© Media Forty Two. All rights reserved. 2021.

The Pitfalls of Perfection in Attribution Modelling

Is perfection in attribution modelling wagging the B2B marketing dog?

There are lots of attribution models you can go with. Some are really flawed, some are a little bit flawed – none are perfect. This doesn’t mean you shouldn’t attempt to attribute your marketing activities, it merely means that you should be aware of the flaws and take them into account before making any final analysis.

Attribution is a mechanism whereby you measure the impact of marketing spend, to the acquisition of customers. Note it is not necessary to define a monetary value to attribution modelling, although many will. For the purposes of this article we’re going to give a couple of examples but if you’d like to learn more about attribution models I really quite like this article from Bizible.

Example 1.

Technology Company A attends an exhibition. In total Company A spends £70,000 including the cost of three sales people for 2 days. Marketing receive the data from 200 people who visit their stand. Sales have 5 meetings which lead to sales. These contracts are worth £500k per annum total. Pipeline modelling shows that the cost of the activity was one seventh, of the revenue generated in year one.

Example 2.

Technology company A invests £50,000 into content syndication paying £25k to a content agency and £25k to third party publishers. 500 leads are generated which are put into marketing automation and within three months 3 opportunities are raised. One of these converts to a sale worth £125k per annum. Pipeline modelling shows the cost of the activity was just under half the revenue generated in a year.

Pitfalls with this type of attribution

Obviously the above models are vastly simplified. Likely each client has an average lifetime value associated based on client lifecycle averages and the costs associated to activity include more internal recharges. However both these simplified examples are examples of last touch attribution.

The biggest problem with the above is; that good marketing, in a data led world, should be based on testing under fair conditions. Judging suppliers by pipeline activity, means judging suppliers on variables outside of their control (the effectiveness of your sales team is not within their control) and judging the effectiveness of a campaign where the resources for follow up (different sales people, salespeople having a bad day etc.) vary. Even if you follow up with a nurture program, unless your messaging is identical for every lead (which would be extremely unwise as conversions would suffer due to a lack of personalisation) you’re not presenting a fair test.

The next problem you have is pipeline lead times. One of the reasons account based marketing has resonated so well with B2B technology marketers is that it recognises the need to focus on large accounts, and the need to map out stakeholders across these organisations. Big software deals worth millions, take time. Twelve month sales cycles are not uncommon. How can a marketer, planning tactical spend, get pipeline results quick enough to know whether to repeat activity?

In the above examples a huge contract would be attributed to one single event. This ignores the brand activity, sponsorships, internal email, social retargeting, and any other activity you’re running entirely, and associates zero attribution towards these other channels that were likely critical in the buying decision. Incidentally this goes both ways – the 195 contacts obtained from the exhibition that didn’t convert to a sale might in the future. B2B marketing involves a very close relationship between marketing and sales. But marketing can’t become sales, even though in B2B it is necessary to work extremely closely together.

Understanding what is possible

A marketing department will always have numerous campaigns and different publishers/channels on trial. That these run concurrently mean a fair split test is never fully possible. It is simply impossible to ensure that the knock-on effect of one campaign does not impact another. I’m yet to see an example of a perfectly attributed campaign – but some attribution is better than no attribution and a balance is required between time spent attributing activity (and opening some rabbit holes) and the decisions you hope to make as a result of attribution.

A simple first step to benchmarking

If you have chosen to advertise then chances are this was in part based on an audience match. Publisher A reaches IT professionals so we’ll put some banners on their site.The publisher’s job is not to convert that person (or it shouldn’t be) into your customer. The conversion rates are as much in the advertiser’s control as the publishers.

You require a way to fairly benchmark the audience provided. If personal data is being generated (events, content syndication – any activity where the call to action results in a form fill) then it is very easy to define the source of each record. As a first step in attribution I always recommend understanding the percentage accuracy match rate of your suppliers – something surprisingly few do. If you know what your customer profile looks like (maybe you set three levels; A customers, B customers and C customers) then you can fairly quickly analyse the percentage of the audience meeting those specifications.

If you run the above process with events you will very quickly have a benchmark for what a good event data match rate looks like. Ignoring the complexities of pipeline etc. you can judge an event company (or any media provider) on whether they have the correct audience. There are more, and usually company specific, variables to consider. But this provides a good start that can help marketers make fast buying decisions.

If you’d like any further information please contact us.

*The plot of Game of Thrones doesn’t seem so complicated now does it?

Leave a comment: