What Netflix Teaches a Small Business About Personalisation and Keeping Customers

In a 2015 paper, two Netflix product leaders, Carlos Gomez-Uribe and Neil Hunt, described a problem every shopkeeper will recognise. Their consumer research suggested that “a typical Netflix member loses interest after perhaps 60 to 90 seconds of choosing, having reviewed 10 to 20 titles (perhaps 3 in detail)”. After that, they wrote, the risk of the member abandoning the service “increases substantially”. (The Netflix Recommender System, ACM Transactions on Management Information Systems.)

Now picture a typical small business in Navi Mumbai: a pharmacy, a salon, an optician, a hardware supplier. It has hundreds or thousands of past customers in a billing app or a WhatsApp chat list, and almost never uses that information to bring anyone back. The question this article answers: what does Netflix actually do with personalisation, and which parts can a business without a data science team realistically copy?

Everything I say about Netflix below comes from Netflix’s own published material: its help centre and the 2015 paper written by its own product leaders. I have no connection with Netflix, and it is used here purely as a case study. The paper is now over ten years old, so read its figures as what Netflix reported at that time.

1. Why Netflix personalises: it is a retention machine

The most useful part of the Gomez-Uribe and Hunt paper is its plain description of the business case. Netflix’s revenue then came only from monthly subscriptions, and the paper says the company makes it “very simple to cancel the subscription at any time”. So the main thing personalisation had to do was keep members from leaving.

Source: Gomez-Uribe and Hunt, “The Netflix Recommender System: Algorithms, Business Value, and Innovation”, ACM TMIS 6(4), December 2015. Company-reported, US-led global service, figures as of 2015.

Two lines in that table matter most for a small business. First, the $1B figure is explained as savings: lower churn means a longer life for each subscriber and fewer new subscribers needed “to replace cancelled members”. Second, much of the churn was payment failure, not people choosing to leave. That is a boring, fixable problem, and small businesses have their own version of it.

This is not unique to streaming. A widely quoted Bain & Company article by Fred Reichheld, “Prescription for Cutting Costs”, says that “in financial services, for example, a 5% increase in customer retention produces more than a 25% increase in profit”, because return customers tend to buy more over time and cost less to serve. That figure is specific to financial services and is now more than two decades old, so I treat it as a direction, not a forecast. But the direction matches what most owners already sense.

2. Use what customers did, not who you think they are

Netflix’s help centre lists what its recommendations are based on: “your interactions with our service (such as your viewing history and how you rated other titles)”, “other members with similar tastes and preferences”, and information about the titles such as genre, actors and release year. It also uses time of day, language, device and how long you watched. And it says the system “does not include demographic information (such as age or gender)” in its decisions.

That last point is worth taking seriously. Many small businesses segment customers by guesses about who they are (“young customers”, “ladies”, “corporate types”). Netflix, with far more data than you will ever have, relies on behaviour instead. Behaviour is also what you already have: bills, bookings, service records, chat history.

Left column: Netflix Help Center, How Netflix’s recommendations system works. Right column: Infoways illustrations.

Recency matters more than you think

Netflix’s help page says “the titles you engage with more recently will outweigh titles you engaged with in the past”. For a small business, the simplest personalisation of all is knowing who has not been in for a while. A salon that sees a regular has not booked in ten weeks, or a pharmacy that sees a monthly refill is overdue, already knows enough to send a useful message.

3. Copy “Continue Watching”: remind people of what they started

One of the rankers the paper describes is the Continue Watching row. It sorts recently viewed titles by Netflix’s estimate of “whether the member intends to resume watching or rewatch, or whether the member has abandoned something not as interesting as anticipated”, using signals such as time since viewing and the point at which the member stopped.

Every small business has unfinished journeys. They are easier to spot than you might think.

• A clinic or dentist: treatment plans with a second or third sitting not yet booked.
• A coaching class: students who attended a demo and did not enrol.
• An online store: abandoned carts, and customers who bought a refillable product.
• A car or bike service centre: vehicles due for their next service by date or kilometres.
• A B2B supplier: quotations sent and not followed up.

Notice the second half of Netflix’s logic: the ranker also tries to tell whether someone abandoned a title because it was not interesting. The small business version is to stop chasing people who clearly said no. One polite reminder is service. Five is spam, and it damages the relationship you are trying to keep.

See also: For timing and message templates, see our guide to email marketing for small businesses and the WhatsApp marketing guide.

4. Explain why you are recommending something

The paper says the recommender’s job is to make sure each member finds something compelling “and will understand why it might be of interest”. It describes “evidence” algorithms that decide whether to show, for example, that a film won an award or that it is similar to something the member recently watched, and even “which image out of several versions” to use to support a recommendation.

For a small business, this is the difference between a broadcast and a personal message. “20% off this weekend” sent to everyone is a broadcast. “Your last order of printer cartridges was in July, and the same model is back in stock” is a recommendation with a reason. The second message is shorter to read, easier to trust and far less likely to be muted.

You do not need software to do this for your best fifty customers. You need the customer list from section 2 and ten minutes a day. For larger lists, WhatsApp Business already offers basic tools. Its features page lists Labels to “organize and filter important customer conversations”, Quick Replies for common answers, and a Catalog to “showcase your products and services in a mobile storefront”. Labelling customers by what they bought is a crude but real form of segmentation.

5. Test, do not guess

Netflix describes deciding between algorithms with randomised, controlled experiments: “our A/B tests randomly assign different members to different experiences”, then compare engagement and cancellation rates across the groups. The paper is frank about why: what customers ask for and “what actually works (a few compelling choices simply presented) are very different”, and the authors add that choosing by intuition, “even collective intuition”, often gives the wrong answer.

A small business cannot run statistically rigorous tests on a list of 400 people, and I would not pretend otherwise. But you can avoid the most common mistake, which is changing everything at once and having no idea what worked.

1. Split your list in two by something random, such as odd and even last digits of the phone number.
2. Send version A to one half and version B to the other, at the same time of day.
3. Change one thing only: the opening line, the offer, or the call to action.
4. Count replies or bookings within a fixed window, say 72 hours.
5. Keep the winner, then test the next change against it.

With small numbers, treat a small difference as noise. If version B gets 14 replies and version A gets 12, you have learned nothing. If B gets 30 and A gets 12, repeat it once to be sure, then adopt it.

6. Fix the boring leaks first

Remember the line from the paper that much of Netflix’s churn was “due to payment failure, rather than an explicit subscriber choice to cancel”. The small business equivalent is customers who wanted to come back and could not: the phone that was not answered, the enquiry form that went to an old email address, the WhatsApp message left unread over a long weekend.

These leaks are not glamorous, but closing them often brings back more customers than any recommendation engine. WhatsApp Business’s own features page lists Away Messages that “respond automatically to messages received outside your business hours” and a Greeting message for new chats. They are free and take minutes to set up.

See also: If you are considering automating replies, read AI chatbots for business first, including where they go wrong.

7. Where this advice breaks down

Personalisation is easy to overdo, and Netflix is a misleading model in several ways.

• Scale. Netflix’s methods depend on huge numbers of members and constant usage. A business with a few hundred customers who buy twice a year cannot build a meaningful “customers like you” model, and should not pay someone to pretend it can.
• Frequency. Netflix customers open the app many times a month. A plumber, a wedding photographer or a property dealer gets few repeat purchases, so retention messages should be rare and genuinely useful.
• Creepiness. A message that reveals more than the customer expects you to know (a pharmacy referencing a sensitive purchase, for example) does more harm than good. Ask yourself whether the customer would be comfortable seeing the reason written down.
• Law. India’s Digital Personal Data Protection Rules were notified on 14 November 2025 with an eighteen-month phased compliance period. According to the PIB explainer, every data fiduciary must issue “a separate consent notice that is clear and easy to understand” explaining the specific purpose, and individuals “may withdraw their consent at any time”. Get permission before using purchase history for marketing, and make opting out easy. This is not legal advice; check with your CA or lawyer.
• Old data. The Netflix figures here are from 2015. They show how Netflix reasoned at the time, not what its systems look like today.

My view is that most small businesses should not buy a “personalisation platform” at all until they have done the manual version for a few months and can show it brings people back. If the simple version does not work, the expensive version usually will not either.

A Netflix-style checklist for a small business

Netflix found that people give up after a minute or so of not finding what they want. Your customers are no more patient, and a well-timed, relevant message from someone who remembers them is something a ten-person business can do better than a giant.