Measurement & Navigating Data in Digital Marketing

Truth is much too complicated to allow anything but approximations.

– John von Neumann

In the fast-paced world of digital marketing, brand success hinges on more than just data—it depends on making smart, timely decisions that connect with customers.

As a marketer, we often act on measurable and available data. Still, these figures, at best, hide critical nuances that affect long-term growth and, at worst, can be downright misleading.

In this chapter, we deep dive into the pitfalls of measurement, the biases lurking in our data, and the strategies to navigate uncertainty.

Introduction: The Human Element in Data-Driven Marketing

In an era dominated by digital analytics, it’s easy to assume that every marketing decision can be guided purely by numbers. Click-through rates, conversion rates, and attribution models fill our dashboards, giving the illusion of complete control. However, at its core, marketing is about people—their emotions, perceptions, and behaviours. While data remains indispensable, it must be balanced with qualitative insights to ensure brands connect meaningfully with their audience.

This requires moving beyond traditional digital metrics and embracing a more holistic approach to measurement. Marketing success is no longer just about immediate conversions, but about a deeper understanding of the interplay of various platforms that lead up to a customer action.

(Fig. 1)

During World War II, Britain suffered heavy aircraft losses, with the Royal Air Force (RAF) losing over 1,000 bombers in 1942 alone during nighttime raids over Germany. Studies showed that less than 50% of bombers returned from some high-risk missions, prompting urgent investigations into the causes of these losses. Damage to the planes (represented by red dots above) were extensively analysed to identify optimum approaches to minimise losses.

The Air Marshal’s Dilemma: How Survivorship Bias Leads Us Astray

During World War II, air marshals analysed returning fighter planes to determine which areas needed reinforcement. The bullet-ridden wings and fuselage seemed like apparent targets for additional armour (Fig 1). However, the real insight lay in what was missing: the planes that never returned had likely suffered fatal hits to the engines or cockpit. By focusing only on surviving aircraft, the military risked reinforcing the wrong areas.

Survivorship bias is not limited to military strategy—it’s everywhere. We tend to study success stories while ignoring failures, leading to distorted conclusions. Businesses celebrate companies that followed a particular growth strategy but forget those that tried the same and failed. Investors idolise entrepreneurs who took massive risks and won but don’t account for the thousands who took similar risks and lost everything. In marketing, we act on “available data”. A campaign with high conversion rates may seem like a model to replicate, but without analysing underperforming channels or disengaged audiences, we risk drawing incomplete conclusions. The key is to broaden measurement frameworks to include both successes and failures.

The Traps of Marketing Measurement

The Last-Click Trap: Mistaking the Finish Line for the Race

Survivorship bias in marketing measurement often manifests in an over-reliance on last-click attribution models (or some variant that attributes it the most credit for a conversion). These models only measure what is visible—successful conversions closest to the point of purchase (i.e. easiest to measure)—without accounting for the unseen, like abandoned customer journeys or brand interactions that don’t immediately result in a sale. Just as the Air Marshal misjudged where reinforcements were needed, marketers who focus only on last-click or similar risk reinforcing the wrong tactics, leading to an underinvestment in top-of-funnel initiatives.

By optimising for last-click-based results, brands inadvertently ignore the role of earlier touchpoints. This bias skewers budgets toward bottom-of-the-funnel activities while neglecting awareness and consideration-stage marketing that cultivates future customers. A more balanced approach acknowledges that every touchpoint contributes to the outcome—even if it doesn’t show up in campaign attribution reports.

The Data Divide: What Siloed Metrics Don’t Tell You

Another common pitfall in marketing measurement is the tendency to evaluate performance within the confines of platform-specific attribution models. Each digital platform operates within its measurement silo, using proprietary algorithms and reporting structures. These models often ignore cross-platform interactions, leading marketers to overestimate the impact of one channel while underestimating the role of others.

For example, a platform like Meta might report a high ROAS for retargeting campaigns, seemingly justifying continued budget allocation. However, without considering how users were initially introduced to the brand—perhaps through a Google search, an influencer mention, or a YouTube top-funnel campaign—marketers risk attributing too much success to the final interaction.

The Illusion of Success: Why Some Metrics Mislead

Marketers often assume that higher ROAS means better efficiency, but yet again, these metrics only reflect what is seen—not necessarily what matters most.

For instance:

• A YouTube ad campaign may appear inefficient in the short term, yet drive new high-value customers into the funnel. If marketers optimise too soon, they risk shutting down efforts that would have paid off with patience.
• A sudden improvement in ROAS might not be sustainable if not accompanied by improvements in acquisition costs as well.
• Advertising only on point on-sale channels, and trying new ad inventories there might provide great dashboard results. Still, those same users might be reached on other channels at a lesser cost improving overall CAC.

“All models are wrong, but some are useful. – George E. P. Box”

Not all past successes translate into future wins. The conditions that made a strategy work before may have shifted, and relying solely on past playbooks can lead to missed opportunities. Instead of assuming that what worked in one context will work again, marketers can take a first-principles approach—breaking down challenges to their fundamental truths and rebuilding solutions from the ground up.

Analysing the Equation

The equation captures how blended ROAS shifts from one snapshot to another, focusing on a single business cycle (i.e. before repeat purchases of newly acquired customers start contributing). Once repeat customers re-enter the equation, ROAS will gradually stabilise toward its original level in the long term, depending upon the CAC Recovery Duration (estimated as LTV/CAC).

It consists of two forward-looking terms, one historical term, and one assumed invariant term:

1. Forward-Looking Terms:

• Budget Increment (IB): A variable under the brand’s control, determining how aggressively the brand scales.
• CAC Penalty (p): A factor partially outside the brand’s control, influenced by market competition, audience saturation, and platform dynamics.

2. Historical Term:

• New-to-Brand Contribution Rate (NTBR): Reflects how much of the current customer base consists of first-time buyers.

3. Invariant Terms (Assumed):

• The ratio of New Customer AOV to Overall Customer AOV:
  $$AOV_N/AOV_O$$
• This remains stable if cross-sell and up-sell paths are well-defined and if there are no significant disruptions in product strategy.

By combining NTB contribution and the AOV ratio, we can infer that when a more significant share of LTV is locked into future cycles and the more the brand relies on these ‘future’ cycles in terms of revenue at present, the short-term impact on ROAS during scaling becomes more pronounced. In other words, brands with a higher reliance on repeat purchases will see a sharper ROAS drop in the short term before stabilising over time.

The same factors that influence short-term RoAS impact—such as NTB contribution, CAC penalty, and AOV differences—also determine how smoothly a brand can scale across its life-stages (Fig. 2 & 3) and whether specific industries face greater constraints when expanding.

We see a clear trend: as a brand transitions from the High Growth stage to the Decline stage, the possibility for efficient scaling declines significantly. Thus, the stage of brand should also be factored in when planning growth.

In strategic decision-making, a marketer should ask how the initiatives are likely to affect core brand health metrics –

• Are my optimisations improving RoAS due to reduction in CAC or is it because my New to Brand Customers are downsized?
• Is the cost of repeat customers increasing? Resulting in my overall RoAS drop, or am I paying higher for reaching new customer cohorts?

Such questions push us to think beyond the obvious and have clarity on who is buying my products and why.

Why Over-Optimization Can Backfire

When a brand shifts its budget toward high-ROI campaigns, which are high-ROI because they are very close to the point-of-sale and vice-versa while cutting early touchpoint investments, it creates a self-reinforcing loop — ROI improves because early touchpoint campaigns are no longer driving as many new customers as before, however, this means in the future there will be fewer repeats. With fewer repeats ROI drops yet again, and if it is below the acceptable ROI, necessitating another round of cuts. This vicious cycle causes brands to lose momentum over time as the supply of new potential customers dwindles.

Therefore, when faced with situations, for example, where marketing efficiency has to be improved, we might ask ourselves:

• What is the big picture? – What are all the variables at play? Which of these are the core variables? Which of these do I have the most control over, and where do I not?
• Control on which variable improves my brand health and which does not? – Improvements in CAC can indicate better brand health while the reduction in the New-to-brand rate typically does not (especially if it is below category average)
• What are the potential secondary consequences? – Will I have a more significant impact than now on RoAS when I reattempt to scale (significantly if NTBR is scaled down)? Is it coming at the cost of scale?

Relying on past performance and benchmarks can be helpful, however, removing the bias of precedent by uncovering the fundamentals ensures decisions are derived ground-up utilising all available opportunities rather than merely imitating past methodologies. This approach encourages adaptability, allowing brands to build strategies that hold up even as market conditions evolve.

The ‘When’ and ‘Where’

Accrual Method of Marketing

The concept of user journey is one of the most fundamental and critical ideas in digital marketing. However, it doesn’t always get incorporated into many media plans. For example, a paid campaign where the median conversion takes 80 days (due to a lengthy consumer journey) will appear ineffective if measured within a 30-day window (in a monthly report – and if we do measure it, it won’t be a unified cause-effect measurement. Similarly, if we take a very long

window (e.g. 1 year), we will end up overestimating the efficacy of campaigns.

Moreover, the user journey is a network of overlapping platform user journeys (Fig. 4). This interplay of channels heavily influences how long a user takes to traverse the platform as well as their overall journeys. Hence, marketers have to take special care to ensure that current measurement methods and tools do not skip source & re-engagement channels.

New Context Unlocks New Opportunities: The Layered Nature of Market Expansion

Many of the most significant market breakthroughs aren’t immediately apparent—they are discovered by unlocking one door, which reveals another behind it. Opportunities that once seemed impractical or unprofitable often become viable only after solving an adjacent problem.

Take Swiggy, Zomato and the evolution of quick commerce, for example:

• Zomato’s dominance in food delivery wasn’t just about aggregating restaurants—it gave them deep operational insight into consumer needs, logistics, and real-time demand fluctuations.
• This context uniquely positioned them to see why hyperlocal quick grocery fulfilment could work where traditional e-commerce models assumed it wouldn’t.
• Without first mastering restaurant delivery, they wouldn’t have had the data, infrastructure, or intuition to recognise the viability of quick commerce.

Similarly, Swiggy Instamart and Zepto didn’t just emerge from consumer impatience—they emerged from growing confidence in the feasibility of sub-30-minute fulfilment, which was only validated through years of experimentation and learning.

This highlights a critical point: Certain insights are inaccessible until the right pre-conditions are met. We can’t always see the second opportunity until we’ve unlocked the first. Brands that move quickly test early and think beyond the immediate market often find opportunities that others don’t even realise exist.

Similarly, many crucial insights in digital marketing are unlocked through a series of experimentations in communication, targeting, product mix and media, unlocking new growth opportunities (Fig. 5).

Summary

The reality, however, is that in organisations – the Growth head, Brand head, and Marketing head are three different people or sometimes one to two people who wear these hats and devise marketing plans. In both cases, there is always pressure to deliver immediate results, and hence, parking success for future returns is not possible. At the same time, immediate results that severely impact future returns should also be called out. In the end, it all depends on the size of the brand, the product, the stage of lifecycle, the size of marketing budgets and the overall risk appetite of the company. As teams we should be clear about how our present efficiency measure can impact our future demand. Also, while scaling, we need to be clear on what is a RoAS drop threshold that is acceptable as a trade-off for scaling. There is no “one” correct answer, but we urge marketers to be cautious of all answers before deciding on any course of action. It should be perfect to list down things we don’t know and data we are not sure of before starting a campaign.

Data-driven marketing is not about “knowing all data” but also about knowing its limitations. Marketing measurement is an evolving discipline; by being humble and receptive, we can stay ahead and gain a long-term competitive advantage

Case Study

Proving cross-channel influence and conversion delay in furniture category

The Challenge:

“Are my efforts to drive website revenue, showing up largely in our retail stores, Amazon, and Flipkart because these other distribution channels are booming?”

Outcome:

Of all customers influenced by Google Ads, over 75% buy from brand direct (75% vs 50% $\rightarrow$ 50% Higher), over $\sim$20% buy from Amazon & 80% of sales are completed by 13th Week.

---

Impact

Channel	Brand-Web	Brand-Store	Amazon	Flipkart	Pepperfry
(%) Sale	49%	27%	21%	3%	0.1%