Omnichannel Marketing: Who Needs to Build It

Being on Instagram, a website, a marketplace and a shop is multichannel. Nothing wrong with it. Omnichannel is harder: the person who messaged you on Tuesday is recognised at your counter on Saturday, by whoever is standing there.

The difference is not pedantry, because the two are bought differently. Multichannel is a question of which channels earn their place — our guide to marketing channels covers that. Omnichannel is a question of whether your systems can agree on who someone is. For many Indian SMBs multichannel done well is the right answer and omnichannel is not worth the cost yet.

Multichannel vs omnichannel, by what the customer experiences

Tell them apart by what the customer notices.

Notice what is absent from the right column: creative, messaging, tone. Keeping one message consistent everywhere is integrated marketing, a separate discipline. This page is the layer underneath — data and experience. You can be perfectly consistent across six channels and still fail every row above.

The four layers omnichannel actually requires

These are dependencies, not a menu. Almost every failed omnichannel programme we are called into failed at the first one.

1. One customer identity across touchpoints. A single key resolving the same human in your website orders, billing or till export, enquiry list and support tool.
2. Availability that is true everywhere. Stock, slot or appointment, accurate per location, refreshed often enough that the promise survives the drive.
3. Service history visible to whoever picks up. One timeline per customer, readable by the person talking to them.
4. Consistent pricing and promotions. One price source the site, the marketplace listing and the counter all read from.

Most buy layer three first, because support software demos well and identity work does not. Then the new timeline holds three records for one customer with no way to tell.

The breakages customers actually notice

Nobody complains about a missing identity layer — only about its symptoms.

• The price that differs in store. Not the discount — the surprise. Someone travelled on a number they saw on their phone.
• The agent who cannot see yesterday's chat. The surest way to cool a warm customer is making them repeat themselves to you.
• The abandoned-cart email for something already bought in store. It says you are not paying attention, right after they paid you.
• The "available" that was not. Google's local inventory specification is blunt: "Accurately submit the product's availability in-store and match the availability to what is in your stores."

Google Merchant Center Help, "Local product inventory data specification" — no last-updated date shown; read 1 October 2026.

What each layer costs — in systems and in discipline

The software bill is rarely the real cost. It is a daily habit that must hold after the project team leaves.

Sequence for early value: layer one first — the duplicate count alone changes decisions within weeks. Then the cheapest slice of layer three that stops customers repeating themselves. Layer two only where travel is involved; layer four last.

The test for whether you need omnichannel at all

One question decides it: do your customers move between channels within a single purchase? If they research, buy and get served in one place, there is nothing to stitch and an identity layer buys nothing.

How to answer it from data you already have

1. The overlap count. Export website orders, billing or till records and your enquiry list. Normalise every phone number to its last ten digits, then count the numbers appearing in two or more files.
2. The cross-channel return. Count online orders that came back through a store, and in-store purchases exchanged online. The cleanest evidence there is: the customer physically did it.
3. Search your support transcripts. Count tickets referencing another channel — "I ordered on the app", "I came to your Andheri store last week".
4. Check the shape of conversion. If nearly every purchase completes in one session on one device, the journey you are planning for does not exist. Our marketing analytics guide covers reading this honestly.

There is no published threshold, so set yours before you look. We treat roughly one in twenty purchases crossing a channel as the line below which the right fix is each channel working properly — a rule of thumb, not a benchmark.

What the overlap check usually shows (our own work)

Digital Hangover, first-hand: when we run that three-file overlap for retail and D2C clients, the share of customers appearing in more than one system is consistently smaller than the client expected. We publish no figure: it varies so widely by category that one number would mislead. The cause is almost never a missing platform — the counter never captured a number, or captured it into a field nobody exports.

Where it matters most and least in India

Retailers should read the store-and-phone seam first in our retail marketing guide: which cross-channel journeys produce your sales decides this. Mapping touchpoints is customer journey work; the groups worth treating apart come from customer segmentation.

A starting sequence when you cannot buy a CDP

Most do not need a customer data platform — just a key, two integrations, and the honesty to leave the rest manual.

Identity resolution with what you already have

In India the practical identity key is the mobile number: customers give it at a counter without hesitating, and payment and delivery flows already carry it.

• Normalise once, in one place. Strip +91, leading zeroes, spaces and hyphens; store the last ten digits as the key, keep the original.
• Do not auto-merge on the number alone. Pair it with one soft signal — name plus pincode, or email — before joining records that carry money. Households share numbers; counters mistype them.
• Treat the number as reassignable. In Rajeswari v. Union of India (Supreme Court of India, 30 October 2023) the court recorded TRAI's position that a deactivated or disconnected number "is not allocated to the new subscriber for at least a period of 90 days". A dormant record keyed on a recycled number can be a different person, so age those out.
• Keep your internal key and your analytics ID separate. Google's policy is that "no data be passed to Google that Google could use or recognize as personally identifiable information (PII)," which covers personal mobile numbers — so the number cannot be your Analytics User-ID.

Google Analytics Help, "Best practices to avoid sending PII" and "User-ID" — neither shows a last-updated date; both read 1 October 2026.

The two integrations worth doing first

1. Till or billing into the customer table your online orders land in, keyed on the normalised number. The software is the easy half; staff capturing the number on a Saturday evening is the hard one.
2. A read-only view of orders and past tickets wherever your team answers messages. Not a CRM rollout — one screen, so the person replying sees what happened before.

What to leave deliberately manual

• Cross-channel loyalty reconciliation. A weekly manual sweep is survivable; a half-built points engine that disagrees with itself is not.
• Returns and exchanges across channels. Document the process and give staff authority before automating it.
• High-value flagging. A human reading a short list weekly beats a model built on uncleaned data.

If you want this scoped against what your systems can support, that groundwork comes before any campaign plan — our services page sets out the sequence.

India specifics that break the clean omnichannel plan

Four realities make the vendor diagrams optimistic here. None is a reason not to build, but all change what you build.

• Phone-number identity is shared and reassignable. One number often serves a household, and numbers get recycled after the quarantine period above. Never let a merge silently combine two people's history.
• Cash on delivery breaks the clean online record. The order is digital, the payment is cash to a delivery partner, and there is no payment instrument to join on. Failed deliveries and returns land in a courier's system, not yours — the commercial side sits in our ecommerce marketing guide.
• Marketplace orders are not your customer records. The platform owns that relationship; you get a shipping address and a masked contact, not an identity you can join to your data. Decide how much demand you want where nobody is recognised twice.
• WhatsApp is where customers actually reply. Behaviour, not a recommendation: a question left unread in email gets answered there in minutes. So your most useful service history accumulates where your CRM cannot see it, often on a staff member's personal phone. Digital Hangover does not sell WhatsApp marketing — this is a data-capture gap, not a channel to buy.

A correct store record is the same job as hyperlocal marketing.

Joining customer data across systems is a consent question

Combining records collected for different reasons is what India's data law governs. The Digital Personal Data Protection Act, 2023 (assented 11 August 2023) ties consent to a stated purpose: section 6(1) provides that consent "shall signify an agreement to the processing of her personal data for the specified purpose and be limited to such personal data as is necessary," disclosed in the notice under section 5(1).

So a number taken at a till to send a bill is not automatically a marketing list, and section 8(7) requires erasure "as soon as it is reasonable to assume that the specified purpose is no longer being served." Write the purpose down at each capture point before building the join. Our DPDP Act guide for marketers covers notices and retention.

Where to go from here

Run the overlap count and the cross-channel returns count this month. They answer the one question that matters.

If your customers do not cross channels inside one purchase, spend the money making each channel genuinely good — for most Indian SMBs that is the correct plan. If they do, build layer one, and nothing else, for a quarter.