When it comes to attributing success metrics, each channel typically has its own way of measuring its impact or performance.
These approaches are generally how we make our decisions on what media to buy:
- Any direct mail/email/tele comms would match the outbound with responses
- Broadcast media like TV & radio would look for an uplift in sales within that period
- Print would typically use offer codes to track response
For many years channels were largely considered independently of one another. Most companies sat in two camps:
- Camp 1 was the pure brand player, they typically would have to invest in TV & Radio and track the overall number of sales. Often these brands had physical stores, or sold through intermediaries – so had to rely on “uplift” or top down approaches.
- Camp 2 was the direct marketers, who were typically companies that dealt with the client, specifically cataglogues or up & coming ecommerce brands. For each transaction, they could usually see some activity prior to the sale or the use of a coupon to track media.
Over time, investors could see the power of the more trackable media, used by those in camp2 – the direct marketers. Rather than having to rely on hope & macro trends, we can now demonstrate statistically the return that marketing spend brings.
Two worlds colliding together
With the advance of technology & social media, these two camps are starting to come together.
When we look at categories like Beauty & cosmetics, we can see that many traditionally camp 1 – pure brand marketers are now starting to create their ‘buy direct’ ecommerce sites, or implement loyalty schemes.
Similarly, it has never been cheaper to create a company or brand. Rather than having to spend millions broadcasting to everyone, we can still create the engaging media that targets specific cohorts of people.
As brands start to invest in more cross channel activtity we start to ask the question; what is really driving marketing return?
Did we have an attribution problem before?
Whilst the problem has always been there, it was not quite a rife as it is today. However, we did have sensible pragmatic approaches to deal with it.
Taking a channel like cold direct mail, list providers would often provide duplicate names; and more often or not those “dupes” would be the most likely to buy. Companies would then choose how they were going to measure this:
- Randomly assign the contact to a provider; this essentailly splits out the share of the result amongst the lists bought
- Implement a hierachy, if you were a new list you had to prove that you could provide incremental names not just the same as everyone else
- Others would be more considered and think, if this person is a really good prospect, that I have their name 2-3+ times – I am going to mail them again
The approach a brand would use would often represent their philosphy or targets. Typically those randomly assigning value would be looking for longer term partnerships where they work together.
Those with a hierachy often needed quick growth, whilst the third option were the more entrepreneurial brands.
The problem is only going to get harder to deal with
Over the past 20 years we have seen a significant change in how people live. With the wider use of technology such as mobile phones, and ever more platforms to enable an easier life for consumers – or a way to improve communications between people.
This has also led to more ways than ever before to spending your marketing budget.
The rise of the number of ways to spend budget has led to a wider awareness of attribution.
If we look forward to the next 20 years, what will happen?
- There may be a consolidation of platforms, with companies like Meta buying up Whats app & Instagram, however this often still means that these are considered separate channels
- We may find new platforms coming in that promote privacy and reduced advertising
- New technology like virtual reality may create new channels
However I think the biggest change will be how an individual can influence marketing.
With the recent rise of brand ambassadors and influencers being used to promote products through organic channels, or approaches like ubers member-get-member; combined with constantly new ways of people to commicate with each other. The personal touch will be both more crowded and important than ever before.
The trouble with this word of mouth media is that it almost impossible to track with technology; due to privacy. Instead you need to get a blend of data & statistics to help unpick the impact.
Regardless of what is actually instore for us: With more ways to spend money, in hard channel spend and in the softer word of mouth approach, the problem will only get harder to solve.
You cannot trust media platforms to mark their own homework
When we look to the larger advertising networks, Google & Facebook, there has been a tightening of data controls.
In the case of Facebook, they have a “walled garden”, meaning that advertisers could not use independent technology to track impressions of their AD. Instead you were reliant on the facebook pixel.
For a long time, Facebook tried to get users to use their attribution tools; which would use their pixel data to evaluate marketing spend.
In a deep dive investigation with a clients tests, we managed to show that the 258 transactions claimed by Facebook were statistically impossible to achieve; instead the answer was closed to 28.
What we realised was happening, was that Facebook would use a linear attribution model and combine all their impression data with visit data to the site. This would be unfair on other channels like Google as their impression data would not be included.
Consider this example, both Google & Facebook have 1000 impressions (on their platforms) that lead to an order. In the same time there are 500 visits to the site, 100 from both channels; so they are performing exactly the same. In total say there were 100 orders (i.e. it took 5 visits to place an order)
A linear attribution approach it weights the assignment of orders to the channel equally, based on the number of visits before. In this case 100 out of 500 visits that led to an order would mean that Facebook contributed around 20% of all orders, so 20% of 100 is 20.
When we looked at Facebook attribution, they were combining their impresssions with visit data. This means that they would say:
(1000 impressions + 100 visits / 1000 total impressions + 500 total visits) = 73% of orders were contributed by Facebook. Or 73 orders.
20 vs 73 out of 100 orders is of a totally different result, but companies around the globe were relying on this metric which in reality is hugely inflated.
Shortly after we completed this study Facebook attribution was removed from the market.
So how did google fair?
Google is not in much of a better situation, having bought Adometry, the feedback from the market was generally that “it didn’t work”.
Since then under the guise of GDPR, there have been significant stirings that Google are removing cookies from browers. Which means that Google will have the most control over user data; and therefore performance tracking moving forward.
You only have to look at the google suite of products to see how by providing free tools, they are lining themselves up for:
- Firstly, easy to place and track marketing
- Secondly, more control over reporting and marketing than other platforms.
It is not just “big tech” either, we regularly see companies commiting significant spend to above the line brand building; without sufficient tracking. They typically rely on the media agencies, “in house” econometrics or brand tracking teams to measure performance.
However the CFO will soon ask the question – “if it has gone so well, why haven’t our revenues gone up?”
The only option is to get an independent view
What it comes down to is finding an independent party that you trust.
As a company, we have made several decisions along the way, such as not becoming media buyers; so that we can remain independent.
For a recent client, where we were asked to evaluate the three year growth plan, this enabled us free reign to really look into what was driving the business; and challenge the status quo.
In this instance we had to tell the client they needed a fundamental change in strategy to meet the growth plan. This was the clients CEO response:
“The results of the analysis constructively evaluated our plans, and it identified what could be done to achieve the growth we require.
Our board were able to digest the findings and had the information required to re-plan & reset with the new strategy.”
What to look for in your attribution partner
In 2016, I was asked to help choose an econometrician for a client (we weren’t allowed to do it as conflict of interest)
We visited four agencies; one was tied to their buying agency, one was tied to another, the third was an independent econometrician and the last was a more general analytics consultancy.
All the econometricians had the same 4 slides about; collecting weekly spends, getting external feeds like weather etc. Then they got to digital, and that’s when the fireworks came out. Each had invested huge amounts of time in finding different ways to model this digital spend; neural networks, nest models.
The general analytics consultancy, said it plainly; there is no standard modelling approach that will work out of the box. Instead to do it analytically you have to do a wide range of projects & integrate the findings together.
For us at Fusion Analytics , we have a similar ethos, but slightly different approach. We have tried to create the tools & process required to:
- Capture more data, consistently; linking online, offline & above the line
- Allow clients to input new findings from other areas of analytics
- Have a proces that allows multiple stakeholders to input and scenario plan
- Provide simple views of this enhanced data, before you do complex modelling
This mixes a blend of technology, statistics & consultancy to get the right fit for you.
When you are considering your attribution partner, you are not just buying a widget, it is important that you can get someone with an opinion on a range of topics including tech and statistics; otherwise they will only be painting half a picture.
Please do get in touch if we can help discuss attribution problems and solutions.