Neil Moree

Instacart not having negative keywords as an option is the best example of what "machine learning" actually costs brands to support. The best solution I've found? Take their auto applied keywords and make their bid the lowest possible (30 cents). I built a really solid ads structure supporting each of our Target sku's and woke up to us bidding on "Ben and Jerry's" as a "optimized auto applied suggesition" for our Vanilla Ice Cream Whey. Gee, I wonder what sophisticated dev team decided "auto bid on product title". Yeah, in the world of CPG there could neeevvverrr be issues with that. Who leads this team and how much do you make. (Hint: too much). Happy Friday to everyone except overpaid leadership that don't actually understand what they're building and get paid boat loads of money for it.

11

4 Comments

Bonus point for the F*ck Survey Fraud shirt, Roddy! Good challenge to the industry about what we have done to create the current situation we are in where data quality is still an issue. First, he unpacked the MRX monster: - constrained budgets - pricing pressure - disengaged participants - survey farms and more sophisticated organized fraud - more bots (non-human) “participants” - AI being used for open-ends The answer is that instead of focusing on CLEANING the data we should be getting focused on it being real, representative and reliable data by not allowing fraud to even get into the survey. It’s a commitment to addressing the root of the problem and not spending money just mopping things up. There’s the dtect differentiation for sure. He had some really great figures to understand the actual hard and soft costs of fraud. Why do we find this totally acceptable??? Also fun scratch ticket below. Great presentation. At Quirk's Media here in New York. Little Bird Marketing #quirks #mrx

27

5 Comments

If we all collectively comment on Google Analytics posts/topics “cool that’s great but where’s the Annotations?” maybe we could have a *checks notes* BASIC FUNCTIONALITY NEED by end of year. UA had this and I’m still blown away that #GA4 launched without it… and still has yet to do it. Salesforce integration? That’s cool. Let’s have some basic functionality back that benefits everyone… not just a corporate partnership where we pat ourselves on the back and call it a better product. #analytics #marketing

4

Ever wonder why some brands can charge a premium while others get commoditized? Our new Blueprint for Brand Growth report sheds light on just that. Our analysis of 6.5 billion attitudinal and shopper data points found that brands that reach “meaningfully different” status, addressing a spectrum of practical and emotional needs in unique ways that competitors don't, have much greater pricing power, higher brand equity and much faster growth. Learn more about how brands can extend their identity and become more differentiated in this @MediaPost article: https://lnkd.in/gHRSmHgw

5

What trends have you seen in Direct Mail over the last few years? Do you still use it in your marketing mix? Let's have a chat about what it takes to continue the use of this craft - and check out some simple but creative campaigns I've found.

7

1 Comment

⚠ALL of your old Google Analytics data is disappearing on Monday! 😱 It will be gone for good, forever, and you won't be able to access it. Here's how you can make sure you don't lose all of your historical data - just a few clicks. I'll post a link to the template in the first comment! #googleanalytics #universalanalytics #ga4

13

7 Comments

The Forrester B2B Marketing and Sales Data Strategy Model and Process have been redesigned to better meet the demands of B2B #DataQuality improvement while your business remains in motion. Changing the tires on a moving car...and other impossible things.

8

1 Comment

I hear a lot about bots and bad respondents in market research. I'm a big proponent of taking respondent quality seriously because it can impact certain types of studies systematically (e.g., segmentations). We made an automated bad respondent detector that looks for internal consistency in survey responses. The last study we used it on it only detected only 1 completely random respondent (e.g., a bot answering completely randomly). But it still detected 364 bad responses (20% of the data) using a more sophisticated method. We detect based on a combination of low internal consistency in responses and low variation. (All of this is in addition to traditional detection methods like speeding.) For example, someone answering 2s and 3s for everything with no discernable pattern would get caught. This is respondent behaviour that would avoid traditional algorithms for bad data quality. I think either our field partners are catching all the random bots, OR the bots are "smart enough" to answer like lazy humans who use subsets of scales and anchoring. The 20% bad respondents are also likely to be many real humans who aren’t paying much attention. This figure may be conservative, but you have to tweak detectors to be the right balance between false positives (throwing away real respondents) vs. false negatives (not throwing away a bot). But it’s interesting that >99% of what we’re catching is due to behaviour that would avoid less sophisticated methods. Either humans evolved from straightlining to minimal (but random) variation or bots have evolved from fully random variation to minimal (but random) variation. (Our tool uses a bunch of past datasets, both real and synthetically created, extracts features that are markers of internal consistency, and uses a neural network on top of all that to make a final prediction of good/bad respondent. Internal consistency is a relative measure and is derived automatically study to study based on available data. We trained the model to differentiate between high quality real data and randomly created data to match known and expected patterns of bad responses.) Dig Insights, Upsiide, by Dig Insights

49

10 Comments

Forbes Article: 5 Priorities for Customer Personalization Through AI by Martin Zwilling I am continuing to find articles about how companies are using AI as part of their CX strategy. This article looks at the book, “Personalized: Customer Strategy in the Age of AI” (Abraham & Edelman) The final point ties it all together: “Make the whole experience memorable.” The Forbes author points out that the entire experience needs to feel magical to the customer.

2

Generative AI can collect responses and analyze them more quickly and efficiently, to glean takeaways quickly and accurately. Move past the “wow” phase and start asking “how can I use this to grow my business?” https://lnkd.in/gnunHz7N Visit our Insights to Activate #GenAI portal to 🎞 Hear how (April 17 webinar) 📚 Read how (Point of View) ❤️ https://lnkd.in/gWhKPjve #insightstoactivate

1

A compelling argument for the importance of experiments and the use of Bayesian modeling. The objection/notion of isolated, walled garden experiments is slowly becoming obsolete. As advertisers increasingly leverage open-source or proprietary geo/user-based experimentation on their own internal metrics, the need for incrementality testing as the gold standard for establishing media causality becomes undeniable. Incorporating experimental data into a Bayesian framework should enable more accurate and actionable insights in this regard.

13

Is Bing intentionally monetizing organic map listings in search? For a large #retail client’s  #affiliatemarketing program, we uncovered an anomaly in the Bing Map SERP (search engine results page) that turned a seemingly benign search into an affiliate-sponsored result. We’ve outlined some findings here: https://lnkd.in/dQr_DqiU #SEM #affiliatemarketing #marketingROI #searchengines

12

Good points here. (A "gatekeeper" company is going to get a lot of unwanted attention if they make it really easy to turn off every competitor's cross-context user tracking while leaving their own cross-context user tracking turned on. Whatever form the cookie preference ends up taking, it's going to also have to control Google tracking that isn't technically cookie-based but would otherwise be unfairly/illegally advantaged)

1

1 Comment

GA4 demographic and interest data isn't even directionally accurate. I work with a lot of brands that make decisions based off of the "best data they have available." It sounds like sage advice, except for when you confuse "best" with "available." Just because something is reported doesn't mean it is accurate, or useful. We have clients who frequently look at the information available in GA4 and make IMPORTANT business decisions like: • Customer strategy • Product strategy • Marketing strategy • Logistics strategy It's tempting, I get it. Google is just so gosh darn convincing. And look, to be fair, a good chunk of the time their data is close enough. But, it's very hard to know if your business is one of the ones collecting a bunch of nonsense. I'll give you a recent example: We had a client with GA4 data that suggested: • Their average age was 25-34 • Heavily skewed male (65%) • Interested in technology, video games, etc We then plugged in their Shopify first party data into nova -- where we actually used their encrypted PII to append verified credit card and demographic details -- and you know what we found? Their actual customer was: • 55+ • Slightly skewed female (55%) • Focused on family, vacations, coupons This made sense with their product, which intuitively seemed to cater more to a retiree focused on their family. This caused us to significantly change our messaging, creative, media, and experiments we were running on CRO. Why is GA4 so frequently off? Outside of the handful of people who actually tell Google the truth when they get those little YouTube popup surveys, the vast majority of their data comes from probabilistic modeling. You searched for knitting so you love knitting. You looked at colleges so you're college age. You watch monster trucks getting crushed so you're male. Etc. For bigger brands that cater to more of the mass market and have huge amounts of data are less affected. But if you're a sub $10m/yr brand, especially one that caters to a specific niche? Fair warning. #strategy #marketing #persona #ga4

107

17 Comments

Have any of you seen an online experimentation case study that can pass statistical scrutiny? 10+ years and I’m yet to see one. - Insufficient data - Cherry picking metrics - Flat out misleading conclusions - Post test “massaging” of the numbers If you’ve seen one published publicly please share it. #experimentation #abtesting #cro #optimization

13

7 Comments

Hot Take: With cookie deprecation coming (at some point), identity resolution will prove increasingly difficult via programmatic desktop and mobile due to the fact users are rarely signed into websites. Streaming (along with commerce) will actually be the easiest way to reach a target addressable audience since you need to be logged in (for the most part) in order to watch the same Office episode for the 500th time (whatever streamer the Office is on now). This can put Amazon Ads in a unique position to win.

19

While this is a neat feature, I wish the GA team would fix some of the more basic problems such as: - Let me click on the month to select the entire month, instead of having to select the first and last day - Let me select monthly as a time slice, instead of limiting me to daily or weekly - Give me back regex search - Allow me to copy and paste from the display instead of forcing me to right click + copy, or download csv. These (and many other issues) is why I'm reliant upon custom Explore and Looker Studio reports. What bugs you about GA? https://lnkd.in/ei6nmHvU

3

I'm seeing a ton of cases lately where organizations are noticing that Landing Page is showing up as "not set" in their GA data. There are a few common reasons for this to occur, but one reason in particular has stood out to me lately. Apparently it's common for organizations to set up timer-based events to measure "engagement" in some increment of time. For example, 15/30/60/90 second "engagement" events being fired off to GA. These events aren't directly tied to any particular user action; they're just firing after a page loads, dependent purely on elapsed time. I'm seeing this cause rampant "not set" issues, and massive session inflation. Here's why: GA4 naturally lists Landing Page as "not set" for sessions that don't have any page_view events. And these kinds of timer-based engagement events are inadvertently creating large volumes of sessions with no page_view events. Consider this scenario. 1. Users are browsing the site 2. Their sessions time out due to inactivity 3. Later, the user does something to trigger events to start firing again (for example, they return to the browser tab that the website was on) 4. Events are dispatched to GA4 without a page_view event happening. These events have to have a session, so a new session is instantiated. I'm not sure how valuable these kinds of events really are, but at a minimum, organizations should be aware of the potential unintended consequences, which include: - Wonky-looking data your stakeholders will notice and question - Serious overcounting of sessions - Major skewing of session-based metrics (e.g. artificially deflated conversion rates) Check your Landing Page data and see if you have a non-trivial amount of "not set" showing up! h/t Todd Bullivant for nudging me down this rabbit hole.

25

4 Comments