The Twelve Errors of Mar-Tech

If you’ve been an avid reader, you might notice things look a little bit different around here.

Welcome to our new website! We hope you like it as much as we do.

It’s been an incredibly tough year, but we are lucky to serve our customers doing two things with you:

  1. Get your data to live up to it’s potential
  2. Evolve your mar-tech and empower your teams along the way.

Without further ado, let’s have a little fun. I present to you Bonsai’s first ever…

Twelve Errors of Mar-Tech

  1. Changing the tag, instead of changing the report
  2. Concatenation above common sense
  3. Mistaking keywords for queries
  4. Overbaking your third-party cookies
  5. Closing attribution-windows
  6. Stale user-interfaces over custom data pipelines
  7. Lost total cost
  8. Wasting time asking vendors impossible questions
  9. Making different metrics match
  10. Letting data intimidate you
  11. Using hand-me-down data instead of your own premium data
  12. Buying the all-in-one promise

1. Changing the tag, instead of changing the report

I think there’s probably nothing in the world of marketing technology that creates more bad data than a practitioner who unwittingly breaks an entire architecture just to see a data point they need. The worst part is that this data was already available – it was simply living in a different report.  Why does this happen? My theory is this: practitioners are trained by marketing technology salespeople that all of the problems they need to solve will be magically handled by their user interface. Practitioners give marketing technology far too much credit.  Standard UIs are pretty, but flawed: at the end of the day, any report you use that you did not explicitly design is simply a canned report. Sometimes, canned reports do not have the data you need.

But if you believe marketing technology is magic, and user interfaces are all-knowing, then the problem must be your data, right? So well-meaning practitioners figure out the workarounds – add the tracking tag to populate the data they want to see in the wrong canned report. They push their changes, see the new data points, feel relieved to answer their question, and ruin the data quality for most of the business reporting dependent on correct tag implementation in the process.

Stop breaking your tags. Tell your analysts what data you need to see to make decisions, let them design you the data pipeline and report you need, and solve your problems without the mess.

2. Concatenation above common sense

In the year 2020, I’ve had very serious practitioners, at very large institutions (not Google!), attempt to impress me with the quantity of keywords their business advertises against. “We manage millions of keywords here Matt, if you can even imagine that.” Sadly, in my worst fever dreams, I can. For those lost souls and others similarly afflicted, let me share two little secrets with you:

First: There is nothing sophisticated about an account with a lot of keywords. In fact, it’s a sign that you are hopelessly behind the times. It might even be an indication that you don’t know how digital search auction media actually works. Please, please, get out of 2006 and come join us in 2020. The most sophisticated accounts bid using every available lever – not just keywords – and do so with automation. Dynamic matching and feed-driven approaches improve the query-to-ad-relevance match at a level far superior to simple keyword expansions. More sophisticated advertisers leverage audience, location, time, and context alongside semantic intent to deliver results without all the waste.

Second: For executives impressed by million-keyword marketers, let me bring you in on a little secret formula called CONCAT(). If you show me an account with over 1 million keywords, I will show you an account with 800K keywords that have never once been matched to a search on Google. Just because you can use Excel, doesn’t mean you’ve built a high-quality paid search program.

3. Mistaking keywords for queries

Admitting that the entire world of SEO is to blame for this misunderstanding, but still. I cannot remember how many times I had this conversation:

“But what keyword did they search?” 

“Do you mean what query did they use?”

“What? No I want the keyword. Did they use [keyword] or “keyword”?”

“I don’t know, let me check. First, excuse me while I go scream into a pillow and break another bottle over my head.”

A keyword is a net an advertiser uses to catch potentially relevant search queries. Ads are shows on Search Engine Results Pages (SERPs) unique to the user’s specific query. When you want to know how you are doing for particular search queries, look at a query report. Stop downloading keyword reports that tell you absolutely nothing about how you are doing on a specific search term.

4. Overbaking Your Third Party Cookies

Most digital marketers remarket to site visitors using 20 different ad platforms – all at once. Most digital publishers not named Facebook, Google or Amazon show ads from as many as 90 of these third party platforms, all at once. Think you are safe just because you only use Facebook, Google, or Amazon? Think again. These platforms show ads using audience technologies from all three, as well as others. Want to know what happens when someone clicks on your ad from ad network A,  converts, and then goes back to their digital lives? They keep seeing your ad – served now only from ad network B. And ad network C, D, E all the way through Z. None of their third-party cookies are the wiser. When the best case scenario for most digital marketers is 96% ad impression waste, can we all agree that strategy is beyond repair? If you haven’t entertained a conversation about how to approach your digital marketing with a first-party dataset, please do so in 2021.

5. Closing Attribution Windows

Our most lauded blogpost of the year was the one where we pointed out that attribution windows take one single customer journey dataset, break it into two incorrect customer journey datasets, report out on two imaginary customer journeys, and ascribe causal impact to the imaginary customer data outcomes in an effort to make an inferior and implausible attribution model jive with basic common sense and intuition. If you use attribution windows in your marketing attribution, you’re marketing attribution is hopelessly incorrect. Measure as much as you can. Attribute using better causal models! Keeping customer journeys and common sense at the forefront. If you need help with this, our contact form is right here.

6. Stale user-interfaces over custom data pipelines

I don’t know why people still want to use this interface. The last time it meaningfully changed was 2012. Why aren’t you designing only the reports you need to run your business? That power – to build your own insight, and nothing more –  is literally the essence of cloud for marketing. Years ago, that capability wasn’t generally available, scalable, or cost effective. It is now. Stop learning platform UIs. Start learning platform APIs!

7. Lost Total Cost

I bet your company has 5 Ph.D.’s researching the best statistical methodology to calculate your firm’s Customer Lifetime Value. I also bet I can’t find 5 people who’ve ever even asked the question: “What’s our total acquisition cost – end to end – to acquire that customer?”

Here’s a few of the costs that go into acquiring customers through your marketing program:

Ad clicks (or impression cost).

The third-party audience collection cost.

The third-party demand side platform’s audience onboarding cost.

The ad serving cost.

The site clickthrough tracking & subsequent conversion tracking cost.

The cloud storage & site analytics hit tracking cost.

Your reporting & analytics SAAS platform’s cost.

Not to mention your management fees if you use an agency(ies), as well as creative services costs, staffing, trafficking, tag management, data governance and customer consent management costs. This is even before we begin to quantify the compliance costs for data regulations in your region.

Does calculating the total cost of acquiring your next customer sound fun? It’s certainly not easy. But is it valuable? More so than you can even imagine. In 2021, if your organization commits to the simple act of intending to learn more about the total cost of acquiring customers, the exercise in and of itself will drive far more bottom line value to your business than predicting another decimal-point difference in your expected customer value.

8. Wasting time asking vendors impossible questions

“But are these conversions incremental?” I used to get this question all the time at my old job. Could I answer it? Within meaningful reason, sure. But the question was always an absurd one to ask someone who’s not privy to all of the data required to answer it. Underlying every one of those questions was a doubt — something wasn’t working as well as intended at the client. It might very well be something completely unrelated to your digital ad programs. But asking third-parties to solve problems they’ll never have the datasets to answer wastes everyone’s time.

9. Breaking data to make different metrics match

“Why don’t my clicks match site visits?”

“Why don’t my sales numbers in Chicago match my sales numbers in San Francisco?”

These are equivalently ridiculous questions. For some reason, we’ve all accepted the second one as such, but still give the first question the benefit of the doubt. I’ve seen entire organizations build tiger teams whose sole purpose was to concoct a way to alter their reporting definitions and processes so that things like “clicks” matched “visits” down to the decimal. Nevermind the fact that they are different things altogether. Here’s a thought: different metrics should be different. They shouldn’t be the same, or you wouldn’t need two metrics! Diagnosing broken data is critical, but observing different datapoints being different is not a symptom of broken data.

10. Letting data intimidate you

Here’s another little secret: I can’t do math in my head either. I routinely forget the definition of the core concepts of the things I’m an expert in. I am literally constantly Googling things to solve data science conundrums. Just jump in. Ask questions. Ask faster. Stop worrying. It’s honestly not that hard once you let go of worrying about being right.

11. Using hand-me-down data instead of your own premium data

You know what your mattress business needs? Someone else’s opinion of who’s in the market to buy a mattress.

You know what your real estate business needs? Someone else’s guess on who’s looking to sell their home.

You know what your fast-fashion clothing shop needs? Someone else’s dataset on fashionista fans.

The amount of marketers who ignore golden customer insights inside of the datasets they already own is mind-boggling. These same marketers are also addicted to buying worse insights from others. Your first-party customer behavior insights are your biggest asset. Commit to learning more from it.

12. Buying the all-in-one promise

Sporks. Universal TV remotes. Roller shoes. Jean shorts. What do they all have in common? They all offer you everything-in-one.

If your platform claims to do unify all of your marketing activities, I can assure you it will do everything badly. The simple fact is that you cannot accomplish everything in successful modern marketing with one technology. Here’s a short sample of the list of things you need to do to well for world-class modern marketing:

  • You must acquire customers at right moment in time, across thousands of digital platforms.
  • You must gain their consent, and serve them personalized offers and services without accidentally creeping them out.
  • You mustn’t waste money selling past the close.
  • You must unify your customer behavior insights across all of your marketing channels and touchpoints.
  • You must coordinate marketing messages across online, offline, onsite and out-of-home.
  • You must invest your next dollar in the marketing efforts most likely to increase incremental returns.
  • You must not undervalue your customers.
  • You must not overvalue your media channels and marketing efforts.
  • You must automate activities responsibly.
  • You must value and treasure your customer data, utilizing it to enhance their experience and perception of your brand.
  • You must adapt to their changing behavior without delay.

You cannot accomplish all of this through a third-party cookie.  No third party identity graph, device graph, or probabilistic identity map will solve all of this for you. You cannot do this by outsourcing the problem to a single tech giants’ identity space within their walled garden. For world-class modern marketing excellence, you must take ownership of your own business and customer information, commit to understanding the technologies required for your unique business model, and operate the systems integration that’s best fit for you. The only ‘everything-in-one’ solution is you.

I hope you enjoyed this silly list. Happy New Year everyone! Here’s to a roaring 2021.

About the Author

Matt Butler
Ex-Googler of 12 years, Matt was a founding member of Google’s analytical consulting team, developing analytics, statistical forecasting, auction modeling, and machine learning for companies such as Procter & Gamble, Coca-Cola, Unilever, Kohl’s, Best Buy, and many more. He went on to lead global technical partnerships before leaving to found Bonsai in February, 2020.