[The Marketing Analytics Intersect, by Avinash Kaushik] [1]
TMAI #480: KEY KPI: COST PER INCREMENTAL SALE.
[ Web Version [2]Â ]
A super simple, super impactful, lesson from our team.
The power of being thoughtful in the choice of a KPI, by accounting
for psychology/culture.
QUICK RECAP: SMART SOLVING INCREMENTALITY.
Iâve emphasized that an effective approach for building a culture
powered by incrementality is to break the problem down into three
impactful layers:
1. CHANNEL-SILO INCREMENTALITY.
_What if I turned off all my Google Brand PPC to zero? What is the
optimal investment allocation across ASC Value, ASC Volume, DABA on
Meta? Yada, yada, yada._
You are trying to improve the incremental impact from a single
channel.
The most common methodology is to use CHANNEL LIFT STUDIES (CLS). Most
digital platforms have it built-in, and provide for free!* For your
offline spend, TV, OOH, Radio, you will have to construct CLS â a
little more work, so worth it.
(*at least to their medium and large customers.)
2. XSTACK INCREMENTALITY.
_When we spend $10 million on YouTube, do we need to spend $40 million
on Television? Can billboards build frequency, while Iâm spending on
TikTok Consideration ads, to achieve a bigger brand lift? Yada, yada,
yada._
You are trying to improve incrementality across channels. It is
harder, and more rewarding.
The most common methodology is to use LARGE SCALE EXPERIMENTS (LSE).
No platform provides it for free, sadly. You, your agency, sets up
experiments and does the analysis. You can also sign up with a
specialized provider like Measured [3] - theyâve built a whole
exciting LSE platform.
3. PORTFOLIO INCREMENTALITY.
_What if we fired everyone in the Marketing department? Compared
apples-to-apples-to-apples, whatâs the incrementality of Google
Search, Meta, Radio, SEO, and SMS?_
Real hard stuff.
You are trying to understand overall incrementality delivered at a
company level by Marketing. You also get the bonus points from me for
identifying incrementality by channels and tactics (for massive
spenders on Mkt). You will earn my love if you build predictive media
budget allocations on top of all that to power truly
incrementality-centric Mkt!
The most common methodology is to build a MARKETING-MIX MODEL (MMM).
Ideally one that uses machine learning algorithms to understand the
ground truth, and measures incrementality as a portfolio (vs. silos,
then âpulled togetherâ as many MMMs do).
In a lovely twist of life, LSEs now help you _Trust, but Verify_ your
MMMs! They help identify how wrong your MMM is (it will always be
wrong, just a matter of degree). Even when 83% accurate, your MMM is
Godâs gift to your Marketing practice. Just make sure it is 83. đ
CONVERSION LIFT STUDIES (CLS) FTW.
I find teams forget the altitude at which MMMs operate (very, very
high). Then they make this critical mistake: They donât realize most
tactic level improvements will actually come from the most accessible
tool⌠Channel Lift Studies (CLS).
So, if your budget is big enough please earn my true love by building
a high quality MMM. BUT, first ensure you have a robust practice of
CLS across all digital channels â at the very minimum.
IF you spend lots offline, great, it is time to build offline channel
silo studies to prove the awesome (/no) incrementality of those
channels.
Building MMMs without CLS is like spending all your money on building
the Starship body, without any focus on the 33 Raptor Engines. High
expense, to go nowhere.
KEY KPI FOCUS: CPIS.
Every CLS is an experiment.
You form a hypothesis.
Your platform provider, say Snapchat, has CLS built in. They will take
your hypothesis, identify the optimal audience for it, distribute them
into test and control groups, and run the experiment for you.
[Note: Test & Control groups don't have to be 50:50. Depending on a
number of factors, it can even be 90:10 or, if you are doing something
crazy, 10:90.]
Here's an example of a Channel-silo hypothesis CLS:
Your ads are delivered to the test group. Performance data (sales,
brand lift) is collected across both test and control groups.
When they (Snapchat, Google, Meta, Tmall) have collected enough data
to reach statistical significance of at least 0.9 (ideally 0.95, and
NEVER lower than 0.9), they will present the Channel-silo
Incrementality results.
This is a typical way an Ad Platform will present CLS results:
This does not look so bad. You made a little bit of money. 1.02 ROAS.
But. These are terrible results. It just does not _feel _that way.
You spent $1 to earn an incremental $1.02 in _JUST_ media costs.
You did not include Agency fees, creative costs, and the cost of goods
sold (the product you sold did not fall off a truck). Oh, and you did
not include employee costs. And, there are hidden opportunity costs
from doing this vs. other things.
See... Terrible results.
[Bonus read: TMAI #455: Minimum Acceptable ROAS? 8. [4]]
I want that reality to sink in. Hence, a little psychological trick.
Snapchat, and other platforms, will usually provide you with one more
view of the CLS results that I want you to focus on more.
This one:
Note: Each conversion might have purchased more than one product. For
the sake of simplicity here, Iâm going to assume each order just
sold one product.
Armed with this lovely view... Ignore the ROAS, the Incremental
Revenue, because at some deep gut level it does not hit as much as it
should to result in a change of Snapchat strategy/tactics.
FOCUS ON THE COST PER INCREMENTAL SALE. THE $227.
Any Cost KPI often causes leaders to lean-in just a bit more.
It gets them to ask:_Â Hey, what was the Average Order Value of those
two hundred odd incremental orders where CPiS was $227?_
That is an excellent question. Did you sell a $90 product at a media
cost of $227?
If it had turned out that I each product I sold had an Average Price
after Discounts and Promotions of $1,500⌠OMG, I love that the
Snapchat Advertising Cost Per Sale was a tiny little $227. Joy.
Happiness. More budget to Snapchat to find the point of diminishing
returns!!
But, $227 to sell $110 worth?
Whatever the answer, the discussion will hit a little deeper than
saying_Â incremental revenue was $53k_ or saying_ the ROAS was 1.02._
Focus on CPiS. Ignore ROAS.
It is a much more psychologically potent KPI.
It will raise good questions, it will facilitate insightful
conversations, it will set higher standards for your internal
Marketing, external Agency, and Ad Platform.
Thatâs how Marketing becomes a competitive advantage.
THE RISK FROM A FOCUS ON COST.
A focus on efficiency can come at the cost of effectiveness.
Sometimes, you might want to lose some or a lot of money if you can
make a lot of revenue. Say, you are one of these AI start-ups who have
been handed billions of VC dollars, and growth is all that matters.
Be aware of this scenario. Check if it applies to you. If it does,
spend where you can get the most incremental revenue / signups. Costs
be dammed.
If you don't fit in this scenario... Focus on CPiS. From your CLS,
LSE, & MMMs.
BOTTOM LINE.
Culture eats strategy for breakfast.
Hence, my nudge to focus on understanding the psychology of
persuasion. In this instance, powered by using the right KPI.
One of the most amazing things about Conversion-Lift Studies is that
they donât need to take a massive amount of time, and the analysis
is relatively simple. That means, they power quicker tactical changes
to your channel-level tactics â and they do it on the basis of
incrementality.
Carpe diem.
-Avinash.
PS:Â Oh, it goes without saying the most powerful thing you are doing
is not running the experiments⌠It is making decisions, taking
actions! Analysis without actions is a perfectly built Starship that
never flies â a heartbreaking waste of time and resources.
Thank you for being a TMAI Premium subscriber - and helping raise
money for charity.
Your Premium subscription covers one person. It's fine to forward
occasionally. Please do not forward or pipe it into Slack for the
whole company. We have group plans, just email me.
[Subscribe [5]] Â | [Web Version [2]] Â | [Unsubscribe [6]]
[7]
[8]
[9]
Š2022 ZQ Insights |  PO Box 10193, San Jose, CA 95157, USA
Links:
------
[1] link
[2] link
[3] link
[4] link
[5] link
[6] link
[7] link
[8] link
[9] link