Вы находитесь на странице: 1из 8

Multi-Touch Attribution by Dr.

Josh
Bradley-Viso
This paper has been written by Dr. Josh Bradley-Viso to help B2C and B2B marketing
managers understand the links between marketing spend and ROI. If you are a marketeer with
budget responsibilities and your performance KPIs are linked to ROI or a sales funnel then this
paper is for you. This paper does not cover top-of-funnel ecommerce click attribution or UTM
click attribution as these models are considered too basic in the context of the multi touch ROI
attribution environment.
By reading this paper you will gain an understanding of ROI attribution and which models are
available in Sweet Analytics.

What is multi touch ROI Attribution?


The million-dollar question!

Ask this question to any marketeer and they will answer it based on their own intuition and not
on their empirical knowledge. It is critical to understand how and why campaigns are delivering
results and how to use this data to continue building on that success.

To know where to invest you need to know where your ROI is coming from, how it contributes to
your overall revenue and track it against different RFM (Recency, Frequency and Monetary
Value) Segments, Geo-demographic Information and Product Information. You also need to
understand your average Funnel Conversion Velocity​®​ for each of these segments. Without this
information you are only guessing where to invest more marketing budget in your next
campaign.

Multi-touch ROI Attribution tools like Sweet Analytics​™​ analyse every touchpoint or interaction
generated by every marketing campaign between the first and last interaction and assigns ROI
and the conversion credit to the channel that delivered the most impact or engagement. In other
words, if a prospect is exposed to 6 radio ads and then lands on your website from a Google
search, the conversion credit should be assigned to Radio not Google which is exactly what
doesn’t happen when you use inflexible models like Google Analytics or attribution tools from
the platforms that are delivering your web traffic.
Multi touch attribution answers how much ROI, Revenue and Lead Generation was
generated by the investment made on advertising. It delivers the clarity we need and
provides us with a reliable model on which we can base our future decisions.

What is Sweet Analytics?


Sweet Analytics is a platform with a powerful tracker that measures prospect engagement from
the first time they interact with your company. Most platforms only start measuring after a
website landing interaction as this is the only data they have access to. Sweet Analytics goes
beyond that and measures each engagement including Impressions on 3rd Party websites,
Radio Ads listed and exposure to TV ads and offline advertising. By gaining visibility of each
engagement we are able to attribute ROI to the channel that truly deserves that credit.

Here is an example of a purchase being attributed by Sweet Analytics vs Google Analytics:

In this example we have 4 channels


taking part in the conversion. Each
channel records a different level of
engagement based on the prospect’s
behaviour, green for most engaged and
red for less engaged.

We can clearly see that the conversion


credit on a last click attribution belongs
to Direct, but what about the other
channels? Use a linear attribution
model and 25% of the conversion credit
is assigned to each channel but this is
also wrong!

Sweet Analytics would assign the credit


to Display (based on the budget used
and engagement recorded) but would
also reward part of the ROI to Digital
Radio and Paid Search. This is a fairer model where each channel gets some credit and where
the advertiser knows that although Direct was the last channel, the channels that generated the
interest are duly credited for finding the prospect and delivering them to the funnel.
Why Common Attribution Models are wrong
In order to understand multi touch attribution we first need to understand why common
attribution models are wrong. But, before we get started, consider the following scenario which I
am going to refer to as I examine each model.

Peter Rubio from Apple saw your company’s Display Advertising Campaign promoting the
conference you were sponsoring. This made Peter register and visit your company’s stand at
the conference. Peter’s badge was scanned and a lead was recorded in your CRM. 24 Hours
later an email was sent to thank Peter for visiting your stand and he clicked on a link to see a
video on your website, signed up for a webinar and downloaded an ebook. Eventually, after a
long time Peter converted and signed up to your product generating an income of €20,000

SINGLE TOUCH ATTRIBUTION (FIRST INTERACTION)

Revenue and ROI are attributed to the first


interaction. All other interactions are ignored.
Peter’s first interaction - seeing your company’s
display advertising ads - receives 100% credit
for the entire deal (€20K)
Revenue is attributed to the last interaction
prior to the conversion - downloading your
eBook - and receives 100% credit for the €20K
deal.

Whilst being the most common, single touch models place disproportionate emphasis on either
lead generation (first interaction) or conversion (last interaction) activity.

The biggest problem with single-touch attribution models is that they overstate the value of a
single activity, and undervalue everything else. For instance, if you have a campaign that
consistently occurs midway through the buyers’ journey (e.g. webinar), it won’t be given any
credit with single-interaction. In reality, your mid-funnel programming may be a critical step in
progressing prospects through the funnel. As a result, marketeers may be tricked into
inappropriately allocating investments, cutting spend in mid-funnel programs because
single-touch attribution doesn’t provide visibility into that portion of the buyers’ journey.
LINEAR ATTRIBUTION OR EVENLY-WEIGHTED ATTRIBUTION

Revenue credit for all of Peter’s interactions is


split evenly among all campaigns (6 campaigns
all receive credit for €3,333, totalling €20K).

This delivers an improvement on single touch.


While this model applies credit to all
interactions, it runs the risk of overvaluing
lower-engagement touches, which leads to
faulty assumptions about marketing results. For
example: A click on an email receives the
same credit as watching and participating in a
webinar.

TIME DECAY ATTRIBUTION

Most revenue credit is given for interactions


that occur closer to the conversion event.

Peter’s last interaction closest to signing up for


the deal -downloading the eBook- receives
most of the credit. All prior interactions receive
less and less the farther back in the history
they are from the conversion event. This is not
fair for the channels that generated the interest
and made the customer aware as without them
there would be no customer to speak of.

POSITION BASED ATTRIBUTION


Greater revenue credit is given to specific touches in the cycle, typically the first and last
touches.

Peter’s first and last interactions receive the majority of credit (€16K total), while the remainder
of credit (€4K) is divided among mid-funnel activity.

Mid-funnel activity receives little credit and it’s not fair as these channels were responsible for
keeping the customer warm.

INTERACTION BASED ATTRIBUTION


Because you’ve historically seen that customers
who’ve engaged with Videos and eBooks are
more likely to convert, more revenue is attributed
to these events than display, email or a webinar.

A custom interaction-based attribution model


relies on historical analysis to apply different
weights to varying interactions. The biggest
danger with interaction-based models is that they
can often be subjective, and marketers must put
considerable thought into which types of user
behaviors are most valuable.
A.I. MARKETING ATTRIBUTION - THE SWEET
ANALYTICS WAY
Thanks to the Sweet Analytics Tracker we know exactly which customers saw an online or TV
ad. The tracker records impressions, views and clicks from all sources which then are
aggregated and anonymised in order to comply with GDPR Regulations. Whilst television and
print media have traditionally offered broad demographic information about potential influence,
Sweet Analytics can extrapolate that information, enrich it and match it to our own internal
databases.

For the past 3 years, we have been developing highly accurate predictive models that capture
the complexity of human behavior, analyse the impact of many touchpoints, and identify which
marketing activities most influence a sale.

Using automated machine learning we have been able to build self-optimising algorithms that
attribute marketing conversions in seconds not hours, feeding the model and improving
accuracy as new conversions continue to come in.

By automating many of the skills traditionally applied only by data scientists we have been able
to teach our algorithms to consider all valid statistical data points and to look for the nuisances
of human behavior that contribute to a conversion. We call this Engagement, a very simple word
that consists of 50 different calculations and complex "what-if" analyses that quantify the
effectiveness of different kinds of marketing activities and different combinations of marketing
touchpoints.

How do we do it?

To begin we determine a baseline -the sales that would naturally occur without any marketing
activity. We use use an algorithm to analyse the impact on sales if we remove all the marketing
touchpoints.

We then determine the marketing contribution to sales. This is the difference between actual
sales and the calculated baseline sales. The more effective your marketing activities, the more
sales are boosted above this baseline.

Finally we assign a contribution for each touchpoint. Our technology then performs a variety of
what-if calculations to determine the impact on sales if you remove one, or multiple touchpoints.
By using historical touchpoints and outcomes, Sweet Analytics automatically finds patterns,
creating a model that predicts sales depending on the touchpoints that apply to each prospect.
The more data we gather the more intelligent the core becomes.

Вам также может понравиться