Deflection Without Hiding: Cut Contacts That Should Not Exist

Deflection has a bad reputation in Indian ecommerce, and it earned it honestly. Too many brands read the word as permission to make the contact button harder to find, watch ticket volume fall, and call it efficiency. The contacts did not go away. They moved to social media, to marketplace claims, and to the review section.

There is a legitimate version of this work. It rests on one distinction. Some contacts exist because the customer needs a human. Some contacts exist only because the business failed to tell the customer something it already knew. The second group should not exist. Removing it is not deflection in the cynical sense. It is fixing an information failure.

The proactive notification link

Start with the arithmetic that most teams never do. Every contact has a trigger event that preceded it. Pull a month of tickets, look at the timestamp of the contact, and look at what happened to that order in the twenty four hours before.

The pattern is consistent. A shipment sits without a scan. The promise date passes. A delivery attempt fails. A refund is initiated but not credited. In each case the business knew something the customer did not, and the customer contacted support to close that gap.

A message sent at the moment the event occurs removes most of that contact volume. Not because the customer is being fobbed off, but because they now have the answer they were going to ask for. The message has to be specific to be worth sending. A delay notice that says your order is delayed does very little. A delay notice that says the shipment is held at a named hub, gives a revised date, and offers a cancel option does the whole job.

Send it before the customer notices, not after. A revised date that arrives after the original promise date has already passed reads as an excuse rather than an update.

Self serve that actually resolves

Self serve only counts if it completes the transaction. A tracking page that shows the same courier status the customer already saw is not self serve. It is a redirect.

The self serve surfaces that carry real volume in the Indian market are narrow and predictable. Order status with a plain language explanation of what the current stage means and a revised date when one exists. Return or exchange initiation with the pickup slot visible. Refund status showing the initiation date, the destination account and the expected settlement window. Address correction while the shipment is still amendable. Cancellation before dispatch.

Each of those has to end in a completed action with a confirmation. If the flow ends in a form that says our team will get back to you, you have created a ticket with extra steps and annoyed the customer on the way.

One India specific note. A large share of these interactions happen on a phone, often on a slow connection, and often through a messaging app rather than a website. Build the flow for that context first. A help centre optimised for desktop reading is not going to move the number.

Write help content from the taxonomy, not from imagination

Most help centres are written once, by someone who guessed at what customers would ask, and then left alone. The result is forty articles that answer questions nobody asks and no article for the thing driving a quarter of your volume.

There is a better source. Your closed ticket data already ranks every question customers actually ask, by volume and by cost. Take the top contact reasons, and write one page for each, in the order of that ranking. Use the words customers used, not internal vocabulary. Customers write parcel and courier, not shipment and logistics partner. They write money not received, not refund settlement pending.

Then close the loop. Track which help pages precede a contact anyway. A page with high traffic and a high contact rate immediately after is a page that failed. Rewrite it or replace it with an action.

The honest line between deflecting and burying

Here is the test we apply. If the customer had reached a human, would the human have told them anything they could not have found themselves. If the answer is no, the contact was avoidable and removing it is a genuine improvement. If the answer is yes, then blocking that contact is not deflection. It is hiding.

Some contact types should never be deflected. Anything involving money the customer has not received. Anything involving a safety or quality issue with the product. Anything where the customer has already tried self serve and it failed. Anything where the customer has contacted more than once about the same order. And anything where the tone indicates real distress, which a well trained team spots quickly.

Practically, that means the path to a human always exists, is visible on every self serve surface, and gets shorter rather than longer as a customer repeats a contact about the same issue. Someone contacting the third time about one order should reach a person faster than a first time contact, not slower.

Deflection metrics are easy to game, so audit them

Be blunt about this internally. Any target expressed as contacts per order can be hit by making contact harder. If the incentive rewards the number without checking the mechanism, someone will move the contact link, add a mandatory article view before the form loads, or shorten support hours, and the dashboard will look excellent.

These are the checks that catch it.

• Watch repeat contact rate. Genuine deflection leaves it flat. Hiding pushes it up, because the same customer keeps trying.
• Watch channel migration. If tickets fall while social media mentions, marketplace claims, payment disputes or negative reviews rise, the volume moved rather than disappeared.
• Watch time to first human contact for the tickets that do arrive. If it is climbing, friction was added somewhere.
• Watch abandonment inside the self serve flow. High entry with low completion means the flow is a wall, not a resolution.
• Watch satisfaction split by whether the customer used self serve first. If the self serve cohort scores materially worse, the flow is failing and pushing people into contact anyway.
• Walk the contact path yourself, on a mid range phone, once a month. Time how long it takes to reach a human. Most gaming is visible in ninety seconds of doing this.

Sequence the work

Do it in this order. Fix proactive notification for the top three trigger events first, because it needs no customer behaviour change and moves volume fastest. Then build self serve for the top two contact reasons that end in a completed action. Then rewrite help content in taxonomy order. Then, only after those are live, look at contact placement.

Brands that reverse this order get a short lived drop in tickets and a longer lived drop in repeat purchase. The volume is a symptom. Treat the cause.