Thread by @darth_na

🧵
::: Attribution - is it really BS ? :::

I've seen a steady rise in tweets and discussions around #Attribution;
and many state/suggest that it's a load of crap.

🧵 Read More >>>

#MarketingTwitter #Marketing #DigitalMarketing

https://twitter.com/very_demanda/status/1449423140082176004

2/*

The majority (but not all!) come from "marketing",
rather than "digital marketing".
And, in many cases, it seems more aimed at things like DigitalAds or SMM, often looking at Last Click etc.

But, why are people saying it;
esp. if they are from "real" marketing?

>>>

3/*

Now, for most of us - the statement is actually true (-ish).
Few of us get to work with hard-attribution (physical tickets etc.).

That means most of us deal with incomplete data,
often with inaccuracies.

But that in no way makes attribution BS or useless!

>>>

4/*

So why the pushback/shit-slinging?

Multiple reasons:
1) Many in "digital" don't really understand attribution/statistics etc.
2) Far to many people take them as "gold"/fact
3) Some people in certain fields/roles (*stares at a % of DigiAds*) take a lot of credit!

>>>

5/*

If you look carefully, you will see that many of the tweets/posts are Not saying attribution is useless.
They aren't saying we can't use it.

But they are pointing out what just about everyone should know/understand;
Attribution is "wrong"

(as in - not 100% right)

>>>

6/*

A shining example of part of the problem,
I saw this yesterday:
@CountXero >

https://twitter.com/CountXero/status/1449756449765158920

Some people think GA is "correct".

The fact that it (and other trackers) may be missing a large % of traffic,
misreporting sources etc. does not occur to some users!

>>>

7/*

Similar issues occur with businesses/clients and Digital Ads (such as SERP PPC).
There are some practitioners/agencies that will include Brand searches/clicks/conversions,
taking credit - when clearly there were contributing factors (else why the brand search?)

>>>

8/*

(Not all Ads are like it, and it's not just Digital!)

Natural misunderstandings can occur via other channels, due to "last click" or "last reference" (if you ask customers at PoS, over the phone etc.).

>>>

9/*

The reality is, if you run multiple campaigns,
multiple channels,
have organic,
or even people talking about you -
your attribution data is incorrect to some degree!

But, for various reasons - people that deal with Digital seem to think the data is more accurate!?

>>>

10/*

TV, Radio, Print - all of these are extrapolated.
Everyone that deals with them knows the figures are "loose", and estimated based on sales/viewership etc.
70K people may have the TV on, but may not "see" the ad etc.

But Digital is viewed as more "concrete".

>>>

11/*

And really - it is Not!

Example of GA data:
Do you really think a "Scroll depth of 60%" means people read 60% of your content? Really?

{Narrator : They in fact, did not!}

So … people (including clients, new marketers, those not used to stats in digital etc.)…

>>>

12/*

... need to stop looking at their data and thinking it's "solid" or "complete".

Chances are,
* you are missing data,
* have partially polluted data,
* are misreading some of it,
* and the model (first/last/degrading etc.) isn't quite right either!

(Sorry, but true!)

>>>

13/*

So ... is Attribution BS?

Yes - kind of.

But it's useful, if used properly and it's limitations and flaws are understood.
Use models, and monitor/annotate and test properly, they are more useful and robust.

>>>

14/*

Make sure you can identify sources
(UTM parameters, custom/vanity URLs (even domain names!)),
Consider using 2 different models (LC is fine, so long as it's Not the only one!),
Run split/sub-test campaigns,
Use hold-out testing/cease sub-campaigns

>>>

15/*

Doing such things will permit you to see just how loose/flawed the data can be.

If you can use a contributory model (track multiple sources over time), you'll start to see how things like Soak and Association work and influence conversions.

>>>

16/*

And if you need to convince clients of this,
ask them what comes to mind when you say

"whassup?" "just chillin"
or
"soft, strong and very"
or even
"big bad dom"

(All of these still pop into my head when I consider certain products ... years after!)

>>>

17/*

Attribution may not be complete bullshit,
but in many cases, it is not neat, clean, tidy or complete/accurate either.

So start looking at your data with different eyes,
and make sure you don't step in something your shouldn't :D