What do you get from Google Ads? I asked Google can or do “predictive analytics improve business performance” and gathered up the ads it returned. From a buyer’s perspective, these results aren’t good. Sellers could do much better as well.

This article consists of five parts:

    1. Raw data
    2. Analysis
    3. Actionable recommendations.
    4. Mission and upcoming research
    5. Appendix: Deeper dive
      1. Google Ads
      2. Sellers’ and Buyers’ needs
      3. Predictive analytics market

1. Raw data

WARNING: Don’t be intimidated! There’s a lot here. Read the first several rows of the thirty-row table, then skip to the Analysis section, and eventually, you’ll want to go back and read more details. The table below includes all ads Google presented to me in two separate forays with the search engine. There’s no cherry-picking here. This is the full set.

Table 1 contains the twenty-four advertisers, embedded hyperlinks to the ads, brief comments, and links to any third-party content referenced within the ad.

Table 1. Advertisements returned by Google Search for the string “predictive analytics improve business performance” on 24 February and 2 March 2022.

Advertiser

(Ad hyperlinked; 3rdparty references cited in ads)

Noted content

Notes

No = No mention of predictive analytics.

Spec = Specialist, not generalist.

1.     Alteryx“Predictive analytics made practical” Brochure has much handwaving, e.g., becoming data-driven, competing on data and analytics. The referenced aging works (McKinsey’s 2018 paper on modeling the Impact of AI and the 2019 NewVantage adoption survey) do not strengthen their arguments.
McKinseyThis paper is out of date. Most of its sources are more than four years old. Its assumptions turn out to be overly optimistic, and the details of its models are inaccessible.Referenced by Alteryx
NewVantage PartnersSignificant flaws in this opaque survey-based research. Spending findings distort. E.g., (a) Claim: 96% of firms invested in AI/Machine Learning. Reality: adoption may be closer to one-eighth that level. With 74% of respondents from financial services, the sample fails to represent the broader business population. It’s also overweighted to largest firms. (b): Claim: in 2019, 21% of firms invested over $500 million on Big Data/AI, almost double the 2018 rate of 13%. Reality: Metrics are too imprecise to determine actual spending change and its components. Hence, doubtful face validity. 77% of respondents said “business adoption” of big data and AI initiatives continues to be a challenge. This suggests an alternative answer: focus on the most significant business impacts by industry and detailed use-cases instead of flogging the need for “data-driven organizations” and “data culture.”Referenced by Alteryx
2.     AmperityAnalytics for marketers – Reviewed both ad page and white paper. Their content covers a “data science for marketers” set of features and capabilities, but their quantitative results (e.g., “$20M incremental revenue from 1st party audience strategies” and “Reduced CPM by 36% driving improvements in media efficiencies”) are unsubstantiated and hence of dubious value.Spec
3.     Black Wood SevenHolistic Marketing Measurement & Optimization. Claims it will increase return on marketing investments via AI-driven key business insights. The platform predicts return on marketing spend and advises on the selection of optimum marketing mix.Spec
4.     Business Intelligence
Market
Huge cross-product feature checklist. Includes suitability by company size (Small, Medium, and Large)
5.     Causal LensEarly-stage. Claims better models that go beyond predictive correlations to implementing or augmenting your decisions, adapting to change, being explainable, and fair.  No-code Enterprise AI platform for better business decisions.
6.     coresignalClaims: Their external data makes your big data bigger, more insightful. They offer additional, external data to augment your existing data and integrate it into your analysis. E.g., employee reviews, geographic data, job postings, resumes. Assisting investors and businesses (recruiting, lead generation.) The external data industry (not just coresignal) accounted for $1.6 billion of spending in 2020, growing to 17.3 billion in 2027.No, Spec
7.     Cribl.ioInnovative and customizable controls to route observability data to where it has the most value. Five trends and predictions regarding observability data.Spec
8.     D&BRevTech, the B2B account-based marketing platform from D&B, helps you grow revenue with unified 1st and 3rd party data, targeted audiences, and personalized activations across channels and tools. Plus, results tracking. All powered by our industry-leading customer-data-platform & without trapping you in a tech silo.Spec
9.     DomoSparse return page, highlighting four capabilities: Get to insights faster, Accelerate discovery and decision-making, Improve data prep efficiency, and Ask questions about your data with natural language queries.
10.   Faraday.AIPrediction cloud. “With a step-by-step guide to creating AI-driven marketing capabilities in-house. Twelve use-cases covered (four featured.) It includes rich consumer profile data on virtually every U.S. adult — almost 300 million. At Faraday, we only predict consumer behavior. After nine years in business, we know which predictions are universally valuable.”Spec
11.   First InsightRetail predictive analytics. Strong unsubstantiated ROI claims. Will perform data collection to help drive decision-making.Spec
12.   IBMThe download answers the question of what’s new (before and after) with predictive analytics. It identifies the business functions that are investing the most in this area. The results on page 9 and onward aren’t meaningfully quantified.
13.   Intuitive Data AnalyticsFeature heavy page with some personable capabilities. The gap between features and capabilities versus business results is enormous (so too for many others on this list.)
14.   LogilitySupply chain planning.Spec
15.   Micro StrategyFrom a short, product-centric page, they offer the download of “MicroStrategy 2021.” To summarize, the report focuses on achieving universal access to real-time analytics with an open architecture and an enterprise scalable platform. They sourced their core business impact claim that firms “with CEOs who encourage data-driven decision-making are 77% more likely to significantly exceed their business goals”, from the Deloitte article below.No
DeloitteThe article “Analytics and Al-driven enterprises thrive in the Age of With” identifies business process improvements and understanding & improving customer experience as the top use-cases for analytics. Their inspirational claim that “Insight driven cultures deliver better business performance” should motivate very little direct buyer interest. Their survey is opaque, and there is no way to determine how well the sample represents the target population. Needs more concrete mixed with the sand.Referenced by MicroStrategy
No
16.   MIT Sloan SchoolA special online, self-paced course for executives to enable them to drive digital transformation in their business and stay ahead of disruptive technology by building a business strategy for the digital future. No quant measures of impact here, only an appeal to the authority of the professors whose content is used in the course materials.Spec
17.   OmniSci (now known as Heavy.AI)Pitching OmniSci as a Data Science resource, the whitepaper linked to their ad page (The Future of Data Science in a Remote Economy) dives into application scenarios that are more likely to inspire readers to explore how this might fit in their environment to create new, specific business outcomes (bravo!)
18.   PegaReal-time decisioning cornerstones: individualized, real-time, continuous: do it in milliseconds, at scale. Not a % or percent found. Pega cites dollars on the closing page: (a) A successful implementation of real-time next best action delivers more than $225 million USD in incremental sales and retention value for every 10 million customers per year. (b) For every month your business is not using a truly best-in-class next-best-action solution that is capable of “real, real-time” capabilities, you are likely costing your business in the order of $20 million for every 10 million customers that you have.” Disappointing that there’s no attribution or other details for those numbers.
19.   PrevedereCloud-based service provider of AI-based predictive analysis and modeling tools, models, external data, and related services.Spec
20.   QlikQlik: BI and Data Trends 2022. “Be Interwoven.” Also included for download; 2021 Top Data and Analytic Trends from Gartner. Both reports are vague, speculative feel-good discussions of all sorts of changes in the world of data and analytics with salutary descriptions of applications of their concepts. Footnotes abound in the Qlik report, but more detail on methodologies and data would go a long way to enhancing their credibility.
GartnerTop Trends in Data and Analytics for 2021, a Gartner research note accessible via Qlik.Referenced by Qlik
21.   RetinaPredict future value from day 1. Use your customer insights to improve targeting, conversion rates, and customer loyalty. Automate marketing operations and get data insights that drive value.Spec
22.   RSMAudit, tax, and consulting services provider. Ad provided a link to an RSM Guide to advanced data analytics. Offered business use-cases (focused on features and capabilities) and success stories (without quantitative claims or measures.)Spec
23.   SAPSAP Analytics Cloud is the analytics part of the broader SAP Business Technology Platform. BARC and Ventana Value Index Report linked in.
BARCLarge scale survey ratings, testimonial quotes. These ratings primarily reflect feature-groupings and usage metrics from the largest of customers. Unlike most other survey work referenced in these ad pages, readers can also find some data on places where SAP’s products don’t appear to be the best fit. E.g., “the product is focused more on trained business power users than the majority of employees.” Bi-survey.com.Referenced by SAP
Ventana Value Index ReportVentana Value Index Report, Analytics, and Data. Quality categories: product-experience related: Usability, Manageability, Reliability, Capability, and Adaptability. And customer-experience: Vendor Validation and TCO/ROI. Arguably the best piece of supporting research cited by any of these advertising pages. Unfortunately, their specific methods and detail ratings are opaque. This impacts the usefulness for buyers making actual investment decisions because those prospects’ needs boil down to specific applications for defined use-cases in particular industries and so on – but I know of no methods in active use that meet those requirements.Referenced by SAP
24.   TeradataCloud Database Management Systems for Analytical Use-cases. Content many times removed from “Can predictive analytics improve business performance?”No

2. Analysis

As I reviewed each ad, I looked for evidence of

  • Targeted seller-intent (why might the advertiser have delivered this information in their ad?)
  • How well the information returned aligned with the substance of the query.
  • The goodness-of-fit for a businessperson who might have posed the query (were their needs addressed?)
  • The ad’s information quality (data.)
  • The apparent target market for the ad (broad vs. narrow, high-risk vs. high-growth.)

At first glance

Four early, non-obvious observations: data irony, the land of ancient-learned-scholars, awful surveys, and consultant-distraction tactics.

Data Irony. What’s missing among most of these ads? Think of the dog that didn’t bark in Arthur Conan Doyle’s Sherlock Holmes mystery “The Adventure of Silver Blaze.”

Hard data is absent and unaccounted for! Most ads I looked at only use soft, qualitative assertions. Whatever happened to deeper quantitative evidence? Where is the hard evidence of business value? The irony’s strong here. Let me pound the obvious into this paragraph: Ads related to predictive analytics are very short on data.

The land of ancient-learned-scholars. Some ads want to seduce us into a trance with the mesmerizing rhetoric of the ancient-learned-scholars they’ve retained, gods who think more deeply and disseminate more wisdom than any of the mortal customers exposed to them. I metaphorically called them out in private marginal notes I wrote alongside many of the ads in this study: “another spurious, unsubstantiated speculative assertion,” “speculates that by 2025, some percentage of the market will change dramatically,” and “who cares?”

The work of ancient-learned-scholars isn’t based on rigorous analytical methods, and their authors don’t generally prioritize their insights in terms of business impact so, beyond entertainment value, ignore them.

Don’t let rock-paper-scissors or other frat-house-derived consensus methods delude you into thinking there’s a robust methodology there.

Awful surveys! In general, this set of ads and related material exemplifies survey worst-practices. They confuse correlation and causation; employ samples distorted by pre-selection and other biases (making the results ungeneralizable); document little (where, oh where have the survey instruments gone?); exhibit no sense of history (where are the literature reviews?) and seemingly have no hypotheses to test.

It doesn’t have to be so sloppy! I exposed some unreasonably great survey research in this note. Contact me if you want my analysis of some of the possibly questionable research on which your decisions might potentially rely.

Let’s document the race to the bottom: Challenge me to find more flaws than you do in the surveys you’re looking at in your investment decision-making process.

Consultant distraction tactics. Too many try to change the discussion to abstractions like “becoming data-driven,” “creating a digital culture,” and so forth. Those abstractions are typically the consequence of successful new investments. They’re results, not antecedents. Of course, it helps to have top-down support for new business initiatives. But don’t put the dogs behind the sled!

Four central business needs – Ignored

For buyers or end-users, compelling ads should address four central issues to help them make progress in their decision-making process:

    1. Where specifically are we likely to get maximum returns? In what use-cases and business processes, with what business models, and in what industries and economies are the most significant business results from predictive analytics achieved?
    2. Where are we unlikely to achieve these significant returns?
    3. How can we best manage risks to our specific business associated with decisions on predictive analytics?
    4. What is the hard, empirical evidence that I can trust?

Ads that do a great job on these four questions should drive qualified readers (and their influencers) to take aggressive action, far beyond downloading another tepid whitepaper.

In my sample of 24 ads served up by Google Search, sellers aren’t hitting those central business-buyer issues.

Most ads failed on all four central issues

Perhaps they’re fishing for a high-volume stream of low-quality leads they’ll leave to their sales and marketing staff to sort out.

Advertising critique

Here’s my opinion of what I gleaned from a review of the ad content (both initial pages, which appeared when I clicked on the hyperlink by the symbol “Ad” and third-party materials linked to those initial pages.)

  • Anyone trying to determine if predictive analytics improve business performance by reading these ads – and the 3rd party content associated with them – is likely to throw their hands up and walk away disappointed.
  • Technical features and specific capabilities abound! Business benefits and business impacts are largely missing. Why? Who do advertisers think their technical content appeals to, business buyers or techno-geeks? I’ve lost count of how often I’ve run across this “drunk searching for his car keys under the streetlight” type of marketing!
  • Use-cases outnumbered testimonials, but both are relatively uncompelling in driving buying decisions because most use-cases were not tied to specific business outcomes.
  • Quantifiable business outcomes were rare and unsubstantiated. Readers are left to do their own “thought experiments” to ask and answer for themselves how a feature or capability converts to a measurable business outcome to which the executive committee might respond positively.
  • Advertisers sprinkle unsubstantiated numbers through their materials to give them the appearance of substance. They cite sources without links, and when they provide links, the underlying research may be missing, the source is questionable, or the conclusions suspect.
  • Survey data, where presented, was often obscure and opaque. Indeed, hard data is hard to find. That which was offered often turns out to be out of date, methodologically suspect, and biased to favor sellers. (I’ll help you rip the mask off this class of problem. Try me. See section 4 below.)
  • Defective culture – the plaintive wail of too many consultants – can always be an issue. But cultural issues are too easy an excuse to raise when, deeper down, the suppliers of technology goods and services fail to clearly define the specific business conditions that deliver the best business benefit. I suspect those sellers fear specificity will make it harder to sell more broadly.
  • There are few, if any, discussions of shortcomings and conditions under which the use of predictive analytics isn’t appropriate.
  • If sellers conceal knowledge of which business conditions (e.g., industry, company size, process patterns, and geography) and business applications deliver weaker (or no) returns; if they encourage buyers to pursue less than ideal opportunities without warning them of the risks, they’re misleading and misdirecting their customer(s). If that’s happening, it’s unethical.

Brynjolfsson, Jin, and McElheran’s paper “The Power of Prediction: Predictive Analytics, Workplace Complements, and Business Performance” – cited in my note on unreasonably great research – surveyed over 30,000 American manufacturing establishments.  They nail it. They identify the conditions under which predictive analytics can deliver significant business benefits and the conditions where it doesn’t.

  • Voluminous trends predictions (over 20 and 30 pages in two different white papers) are awesome, fascinating in their technical and architectural complexity, but generally worthless for business people seeking business advantage. They’re intuitively suggestive but operationally non-directive. Often, they contain specific, quantitative predictions – of the form X% will do such and such by year Y – that they present without quantitative or qualitative justification. They fail to clearly answer the question, “which prediction is the most important for which buying constituency and under what conditions?”
  • On the other hand, sellers are convinced that offering these white papers to the marketplace will generate leads, so they pay for the right to collect reader data in exchange for giving out paid-reprint rights. It’s their advertising money, but it’s also your time. What do you really need to make sound business decisions?
  • Half of the advertisers were specialists, half generalists. Specialists focused on a particular role or industry, or function get closer than generalists to being specific on actionable business needs and impacts. Examples: predictive analytics for marketers and predictive analytics for retail sales. By offering “everything you could ever want,” generalists (such as SAP and IBM) may be more effective in keeping out specialists. But specialists may be more effective penetrating smaller or under-invested organizations.
  • Buyers are misdirected and cheated, and sellers are deluded. Vendor-cost-per raw lead is low, but the cost of qualifying these leads has to be high. For buyers, it’s worse.

3. Actionable recommendations

For buyers

  • Search for use-cases that are proven in your industry for enterprises of your size and so forth. Seek solid evidence. Then open the door to sellers.
  • Ask sellers to
    • Address the four central issues in section 2.
    • Align their use-cases, testimonials, success stories, and business proposals to their answers to the four central issues.
    • Filter their material against the advertising critique in section 2.
    • Attest that they’ve disclosed all significant limitations and risks that you should be aware of. Get that in writing.
  • Run material of the types referenced in this note by me for a quick take on issues I would have with that material.
  • Have you run across a novel claim on which you’d like an experienced, independent third party’s take? Let’s discuss.

Sellers

Get your ads, whitepapers, surveys, brochures and other sales materials critiqued pre-publication!

Use us!

  • First calls are free![1]
  • In-depth reviews are also available.
  • Discussions are confidential!

4. Mission and upcoming research

Quality & Impact Analysis Agenda

This research is part of a broader initiative aiming to fill two major vacuums: people need better advice on how to rate the quality and impact of information provided to them by sellers. And sellers need better advice on how to raise the objectively measured quality and impact of what they say to the market.

Beyond Ads, I am analyzing returns from other sources, such as Google Scholar.

I am also in the middle of a series of articles related to survey-based research. I’ve already completed four. The first,  https://thansyn.com/discover-unreasonably-great-research-and-exploit-it/, delivered a simple framework for unreasonably great research. The next three,  https://thansyn.com/organizations-and-people-in-the-unreasonably-great-research-framework/, https://thansyn.com/science-in-the-unreasonably-great-research-framework/ and https://thansyn.com/survey-design-execution-in-the-unreasonably-great-research-framework/, analyzed three components of the framework.  More are coming.

Contact me or any other Analyst Syndicate member who has published under “Quality & Impact Analysis.”

© 2022 – Tom Austin — All Rights Reserved.
This research reflects my personal opinion at the time of publication. I declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Update history: 14 March 2022. V2: Re-sequenced sections for improved readability.

Appendix: Deeper dive

  1. Google Ads
  2. Sellers’ and buyers’ needs
  3. Predictive analytics market

Google Ads

When you launch a Google Search, a very sophisticated engine[2] returns a page of results (the first of many it may have assembled for you.) Results consist of content links and page excerpts that the engine predicts best match your query. In addition, Google will append ads at the top and bottom of the results list. Ads look like search results, but they have the word “Ad” prepended to the first line of the sponsored result, as shown in Figure 1. There could be ten or more ads on a results page.

Figure 1. A Google Ad at work

Sellers buy Google Ad Words to get Google to surface their advertising content on your search results page. They’re asking Google to show their ad on any results page that matches their ad words. (In reality, it’s a bit richer. It’s an algorithmic, real-time auction system that supports the sellers’ desires to optimally expose their ads while controlling their ad spend and maximizing Google’s ad revenue.)

  1. Sellers’ and buyers needs

Ads help sellers:

  • Promote an identity, communicate and reinforce a vision and mission, establish an impression or market position. For example, in the 1950s, auto manufacturers directly subsidized automobile racing because they believed in the adage “win on Sunday, sell on Monday.”
  • Gain attention – either as an indirect way to bask in the glow of the work of others or just a naked attention grab. For example, the press ingests and rebroadcasts predictions at years’ end, so sellers sponsor and distribute predictions-white-papers and feed them to the media to promote.
  • Exploit the fear of missing out and of losing out — the two classic underlying themes of the Technology-Industry-Complex, FOMO, and FUD[3]
  • Increase awareness of the advertiser’s value proposition(s)
  • Collect information as in test-marketing concepts and testing marketing readiness
  • Feed the sales funnel: identify potential buyers, convert them to leads, and convert the leads to orders, landing and expanding from there.

As a corollary, in ads, the biggest, broadest players in a space usually rely on making the most generalized claims to fend off attacks by smaller, more specialized competitors who focus on developing new niches, unoccupied “white space” in the market[4].

Buyers?

Why do business people seek information from the Internet? What’s their intent?

Like sellers, buyers’ motives are many. But if a business person asks questions about “predictive analytics’ impact on business performance,” odds are that there’s a battle going on between confirmation bias (looking for proof their earlier decisions were right) and finding new information potentially of value to them.

  1. The predictive analytics market

This is more complicated than it might seem at first glance.

Predictive analytics is not a new technology category. One history[5] traces the term back to usage in 1689 by the Lloyd company to predict risk in various sea voyages and underwrite insurance accordingly. There are references to predictive analytics dating to the 1940s and 1950s. There are “white papers” published 15 years ago that measured the level of market awareness of predictive analytics[6].

Predictive analytics is an enormous topic. It includes vast swaths of technology, target markets, buyers, decision-makers and influencers, target users, business models, financial details, and so on. Technology categories differ in many ways, including maturity and rate and level of diffusion across the overall marketplace[7]. Sellers could be promoting horizontal analytics-related offerings or industry-specific ones. Products could be focused on enterprises with specific data platforms, and others could be “universal” offerings.

Simplify the target: High-risk buyers vs. high-growth buyers

Figure 2 illustrates how, in Roger’s model, innovations diffuse across the marketplace with Moore’s market “Chasm” superimposed. The blue line in the large diffusion chart is the level of diffusion per unit time (with time moving from left to right and typically taking many years overall.) It’s broken into five phases, from innovators to laggards. The yellow line represents total (cumulative) diffusion across time. The Chasm, as in “Crossing the Chasm[8],” is somewhere near the start of the “early majority” phase.

Figure 2. Everett Roger’s and Geoffrey Moore’s Models

Source: Author’s illustration of the Chasm superimposed on [File: Diffusion of ideas.svg | Diffusion_of_ideas] Public Domain per  https://upload.wikimedia.org/wikipedia/commons/thumb/1/11/Diffusion_of_ideas.svg/330px-Diffusion_of_ideas.svg.png

Let’s assume that there are two different potential target markets:

  • High-risk buyers, the innovators, and early adopters, the first 16 percent of the total market, a market that has not yet “crossed the chasm” and,
  • High-growth buyers, the “early majority,” the next 34 percent of the market, momentum buyers who appear once the market has crossed the Chasm.

Size versus specificity.

They’re negatively correlated. The bigger the firm and the bigger its installed base in the analytics segment they’re pursuing, the less specific their claims are. The smaller the firm and the smaller the installed base in the market segment, the more specific their claims. I trace this contrast to differences in buyer and seller targets. Sellers pursue either high-risk or high-growth buyers, and they focus on either conquest sales or retention and expansion sales.

Sales processes

High-risk. The sales process starts with features that appeal to the “geeks” of the discipline (in this case, analytics). These buyers are more willing to assume higher risk to accomplish the out of the ordinary. They may also include “conquest sales” to users of alternative technologies. Sales costs are high, feature-based competition reigns and repeat business is less likely because, by definition, the buyers are taking on extra risk. Testimonials tend to be more influential here.

Hi-growth. As sales and market penetration accelerates towards and past the Chasm, sales efforts shift to business people seeking specific business outcomes. These buyers are less likely to focus on features, more on results. Large, established sellers will try to maximize retention on top of “land-and-expand” successes. The cost of sales is lower, and lock-in grows. Advertising tends toward global strategy, partner ecosystem, platform business models, and downplays engaging publicly on “feature wars.”

Contact me or any other Analyst Syndicate member who has published under “Quality & Impact Analysis.”

© 2022 – Tom Austin — All Rights Reserved.
This research reflects my personal opinion at the time of publication. I declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Update history: 14 March 2022. V2: Re-sequenced sections for improved readability.

[1] I’m on calendly.com as https://www.calendly.com/tom-austin. Or look me up on https://thansyn.com/our-analysts/ to get email, LinkedIn and Twitter coordinates.

[2] See https://blog.google/products/search/introducing-mum/ and https://arxiv.org/pdf/2105.02274.pdf for a discussion of the evolving state of the art of search engines at Google.

[3] Technology-Industry-Complex, FUD and FOMO are analyzed here https://thansyn.com/confronted-with-too-much-tech-fud-and-fomo/

[4] See https://blog.hubspot.com/marketing/white-space-opportunity.

[5] See https://medium.com/@predictivesuccess/a-brief-history-of-predictive-analytics-f05a9e55145f

[6] See, for example, http://download.101com.com/pub/tdwi/files/pa_report_q107_f.pdf  (Unfortunately, there is no reconciliation of the sample used with the overall market so treat the quantitative data in this report with caution.)

[7] Everett Rogers’ Diffusion of Innovation, https://en.wikipedia.org/wiki/Diffusion_of_innovations

[8] Once the product has diffused into 50 percent of the market, growth slows.