Inform, Suggest, Then Let Them Decide
Inform, Suggest, Then Let Them Decide
More choice doesn't always help customers. Sometimes it leaves them stuck.
I work on the platform at Kiddocare, a Malaysian marketplace that connects parents with Kiddocarers (our babysitters) and tutors. When I looked at bookings that fell through, many of them already had Kiddocarers waiting. Parents had opened the list of applicants and looked through it, but never picked one. On the tuition side, most parents who received a shortlist of tutors never requested any of them. The ones who did request usually decided within minutes.
That last part stayed with me. The parents who chose didn't need days of comparison. They needed something to click. The ones who didn't choose weren't short of options. They had plenty.
The pattern we already accept from AI
Most of us now work with AI chat assistants in roughly the same way. We ask a question. The assistant suggests an answer and explains why. Then we decide: accept it, change it, or ignore it. The final call stays with us.
Nobody finds this strange. We don't ask the assistant for twenty possible answers ranked by relevance. We want its best guess, its reasoning, and the freedom to disagree.
I think businesses should work the same way with their customers:
- Inform: give customers the information they need to understand what they actually need.
- Suggest: offer one best option based on what you know about them, especially their needs, and explain why.
- Hand over the decision: the customer always makes the final call.
None of this requires AI. It's a way of structuring a decision. AI assistants just made it familiar enough that customers now expect it.
Step one: inform, with the right signal
Informing sounds like the easy step, and in one sense it is. Most businesses already do it. The problem is that they often do too much of it.
It's tempting to believe that if a customer is hesitating, they must be missing something, so you add another detail, another badge, another line of context. At Kiddocare we tried giving parents more context in a message meant to help them choose. It didn't help. If anything, it trended worse.
More information isn't automatically better. Every extra detail is one more thing to weigh, and weighing is exactly where people get stuck. The job of this step is to help the customer understand their own need, not to show them everything you know. A parent booking a sitter for a toddler at night needs to know who has handled that situation before. They don't need every Kiddocarer's full history at once.
Show the right signal, not more signal. If you can't say which piece of information changes the decision, you probably don't know yet, and adding more won't tell you.
Step two: suggest one, and say why
This is the step most businesses skip. They show a list, maybe sorted, and leave the rest to the customer.
A list puts all the work back on the customer. A suggestion does some of it for them. "This one" is easier to act on than "one of these".
What makes a suggestion trustworthy is the reason attached to it. "Recommended" on its own means nothing; customers have learned to read it as "promoted". A reason is different: "She has looked after your children before." "He teaches this subject at this level, near you, at the times you asked for." A reason lets the customer check the suggestion against what they know about their own situation. It also tells them what you know about them, which is a form of respect.
The reasons don't need to be clever. Simple signals like past bookings, favourites and ratings predict what people will choose far better than chance. Most businesses already collect these. They just don't turn them into a suggestion.
Step three: hand over the decision
The point of suggesting is to make choosing easier, not to choose on the customer's behalf. A few rules I hold to:
- One tap to accept. If taking the suggestion is as much work as browsing the list, you haven't helped.
- Always easy to choose differently. The other options stay one step away. Declining the suggestion should never feel like going against the system.
- Never auto-confirm when the stakes are high. In childcare, a parent is deciding who looks after their child. No suggestion, however good, should go through without them saying yes. The same goes for health, money, and anything else people would be upset to have decided for them.
- Keep the reasoning visible. The customer should be able to see why this option came first, so they can disagree with the reason, not only the result.
This is where the AI analogy matters most. People trust assistants partly because the assistant doesn't act on its own. It proposes, and the person decides. Take that away and a helpful suggestion turns into something people work around.
The catch: suggestions that shut out newcomers
There is a cost hidden in step two. The easiest suggestion is always the familiar one: the Kiddocarer the family has booked before, the product the customer already buys. Those suggestions are often right, and they're safe.
But in a marketplace, suggesting the familiar option has a side effect. When I looked at why new Kiddocarers lost jobs at Kiddocare, a large share of those losses went to someone the family already knew. When we set those contests aside, new Kiddocarers won about as often as anyone else. Being new wasn't the obstacle. Being up against the family's existing Kiddocarer was.
A suggestion engine that always picks the incumbent makes that worse. New Kiddocarers get fewer first bookings, so they never become anyone's familiar option, so they leave. Those newcomers are tomorrow's supply. A marketplace that only rewards incumbents slowly runs out of people to suggest.
The fix is to budget for exploration on purpose. Some share of suggestions should go to a strong newcomer whose fit is good but whose track record is still short, with a reason that is honest about it: "New to Kiddocare, experienced with infants, lives nearby." The customer still decides. You're only making sure the newcomer gets seen.
How to tell whether it works
Suggestions are easy to fool yourself with. A high acceptance rate looks like success, but it isn't impact. If you suggest the Kiddocarer a family was going to pick anyway, they'll accept it, and nothing has changed.
Two habits keep this honest:
- Judge it against a holdout. Keep a group of customers who don't see the suggestion, and compare outcomes. The question is whether more people completed a booking, not whether people clicked the suggestion.
- Count an accepted suggestion only if it changed the outcome. Accepting a suggestion for the option they'd have chosen regardless is not a win. The cases that matter are customers who would otherwise have left without choosing.
If you can't run a holdout, at least be careful about what you claim. "People liked the suggestion" and "the suggestion created bookings" are different statements.
Why most businesses stall at step two
Most businesses have done step one. They have product pages, profiles, reviews, filters. They get stuck at step two, and in my experience it's for one of two reasons.
The first is data. To suggest one option with a reason, you need to know something about the customer: what they've done before, what they've said they need, who they trust. Many businesses never recorded that in a form they can use.
The second is more common. They have the data and don't know what to do with it. It sits in reports that describe customers in aggregate, rather than being turned into a single suggestion for a single person at the moment they need to decide. Getting from "we know this" to "so we suggest this, because" is mostly a product decision, not a technical one.
Customers are already used to being informed, given a suggestion with a reason, and left to decide. The businesses that work this way will feel easier to deal with than the ones still handing over a long list.
If you run a product where customers have to choose, which step are you stuck on?