Measuring the halo effect: proof your D2C ads are selling on every shelf

The problem: your dashboard sees one shelf, your buyers use four

An Indian consumer who meets your brand in a Meta reel might buy on your site tonight, on Amazon this weekend when their Prime order goes out, or on Zepto next Tuesday at 11pm. Your ad platform credits only the first path. The other two register as "organic" marketplace growth — and in the next budget meeting, someone proposes cutting the Meta spend that caused them.

This is the halo effect: paid demand creation on the direct channel converting on shelves you cannot pixel. It is real, it is large in low-consideration categories, and it is measurable with tools you already have — no media-mix model required.

If you only measure the click, you will cut the very spend that is filling your other shelves.

Method 1: the geo split (the cleanest signal you can get for free)

Pick two sets of comparable cities. Run elevated D2C ad spend in one set, hold the other as control, keep everything else constant for six to eight weeks. Then compare quick-commerce sell-through and marketplace units city by city. Blinkit, Zepto and Instamart all report city or dark-store level sales; Amazon gives you pincode-ish signals through its reports. If the treated cities' off-site sales outgrow control cities, the delta divided by the extra spend is your halo ROAS.

India makes this unusually practical: quick commerce is city-granular, and metro demand pools are comparable enough (Pune vs Hyderabad, Jaipur vs Lucknow) to build honest controls.

Method 2: branded search as the leading indicator

Brand queries are the halo's early-warning system. Watch four series weekly: Google Search Console brand-term impressions, Amazon search-term report brand queries, your own site's direct traffic, and autocomplete presence for "your brand + category". D2C campaigns that are building memory move these within two to four weeks — before the marketplace sales lift lands. If spend rises and branded search does not, your creative is renting attention, not building memory; fix the creative before scaling further (the compounding mechanics are in the branded-search loop).

Method 3: the pause test (cheap, brutal, honest)

The reverse experiment. Pause D2C prospecting in one region — or, if you are brave, entirely — for three to four weeks, and watch what happens to "organic" marketplace and quick-commerce sales with a two-to-six-week lag. Brands that run this almost always discover the marketplace baseline was never a baseline; it was the echo of paid demand creation. Document the decay curve — it becomes your strongest budget-defence artefact and tells you your brand's memory half-life.

Method 4: cohort surveys and the maths of triangulation

One question at checkout and in post-purchase WhatsApp — "where did you first hear about us?" — costs nothing and catches paths no pixel sees. Layer the four methods and triangulate: geo splits give magnitude, branded search gives lead time, pause tests give causality, surveys give the story. You will not get one perfect number; you will get a defensible range, which is what budget decisions actually need.

A working rule from our client base: in impulse-friendly FMCG and F&B categories, total commerce lift typically lands at 1.3–1.8× what the D2C dashboard alone reports. Your multiplier is knowable — measure it once a year and use it in every planning cycle.

What to do with the answer

Once the halo is quantified, three decisions change. Budget: judge demand-creation spend on blended, all-channel CAC — not site-only ROAS. Channel negotiations: walking into a quick-commerce margin discussion with proof that your ads drive their category growth changes the conversation. And forecasting: marketplace projections become downstream of your ad plan instead of independent guesses. That is what running the flywheel as a system looks like — the full model is in the D2C flywheel.

Designing the geo split: a field checklist

The practical design, start to finish. Pick four to six cities you can pair honestly — similar population, similar current sales, similar quick-commerce penetration (pair Pune with Hyderabad, not with Patna). Randomise which side of each pair gets treatment. Lift D2C prospecting spend 50–100 percent in treated cities only, holding creative, offers and marketplace activity identical everywhere. Run six to eight weeks minimum — shorter windows drown in noise. Collect weekly: quick-commerce sell-through by city, marketplace units by region, branded search by geography where available, and your own site orders by pincode. Read the result as a difference-in-differences: (treated growth minus control growth) across off-site channels, valued at contribution margin, divided by the incremental spend. One honest caveat: stock-outs wreck the experiment — confirm dark-store inventory depth in treated cities before you start, or the test measures your supply chain instead of your brand.