A family plan should begin with a conversation about how people actually communicate. Who answers calls? Who prefers text? Who is hard to reach during work? Who uses a landline, a relay service or another person’s help? A plan that ignores these facts may look sensible on paper and fail on its first use.
It should also be mutual. The older person is not the only one allowed to make a mistake, and the younger person is not automatically the authority on every request. Anyone may ask for a check. Anyone making an unusual request should expect to explain it through the agreed route.
The purpose is to make help easier to give safely. It is not to place one member of the family under constant supervision.
Warnings can help, but the evidence is mixed
Scheibe and colleagues tested warnings among people who had been victims of mail fraud. Of 895 people assigned to warning or control conditions, 511 were reached for a later mock telephone offer. The warnings shifted clear acceptance toward skepticism. No money was collected. The study offers evidence that a prior warning can change responses to a later approach, while leaving actual loss prevention unmeasured. It tested a sales offer, not a familiar cloned voice or a family emergency. Scheibe and colleagues, 2014
A different study examined a Danish bank’s mass warning using customer and fraud records. Its analyses found no significant protective effect. This was not an individually randomized trial, and the researchers could not establish who actually read the message. Even so, it is important contrary evidence to the idea that sending advice is enough. Jensen, Gerlings and Ferwerda
These studies should not be averaged into a single verdict on education. They used different people, messages, comparisons and outcomes. One observed responses to a mock offer; the other examined recorded fraud after a bank campaign. Their differences tell us what to test next.
For a family agreement, reading a warning is only the beginning. Can the person find the established contact? Does someone answer? Can they use the fallback without being shamed? Does the plan still make sense when the request is genuine? Those are practical questions that a general awareness score will miss.
A calm walkthrough may reveal missing details before a crisis. Use an openly invented, low-stakes request and let everyone know it is practice. Do not surprise a relative with a fake emergency or clone their voice to teach a lesson. Such a test can frighten people and damage the trust the plan is meant to support. Practice should be agreed, brief and easy to stop. Its benefit remains a proposal to test, rather than an outcome established by the studies above.
A phrase is one possible check
The FBI recommends that families consider a secret word or phrase when checking identity. It provides no measured effectiveness rate for that step. A family may find a phrase helpful, but it should be an additional agreed check with a fallback. FBI guidance, December 2024
Forgetting the phrase is not proof of fraud. Someone may be distressed, unable to speak or simply unable to recall it. Knowing the phrase is not unlimited permission to send money either. A shared phrase could become known beyond the people who chose it. The particular request still needs checking.
There is no need to include a real phrase in a public example, shared research record or online discussion. It is also distinct from a bank password, a one-time account code or any other credential. A caller asking for those account secrets has not become safer because the conversation includes a family word.
Some households will prefer another arrangement. The useful principle is a mutually agreed way to reconnect the request to the person. The right form depends on memory, language, speech, hearing and access. There is no virtue in choosing a clever test that someone cannot use.
Design with the person who will use it
A 2025 SOUPS poster reported six focus groups with 37 adults aged 70–94 at one US independent-living facility. Participants described trusted helpers and ways of checking suspicious contacts. They also raised problems such as family members being unavailable and missed medical calls when avoiding unfamiliar numbers. These are accounts from one setting, not a trial proving a prevention method. They show why a plan must preserve access to legitimate help as well as reduce exposure to deception. LaRubbio and colleagues, 2025
Ask what the person wants from the arrangement. A printed card may be easier than a contact hidden several screens inside an app. Large text may help. An agreed text route may be preferable to speech. Two willing backup contacts can reduce reliance on a single busy person. These are design options, not tested equivalents with identical protection.
Avoid asking someone to share every message, surrender control of an account or accept constant location tracking as the price of support. Assistance should be limited to what the person wants and what the particular decision needs. A family agreement can protect privacy by naming who may be contacted and what may be shared when checking a request.
We should also distinguish an accessible warning from a proven reduction in harm. Sharevski and colleagues studied aural telephone scam warnings with 72 people, half legally blind and half sighted. Nobody in the no-warning baseline pressed the scam prompt. Two presses occurred under a contextual warning, one accidental and one deliberate engagement. The study supplied observations about usability and preferences, but its baseline leaves no demonstrated warning benefit on that action. It did not test actual financial loss. Sharevski and colleagues, preprint version 1, 2024
This is a useful reminder about evidence. Participants can like a feature without a trial proving that it reduces victimization. Both questions matter. We need to ask them separately and include people who use different forms of communication in the work.
Keep the agreement small enough to remember
A household can write down its own answers to a few practical questions. Which requests trigger a check? Which known route should be tried first? Who has agreed to be a backup? What accessible option is available if speaking is difficult? How will the household seek emergency help locally?
The answers should be revisited when contact details or circumstances change. A new phone, a move or a change in caring responsibilities can make an old plan unusable. Updating the plan is ordinary maintenance, not evidence that someone has become less trustworthy.
There should also be a shared response to mistakes. If someone forgets the routine, the next conversation should repair the gap. What made it hard to use? Was the contact unavailable? Was the rule unclear? Did fear of offending someone get in the way? Blame supplies little information about any of these questions.
A plan succeeds in its purpose when people can use it without humiliation and resume ordinary contact after the check. That is the standard I would put before a household. Research still needs to establish how much particular plans change behavior and harm. Meanwhile, we can make the plans more respectful, more concrete and easier to use.