This article was originally published on AdAge.com on Sept. 28, 2017.

Today's marketers are facing a paradox: We have better intelligence and tools for driving growth than ever before. Data-driven digital marketing, programmatic and audience buying are proven to be delivering results. But if return on investment (ROI) is going up, why are sales in decline?

There are some simple causes and effects: Media inflation continues to increase across all channels, while reach has declined for most every channel and publisher. Rising costs combined with declining reach drives up Cost-Per-Point (CPP), which means creative needs to be exponentially more effective to deliver the same results (and that's before attenuation of media attention is factored in). In most cases, even the best creative cannot offset the decline in efficiency. It's just math. And as the math suggests, in most cases we'd expect to see a decline in payback. Yet most channel-specific measurements show a positive ROI.

Why? A combination of factors is at play. Many ROI analyses today are fundamentally flawed. Some of the most common problems with today's analytic tools, or the deployment or interpretation of them, include:

• Myopia (The value-added study problem): Many analyses seek to evaluate a single channel against a hold-out sample and therefore over-attribute sales impact to that single channel.

• Misinterpretation (The user error problem): It's critical to know what you are measuring and evaluate results accordingly. Different types of metrics should not be compared: e.g., short-term versus long-term ROI, upper-funnel versus lower-funnel tactics, new penetration versus buy rate or marginal returns versus last dollar spent.

• Limitations in scope (The blind spot problem): Many “360” studies leave out difficult-to-measure factors, limiting their explanatory power. This includes issues like the exclusion of walled-garden data and its impact, inability to read non-addressable media, or omission of critical endogenous factors like competitive activity or consumer confidence.

• Monolithism (The one-size-fits-all problem): Many approaches read all targets/customer segments the same way, even though channel allocations and tactics are planned at the segment/target level. This may also be true at the market level, with different markets using different models, so no direct comparisons can be made.

• Inaccurate or flawed (The “It's just wrong” problem): Be wary of analyses that don't measure what they purport to measure — for example, return on advertising spend (ROAS) models that claim any sale in the payout model versus only incremental sales.

• Non-neutral (The vested interest problem): Watch out for analytics that aren't conducted by an independent third party, creating analytic biases that color results.

• Backward-looking only (The rearview mirror problem): Beware analyses that explain the past, but do not provide any forward-looking planning or decision-making capabilities, so the application of learnings is of limited use.

• Diminishing Returns Effect (The what-comes-up-must-come-down problem): ROI can rise while the rate of sales growth drops. After a certain investment level, marketing effectiveness declines. But that doesn't mean you should stop investing. Profits may still rise … just not as fast.

Even with strong analytic models, it's easy to misread results. In a marketing mix model, you may see switching between competitors as Brand A steals share from Brand B in Month 1, only to have Brand B win it back in Month 2. Although there is neither category growth nor topline growth for either brand in this model, both would count it as positive ROI. From the brand's point of view it actually won some sales, even though in reality it was a zero-sum game.

Before the tyranny of ROI, marketers understood the need to evaluate campaign payback alongside other metrics — brand health, penetration and switching analyses, elasticity, etc. — to get a sense of their overall brand performance. Today, the art of understanding the basket of measures required to manage a business is being lost as the science of quantifying financial impact has taken hold.

Marketing analytic acumen must become a requisite for marketing talent. It is no longer good enough to rely on a marketing analytics function to own measurement, without the deep understanding of their marketing counterparts. The only way to effectively grow a business is to learn what works and doesn't work. Learning requires feedback. Analytics must accurately and reliably provide that feedback to drive decision-making. And marketers must understand the core building blocks of advanced analytics.

This includes a comprehensive understanding of topics, such as:

Defining problems and choosing the right measurement metrics to answer them

Understanding attribution and what it means

Data essentials for measurement and analytics

Short-term versus long-term measurement

How to calculate marketing ROI

Connected analytics: Bridging offline and online

Marketing mix model and multi-touch attribution: top down versus bottom up, how to use each and how and why to connect them

The building blocks of data and modeling

A marketer's job is to profitably drive growth. If the business isn't growing faster than the competition, it is losing share. If it is not acquiring new customers, it is losing ground. And if it isn't doing it profitably, it is losing margin (and probably losing marketers their jobs). With better analytics, marketers will be able to build their businesses profitably, so that sales and ROI both go in the same direction.