Attribution: Or, Why Everyone Is Lying To You

Have you even been in the situation where everyone is reporting fantastic channel results but the business result still sucks?  Never been able to work out why the numbers your agencies report always exceed the numbers that you get from Finance?

There is an evil gremlin in your data and his name is attribution.

What is attribution?

Quite simply attribution is about assigning value to the right parties.  In the marketing world, this means assigning the correct value to each one of your marketing channels.

Anecdotally you know that TV advertising is great for branding and will have an influence on your ability to acquire customers; however it’s really hard to actually assign the value of that TV advert to each of your newly acquired customers.

It’s much easier to understand the value of paid search; someone clicks an ad, you pay a prescribed amount and they either become a customer or they don’t.  If they become a customer you (should) know exactly how much you paid to get them and can quickly extrapolate the value of the channel.  In this case it was easy to attribute the customer to the paid search channel.

It’s not quite that simple

It would be fantastic if attribution was as simple as knowing where someone came from and what they did, but that’s a greatly over simplified view of the world.  The reality is that there are many things that influence our behaviour and each one of those factors has ‘value’ and consequently we need to include it in our attribution model.

To put it simply, vary rarely will someone become a customer, lead, sale etc, on their very first interaction with your brand, be it online or offline.

Here’s a short story to put attribution into context:

You start by making a generic search query like ‘buy shoes’ in your favourite search engine.  You click through to the number 1 organic site and browse around for a while.  We’re going to call this site ‘Store A’.  You find a pair of Nike trainers you really like and take note of the model number.

Before buying you want to see if you can find the shoe cheaper elsewhere, so you type in another search like ‘Nike xxx compare prices’.  You now click through to an organic comparison site that shows you 10 stores offering the same shoe at varying prices.  You notice that there are some banner ads from ‘Store A’ offering the very same shoe you just looked at and a 5% off voucher code.  You write down the voucher code for ‘Store A’, just in case you want to purchase there later.

After looking at all the stores offering that shoe, you find their prices are all similar after you take into account shipping costs. Now that you have the coupon from ‘Store A’ – they are the best deal.

You decide that it’s still a lot of money to spend on shoes so you decide to sleep on it and if you wake up still wanting to buy the shoes in the morning, you will.

You wake up and it’s a beautiful day outside and you want to go for a run, you think about how torn up your old trainers are and you decide that you do want those shoes and you’re going to purchase them from ‘Store A’.

You go straight to your browser and type ‘Store A’ into your search engine and click the Paid Search ad for the brand, you navigate to the shoe page and complete the purchase online.

You congratulate yourself on being a savvy and restrained online shopper and look forward to receiving your new shoes.

Ask yourself quickly, which was the most important step in this process for ‘Store A’ in turning you into a new customer?

a)     Your original Organic Search for the generic term

b)      Seeing the special offer display ad

c)      Your branded search query

Under the typical Last Click wins attribution model that 99% of agencies and marketers employ, the branded Paid Search query receives all the attributed value for the sale.  The brand and TV teams pat themselves on the back for another $50M well spent and everyone goes home knowing that big dollars win in marketing.

But the truth is, without our customer originally seeing ‘Store A’ in the organic search results, they would never have known about ‘Store A’, and consequently would have never made that purchase.  If ‘Store A’ had not been undertaking dynamic creative remarketing with coupon codes, the customer may have opted to purchase their shoes at an alternate, cheaper store.  In the typical Last Click wins model neither of these channels would have received credit for a sale they “created” and consequently, may see reduced funding in the future in favour of branding activity and Paid Search.

That’s fine, but why do my figures not balance?

The issue with multiple parties producing reports from multiple sources is that they suffer from Channel Blindness (TM Pending).  Often, media buying teams can only see what occurs as a result of their media activity, for instance if the Paid Search team is reporting from AdWords or Last Click Analytics, they won’t even know that the Organic Click and Dynamic Creative Ad View occurred.  They will simply report on the conversion as they saw it, which was a transaction initiated by a paid click.

This problem compounds as the conversion funnel lengthens. Imagine if our customer had slept on their purchase for a week instead of a night, seen 100 (or 1000) display ads and clicked on them 2 or 3 times.  Now the display buying team would be seeing a conversion in their console as they are completely unaware of the interactions pre and post display click.  The headache for the marketing team is the single conversion could be reported on 2, 3, 4 or 10 times by different teams with a single channel view.

How do we better understand our marketing channels?

Luckily for us, there is a solution to our dilemma.  We need to start looking at a more complex attribution model that assigns some value in a more realistic fashion to each of our marketing channels in order to prioritise and allocate budget to them going forward.

After the simple Last Click model, we could look at First Click attribution.  In this model, instead of assigning all value to the final interaction we can assign value to the initial interaction.  After all, if our customer had not searched for ‘Nike Shoes’ and visited ‘Store A’, they would never have been likely to transact there.

In our opinion, a First Click model is a little more realistic than Last Click, it assigns value to the event that created the momentum towards a conversion, rather than the event that was a victim of that momentum.  But, using the First Click model still fails to take into account the importance of the Dynamic Creative impression that was a catalyst for the conversion based on price so we’re still missing a piece of the puzzle.

Now we start to move into tricky territory, where we allocate varying importance to events based on their type and proximity to the actual conversion, typically we’d use Time Decay or Linear Decay.

Time Decay essentially assigns more value to the interaction the closer in time the interaction is to the conversion.  Similarly, a Linear Decay model assigns increasing importance to an event the closer it is to the conversion, based on the number of interactions.

Ok, so which model do I choose?

That is the million dollar question!  There is no correct model for every business or vertical.  For businesses with long sales cycles, like travel, a very different model is needed from those in fast moving industries like fashion.

As a top line rule, always think about how you can better understand your customers’ interaction with your business.  For those using a Last Click model, take a look at First Click.   For those that have graduated to a First Click model, think about a Linear Model.  For those that are looking for a really granular, personalised model, start having a conversation with an analytics agency.

What’s the point of all this?

Why do we need to understand attribution? There are a whole host of reasons.  The question you could ask is:

If I had an extra dollar to spend, where would I put it?

By understanding the value of each of your channels, you can start to think about planning from that data – if you know Channel X has a 105% return and Channel Y has 150% return, you’d want to invest your dollar where you will see the most incremental revenue, across the entire customer lifetime and conversion cycle and not just the last click.


This chart shows indexed data comparing a First Click and Last Click model, aggregated by Medium.  You will see that when we use a First Click model, the value of the direct channel falls greatly and the value of the Organic and Paid Search Mediums increase dramatically.  In fact, in this example, the value of Paid Search increases by 42%.

By moving from a Last to First Click model, an organisation like the one above could make the assumption that they have been under estimating the value of Paid and Organic Search and potentially over investing in Brand.  By reallocating budgets to those channels which ‘introduce’ customers to the company could increase their total customer acquisitions for the same investment and increase organisational profitability.

Which model are you using and is it telling you the whole story? Our team can help with your attribution modelling so contact us today for further information.

No comments yet.

Leave a Reply

  • badges-women-top10-logo
  • AwardsBadge#2
  • GACP resize
  • Yahoo Certified
  • Adwords Qualified Company
We've joined Performics logo

We've joined the global performance marketing network Performics. Visit our new website today

You have Successfully Subscribed!