The concept of moving beyond the 'last click' has existing in the ad tech industry for over a decade. I clearly recall conversations with advertisers in 2004 where they asked: "How often did potential customers see my ads on publisher X before clicking on the same ad on publisher Y?" It's a frustrating experience that a dozen years have passed and we (the industry) are still asking the same basic questions.
There are obviously several serious stumbling-blocks with a successful deployment of Conversion Attribution technology. If this type of ad tech was simple to deploy then it would be rolled out as 'standard' for all digital campaigns.
In order to talk about some of tricky troubles with in this area I've interviewed several Conversion Attribution ad tech companies and asked them to share their top 10 tips for a successful deployment. Below is a summary of these tips.
1. Conversion Attribution isn't for Everybody: The rollout of this technology requires a considerable amount of project management resource. It's not a 'plug and play' solution. One leading CA technology found that (a) the technology works better for finance, travel and auto clients but not so well for FMCG, pharma and retail clients and (b) the technology works better for spends of $2M+. Anything less became harder to justify the investment. In summary: The technology is better suited towards larger and more digitally savvy advertisers.
2. Data Hygiene is half the Battle: Conversion Attribution need good quality data in order to produce good quality results. Remember that it's 'garbage-in garbage-out'. There are cases when a single publisher is spelled several different ways. (e.g. Yahoo, Yahoo!, Yahoo.com. Yahoo Inc, etc...). To a conversion attribution engine these are different publishers competing for the same budget. Image the disappointment when the CA report suggests to move 50% of the budget from Yahoo.com to Yahoo!. This type of report makes no sense and the entire CA model fails. In summary: The advertiser needs to ask your media agency to deploy a 'Naming Conversion' technology. Deploying a good naming conversion tool may be time consuming in the beginning but it will pay dividends in the end.
3. Access to Granular Data: Not only does the data need to be accurate (see point above) but the data also needs to be granular. A whole new article could be written on just a 'data check list' however to keep it brief consider sharing the following information (a) cost data from your media booking system e.g. Symphony or Mediaocean (b) cookie level data from the ad-server e.g. Sizmek or DCM (c) web-site analytics information e.g. Adobe or Google Analytics. All this information (plus more) will be required to be fed into the Conversion Attribution technology in order to produce accurate results. In summary: Any successful Conversion Attribution project should start by connecting the various data sources into the CA tool.
4. Selecting the Right Model: The goal of Conversion Attribution is to analyse the cumulative effect of each piece of digital advertising prior to a potential customer purchasing your product. However, how do you give different weighting to each advert? Does a recent exposure have more weighting than an older exposure? Does a larger video ad have more impact than a smaller static ad? Each of these questions is different for each advertiser. There isn't a 'one size fits' all model. There are many standard models to choose from including "Linear, Positional, Time Decay, etc..." Too often advertisers say "I don't know which model to choose so I'll let the algorithm decide". This is generally a bad idea. In summary: It's not rocket science and CMO's should be able to articulate the model that works best for their business.
5. Allocate Appropriate Budget: Conversion Attribution isn't cheap. However, somebody once told me that ad tech that was "good and cost nothing" was also most likely "good for nothing". Investing in an attribution technology should not be looked at as a hard cost to the business. The insights gained from attribution should drive efficiencies and pay for themselves over time. In summary: Expect that CA will be costly but advertisers should start to see a return in the end.
6. Select a 'Full Service' Agency: Separating your media spend to different media agencies will cause troubles when it comes to Conversion Attribution. Remember that the output of a CA report will always suggest to move media spend from one publisher (or media channel) to another. For instance; Consider the hypothetical example whereby the advertiser has a specialist for search and another specialist for programmatic. The output of the CA report will certainly move money between search and programmatic display. Neither of these two separate specialist agencies will want their budgets cut and may resist the suggestions of the CA technology. In summary: Select a Full Service agency where possible as it will have less resistance with rolling out the CA reporting.
7. How to Deal with Diminishing Returns: One of the core challenges of the product from a commercial point of view is dealing with the concept of diminishing returns. For instance; assume that the technology works perfectly and saves the advertiser several thousand dollars in the first month. Assume also that the savings from the second month will be less than the savings from the first month. This stands to reason as some of the learnings from the first month were automatically applied to the media schedule in the second month. In other words; as your campaign becomes more and more optimal then there are fewer and fewer adjustments for the technology to make. If you follow this logic to it's completion then it means that eventually the cost of the technology will outweigh the savings that the technology generates. This is the point of diminishing returns. In summary: Make sure your contract with the ad tech provider is a "dip in / dip out" contract whereby you can test your media plan without being locked into a monthly contract.
8. Education, Education, Education: This is a field that is updating quickly. Each of the ad tech companies has a development road-map and they are evolving at a frenetic pace. Each advertiser is also looking for an edge against their competitors and Conversion Attribution may be just the edge that they are looking for. The goal is to remain educated in this field. In summary: Have regular training sessions, lunch and learn, attend summits and stay abreast of other market case studies.
9. Timeliness and "Actionability" of Results: One of the main challenges of Conversion Attribution is the actionability of the results. At the end of the day Conversion Attribution will come up with suggestions of how to update a media plan. If these suggestions are put into the bottom draw of your filing cabinet and never used then Conversion Attribution is a waste of money. In summary: Be prepared to analyse your entire media plan each week and also be prepared to make wholesale changes to your plan if required.
10. Don't Expect to Save Money on day one: The technology has a relatively high monthly cost and it takes time to collect and analyse the results. It also takes more time for these recommendations to be applied and the media plan updated. Based on experience it will take three to four months before the savings are greater than the costs. In summary: don't expect to break even in the first month.
In conclusion: For many advertisers with a strong digital presence then attribution is a necessity not a luxury in this digital media age. Don't ask if you should be adopting Conversion Attribution technology you should be asking how you should be adopting it.
By Timothy Whitfield, director, technical operations at GroupM