E-commerce marketers will always be the richest of us in their access to data but, as 2021 progresses, their ability to properly evaluate decisions is increasingly being eroded.

It’s mainly because of a problem with existing evaluation methods: multi-touch attribution is being undermined as ever more people and platforms are blocking the third-party cookies that make it possible. Last-click is more resilient, but most now agree that its findings on payback are unreliable.

It’s also because of an opportunity: e-commerce businesses had a good year in 2020. With money in the bank and soaring ambitions, many are exploring big brand strategies whose effects can’t be tracked.

The obvious place to look for a new analytics solution is in what brick-and-mortar businesses do. They’ve never been able to track people’s journeys from communications to sales, but they evaluate and optimise all the same, many using market mix modelling.

Market mix can – with some important adaptations – absolutely help. Plentiful and well organised data in e-commerce businesses makes it possible to standardise and automate for faster and cheaper projects. Teams like mine, with experience in e-commerce, have new, more appropriate, modelling techniques too.

The problem: Cookies are crumbling

It’s an axiom in forecasting that technology evolves fast, societies more slowly. It takes time for governments, institutions and markets to evolve a response when tech delivers a capability that we just don’t want.

And we – as consumers and individuals – don’t want marketers to track our actions. The minority that believes tracking is a fair trade for the benefits that online platforms bring is vastly outweighed by the majority who find it worrying.

Over time, the use of products that block tracking – like the brave browser – has continued to grow. Now, to protect their market share and comply with regulation, Google and Apple are also blocking it, or planning to soon.

In a world with far fewer third-party cookies, the remaining tracking options are limited to using last-click or paying for a “clean room.” Both of these have significant problems.

Making budget decisions on the basis of last-click is – to use a football analogy I borrowed from effectiveness expert Iain Noakes – like choosing a whole team based on who scores the goals. He’s right when he says that a team full of Alan Shearer and Harry Kane might be interesting, but it wouldn’t be effective.

Clean rooms, on the other hand, do work. They’re secure locations where confidential data from different platforms can be shared and analysed alongside advertisers’ sales data.

The problem is that they’re expensive and awkward. Analysts have to work away from home and be searched on the way in and out. According to a 2020 report by Gartner their use is growing, but mainly amongst marketers with a $1bn+ budget.

The opportunity: Investing into un-trackable advertising

With people stuck indoors during COVID-19 lockdowns, 2020 was a good year for e-commerce. In the UK in May, nearly £1 out of every £3 of retail sales was spent online, and even during the summer when shops were open it was £1 in every £4.

Source: ONS

The choice to invest some of last year’s extra profit into un-trackable advertising is not clear cut. Most digital-first marketers are used to communications that can be tried and tested in a small way before committing more budget. TV and other un-trackable investments have higher production costs and media investments are lumpier. They feel risky.

On the other hand, the evidence that TV works for brands in this situation is mounting up. Tom Roach recently summarised what we know on the subject in his article Scaling up without screwing up, along with advice on how to get the first step into TV right. It’s a convincing case.

And e-commerce brands are trying it. According to Thinkbox and BARB, nearly 70 e-commerce brands undertook their first foray into TV between April and September 2020 alone.

These businesses quite rightly have an evaluate-everything culture, but even though it’s the go-to technique for traditional advertisers, many are not aware of market mix modelling. Others have only come across versions of the technique that aren’t appropriate for them.

Adapting market mix for e-commerce businesses

“There does seem to be a big gulf between the cost and time of econometrics versus a smart data scientist that can take half a dozen datasets and look at some relationships. I think there’s a lot that clients can learn from more agile analysis.”

The quote above is from Simon Wilden, who is partner at Goodstuff, a media agency that works with lots of scale ups. His point of view is a common one. Conventional market mix modelling takes a long time, and it’s expensive. E-commerce businesses that have heard of it don’t understand why it’s so unwieldy. They have data scientists in-house and already undertake a wide range of complex analysis.

Although market mix is not trivial to pick up and get right, they’ve got a point. The kind of market mix modelling that's evolved for evaluating offline-only campaigns for big brick-and-mortar businesses isn't an exact fit for digital first brands.

Some elements of the traditional approach – like experienced human beings understanding the business being modelled and ensuring the analysis reflects the real world – do need to be kept.

Others, however, – like data collection, preparation, and exploration – need not be so cumbersome for businesses that have their data in order. They can be standardised and automated.

My team and I have developed and tested these, and other innovations, while working closely with e-commerce clients on a number of projects. We’ve built techniques for joining the dots between market mix and the methods that e-commerce businesses are used to.

We’re convinced that the adaptations work and that in 2021 and onwards, market mix will restore e-commerce marketers to their status as the best-informed of us all.