Chatting with a TagMan client (who I can’t name) about the attribution data we provide them, I was really impressed with the approach they have taken in assessing the quality (and ROI) of campaigns and how they use this data in their media planning.

The digital currency of awarding credit is still on the last click that generated the sale, and this is what they use for awarding their affiliates and other CPA channels commission for the business they generate. However, they use the attribution analysis of the campaigns to work out if a CPA channel is producing a positive ROI – and therefore if they should continue to invest in it.

Non-brand SEO vs. Affiliates

I’d like to illustrate this by looking at two of the campaigns we are tracking for them: non branded SEO and the affiliate sales through a well known and respected network.

On a last-click win analysis (how commission is awarded), non-brand terms in natural search results generated 600 conversions with revenue of £18,000 and the affiliate generated 4,300 conversions with revenue of £170,000.

On the face of it, it doesn’t look like SEO non brand really does much for them, and that the affiliate is doing a far better job.

The catch comes when marketers have a hunch that due to cash-back and voucher-code sites, the affiliate is cannibalising the sales of the other campaigns – shall we call it goal-hanging – and make a decision to stop working with the affiliate on this hunch.

Applied attribution

However, if you look at the sales and revenue each campaign generated not by last click, but by an attribution model it tells a very different story and with the data you can make a much better decision.

Using a flat attribution model where the credit and revenue of each sale is split evenly between all the campaigns that show up in the path to conversion, we see that, over the same date range, the non-brand SEO attributed sales (that is the sales where non-brand natural results show in the conversion path) were 4,050 with revenue of £145,000 and the affiliate generated 1,900 attributed sales with £73,000 revenue.

This shows the marketers hunch was partly right, but the key number is the attributed revenue by both campaigns.  For ease of numbers, let’s say this client had a profit margin of 10%.  Therefore the profit on the SEO work was £14,500 while the profit of the affiliate was £7,300.

Change in budget spend

As it happened, this client didn’t spent nearly £14,500 on SEO marketing and as a result of this data now spend incredibly more and are looking forward to seeing this channel push up last click conversions to other channels.

Moreover, while the affiliate wasn’t generating as much value as reported by last click, the profit was still higher than the commission paid out – i.e. the affiliate is still a channel with positive ROI even with the cash-back and voucher-code sites, and so the client also continues to invest heavily in this area.

I purposefully haven’t provided the length of time this analysis was over as the idea can work for smaller companies just as much for larger companies.  Whether this data spans a single day or three months, it still ensures that as a marketer, you are basing decisions on data and not hunches.

I’d like to thank Dixons for their latest tube ads which should bring to light how important the consideration and selection of products are as well as the final purchase in the sales process.

Now while they happily suggest consumers go elsewhere to pick which TV they want and visit their website to make the sale, many brands should be asking for consumers to consider and select products from within their own site as they will have a much higher chance of eventually selling that product to the prospect when the time comes.

With this in mind, using the traditional mechanism of a ‘last click’ wins model, only the marketing efforts which drive the visit of the eventual sale will get the credit, whereas Dixons rightly highlights the prospect’s need to spend time considering what to buy before the final plunge.

By using a last click win model, all the visits to your website which are for selecting and considering products will get no credit for the sale, even though they have an immense effect on the eventual conversion on that customer.

By running an attribution model which shares the credit of a campaign against all marketing events the prospect has responded to, your media planning will be more wisely spent, and the eventual returns much higher.  Ie, users who have visited the site already and looked for products are far more likely to buy that product from you on an eventual ‘purchase’ visit.

A few case studies to prove this are on the way from a number of major high street retailer and travel companies.  Get in touch and I’ll make sure you see them first!

We know that consumers will engage with a brand several times before purchasing and we’re moving to a point where this journey can be accurately mapped and assumptions made.

While retail clients consistently favour paying on a last-click wins basis – ‘the last argument before purchase must be the most convincing’ – there are many examples of how marketers are more sensitive to their users’ psychology.

Examples of ‘engagement’

Here’s a few examples of where engagement mapping across the standard digital touch points – display (banners on publisher sites, ad networks), affiliates, email marketing, paid & natural search -has been relevant:

1. I went to a presentation a few years ago by a rich media vendor where a study showed the optimum number of times a banner should be seen (to maximise click-to-conversion rate) is three.

2. Another study showed that video ads should be between 10-20 seconds in length to optimise click-through rate.

3. A marketing manager at a mobile phone reseller spent greater budget on display than the click-conversions warranted – he could see data that showed users were more likely to convert if they’d seen banner ads, even though the last touch point before conversion was consistently clicks via PPC or SEO.

4. A gambling client of mine would pay the CPA bounty on a first-click wins basis; they were certain that getting a user to visit their site in the first instance was the hardest part of the sales process given the saturated market (their target market was users who were already a member of 3-5 gambling sites).

So some verticals rely on exploiting their users’ impulse to purchase with timed-limited offers specific to them while others feel they’re racing to reach their target market, and use engaging (sometimes uncomfortably distracting) copy to stand out; interrupting a user’s typical ‘ad blind’ journey. Others contrive for the user to encounter and compare their product in search cycles and user reviews; convinced theirs will win through on merit.

My point is that marketers are responding to their users’ psychology with competitors, buying cycles and the holistic digital journey in mind. It’s apparent that the brain reacts to repetition, unusual stimuli, and that our thought processes before ad engagement can positively or negatively affect that engagement (what was the user doing before we dragged them to our site?).

We are thinking in engagement maps, touch points and ad sequencing; and it’s clear this is having an impact on purchase probability. We’re moving away from dividing budget by sales channel and to spending on the combination and timing of digital media.

So why one commission?

But, why do we still tend to pay 100% of CPA bounty to the significant channel (whether first or last click, for example), instead of paying a significant percentage of CPA bounty to the significant channel? I don’t believe first or last click is how we should attribute credit, pay our acquisition bounties or plan our budgets. Many sales channels are adept at prolific cookie dropping, so that they exist in the majority of conversion paths. Others will even persuade users to delete their cookies so that a specific newly dropped cookie overrides them.

So what I’d recommend is a ‘best click’ model. This is how that might be achieved:

1. Use a first or last-click model as a foundation and rank your engagement points: highest for those you feel are doing the most selling (e.g. affiliates on a one-day click window, perhaps banners on a one-hour click window); lowest for those you feel are the most transactional (i.e. those that are the obvious route to buying a product – and retailer – a customer has already decided on) or are serial cookie droppers (e.g. PPC on brand terms, maybe one-day view windows on reach ad networks).

2. Arrange for the correct conversion tag to be written into the conversion page based on rank and timing, using them to award higher proportions of commission (and overall credit) to those high-rank channels where they appear in the conversion path.

Dynamic awarding & attribution: proportionate commissions

Dynamic awarding & attribution: PPC given higher priority

Our charts show how – by giving one channel a higher priority (because you rank it as being a better driver of ‘unique’ sales) – you can also give it a higher proportion of credit and commission. In the second chart, PPC is ranked as highest priority so – even when it appears as the second to last click – it gets the most credit.

This is a complex process but one that requires a useful exercise – to really think through which channels you believe in as drivers of sales you wouldn’t otherwise have, rather than those that deliver customers who were already committed to you and your product. Once in place, the way you rank channels and executions (like keywords) and attribute commission can be constantly refined.

Of course, even this model is flawed; any model that tries to pinpoint the significant engagement event among many is inconsistent with our users’ psychologies and prone to abuse. Still, attributing credit across multiple touch points, from both a reporting and awarding point of view helps to alleviate the distortion created by the sales channels currently fighting for that first or last click.

Progressive attribution

CPA sales channels should be incentivised for finding the right users and speaking to their psychology, and discouraged from just putting themselves at the right point of the conversion path to claim the final click. We can now pass a weighted proportion of the CPA bounty to an affiliate, or that can load a conversion tag for a weighted proportion of the time; all according to the attribution model selected. You may know this method as dynamic awarding or applied attribution.

Despite the flaws in drawing accurate conclusions from user journeys, I feel the marketing community intuitively builds a critical mass of engagement. I’ve lost count of the marketers I know that can’t prove why display contributes to post-view conversions, but are convinced it does (there’s a fair few that can prove it too!). Ad saturation here or under-exposure there can reverse the build-up of pressure – the critical sales point must be capitalised on before the pressure fades or is directed away by a competitor. Modelling this critical mass of engagement with correct statistical treatment (which is what the digital marketplace is acclaimed for), will obviously yield dividends.