that can help gauge a site’s popularity against its competitors, either by analysing the logs of Internet Service Providers, or by conducting panel-based research. Whilst these services can help benchmark a site’s performance compared to the rest of the industry, they can’t usually give an idea of the impact that specific advertising campaigns have had.
Comprehensive campaign analysis will likely combine elements of several of these types of solution. In order to understand which kinds of measurement to use, you need to ask yourself what you’re trying to measure, and what the campaign is trying to achieve.
Deciding what you want to measure, and why
If you buy online advertising, then the terms impressions, clickthroughs and conversions will be familiar to you. These represent different ways of measuring the reach and effectiveness of a campaign. Impressions represents the number of times an ad was shown. Clickthroughs is a measure of the number of times an ad was clicked on. Finally, Conversions measures the number of times that users who responded to the advertising performed some desired action, such as buying something, or signing up for a newsletter.
If you are primarily interested in brand exposure, then impressions may be all you need to measure. If you need to get a better picture of audience engagement, then clickthroughs may give you all the information you need. However, if you want to really build an accurate picture of ad campaign effectiveness and ROI, then you will need to measure conversion.
Impressions and clickthroughs can be measured with a delivery system (i.e. ad server) measurement system, and such systems can also deliver some conversion information. But to get the fullest picture of the activity of those users who have seen or responded to online advertising, you need to integrate target site behaviour analysis. Ideally, you should look for a solution that integrates both data collection approaches.
Making it work
For the most complete campaign analysis, data from your ad server and from your web site should be integrated. Broadly speaking, you’re looking to segment the behaviour of visitors to the site according to the information you have about them, such as which creative execution they responded to, or which page (on which site) they clicked through from (the so-called ‘placement’ of the ad).
The ad networks or media owners who are actually running the ads may be able to provide you with demographic information about their user base (in fact, before you even start a campaign, you’ll be using this information to decide which sites to buy inventory on). But once in a campaign, the best indicator of the fit between your message, execution and placement and your target audience is the response and conversion rates you see.
Understanding the detail about conversions is where the data from the target site comes in. It’s possible to look at the URLs that a user requests once they’ve arrived on the site from an ad, but this raw information is rarely revealing enough on its own; what is needed is to relate the URLs to content from the site’s content management system or e-commerce platform. And if you can also gather some information about the user, then you can use this to further segment your audience, according to, say, age group, or gender.
For example, knowing that someone visited your site at 2.25pm having clicked on an ad, bought product XC345 and then left the site is not that useful. But by integrating the data trails from your web site with data from your marketing service provider and your content management system, you can add a great deal of value to the basic data; in the above example, you might learn that the visitor was male, aged 45, from Birmingham, who arrived at the site from an advert on Lycos offering a discount on shoes. He bought a pair of shoes, and a shirt that was also reduced. He looked at trousers, all at full price, but didn’t purchase.
Once you have this enhanced information to hand, you can decide what your conversion metric is. Is it the number of purchases on an e-commerce site, or their value? Are you more interested in people who moved quickly to buy, or those who took more time to research their purchase? Are there other aspects of site usage that you’re keen to encourage, such as repeat visits?
What most media buyers are looking for is the same effectiveness (measured in terms of click-through or conversion rates) across all the different placements and executions in a campaign. It’s possible to calculate what’s known as a ‘relative contribution’ figure for a component of a campaign (such as all the different sites where ads are being served). The relative contribution of, say, a publisher site is calculated by dividing the response rate for that site by the overall response rate for the campaign; it’s expressed as a percentage.
If a site has a relative contribution of over 100%, it’s punching above its weight, and could possibly cope with a greater volume of ads. If, on the other hand, a site is delivering a relative contribution of only 50%, it’s not pulling its weight, and the ads that are running on it should be diverted elsewhere. Ideally, the relative contribution of the sites, placements or executions in a campaign should be close to 100%, indicating that the campaign is perfectly balanced and delivering maximum effectiveness.
In addition to this kind of data, analysis of behaviour patterns can reveal interesting ‘non-conversion’ behaviour, which can lead to improved campaigns for the future, or even qualify as quite robust market research in its own right. In the above example, the campaign for shoes might generate a low conversion rate, and be judged a failure; but further analysis might reveal that ad respondents bought lots of shirts, even though these weren’t the focus of the campaign. This kind of intelligence is valuable for planning future product strategy, not just ad campaigns. It can also help you to distinguish between a campaign that is under-performing because it has weak creative executions, and one that is failing because there is a mismatch between your target demographic and the types of people who use the sites where your ads are being run.
Finally, performing sophisticated target-site analysis allows you to compare online advertising with other types of online marketing, such as e-mail marketing or search engine placements. In the latter case in particular, there is a lack of tools provided by the search engine companies themselves to measure post-click behaviour, so web analytics can fill this gap.
Brave new world
Attitudes to online advertising have taken a roller coaster ride; the scepticism that accompanied the dotcom collapse is starting to evaporate as the medium shows its value through the increasing sophistication of the measurement tools available for it.
We are now at a stage where online advertising can be analysed in granular detail in real time. The measurement of online advertising no longer revolves around simply driving traffic to a website; it can carry right through to bottom line profit. The advert that drove 100,000 people to a particular site might well prove less cost-effective than the advert that only attracted 10,000, because the 10,000 spent money while the 100,000 merely browsed.
The advertising industry has craved immediate and accurate measurement since its inception. In the online world, where there is an abundance of data that is relatively easy to capture, measure and analyse, that is finally starting to become a reality.
Ian Thomas is a founding member and is Strategic Development Director of WebAbacus.