Product
Email and conversations
What was agreed with a customer is usually in an email, and which email depends on who was on the thread. That is the failure this addresses.
Where the agreement lives
The most important records in most businesses are in a mailbox
Ask any small organisation where the record of what was agreed with a customer lives and the honest answer is email. Not the CRM, not the file, not the notes. The substance of the relationship, the promises, the caveats, the thing somebody said they would do by Friday, is in a thread.
That has three consequences and every organisation of a certain size has met all three.
It is invisible. Whoever was on the thread knows. No one else does, including the person covering for them next week and the person who takes over the account next year.
It is unsearchable in practice. Technically an inbox can be searched. Practically, finding what was agreed eighteen months ago requires knowing roughly when and roughly who, and the person who knows both is the one who is not available.
It leaves with people. When somebody goes, their mailbox goes with them, and what the organisation knew about those customers reduces to whatever happened to be written down somewhere else.
Putting the conversation on the customer
Correspondence about a relationship belongs on that relationship. It appears in the timeline in order with the stage changes, the notes, the files and the appointments, so reading a customer record means reading what actually happened rather than a summary somebody wrote afterwards.
The test is the same one the timeline is built for: someone picking up a customer for the first time should be able to read down and know where things stand. Without the conversation, that test fails for most customers, because the interesting part of the history is precisely the part that was in an email.
Without asking anyone to work somewhere new
People keep using their normal mail application. This is not a second inbox to live in, and a product that requires everybody to change where they read email will be abandoned by the third person who is busy.
What changes is that correspondence with the people in your records also lands on those records. The behaviour that used to be a discipline, remembering to copy the important thread into the CRM, becomes something that happens without anyone deciding to do it.
Inbox against record
Six situations everybody has been in
The middle column is what most organisations have now.
| The situation | A shared inbox | Messages on the record |
|---|---|---|
| What did we agree with them? | In an email. Which email depends on who was on the thread and whether they filed it. | On the customer's timeline, in order, with everything else that happened. |
| The person who handled it is on leave | Two days and a guess, or a request to somebody to search their own mailbox. | Whoever is covering opens the record and reads it. |
| Somebody new joins | They inherit an inbox with no history and learn the customers by asking. | They open a customer and read what has happened, in order. |
| A customer says you promised something | A search through email, hoping the person who promised it still works there. | The message is on the record with a date, next to what happened afterwards. |
| Someone withdraws consent | It is applied in the mailing tool. The next campaign is built from somewhere else. | Consent is a fact on the person, checked at the moment of sending, by everything that sends. |
| The same customer emails three people | Three threads, three partial pictures, and occasionally three different answers. | One record, one history, and whoever replies can see what the others said. |
Consent
Checked when it is sent, not when the list was built
The product knows whether somebody has agreed to be contacted, about what, through which channel, and since when. Every send checks that again for that individual at the moment of sending.
The distinction between checking the list and checking the person is where almost every real failure happens. A list built for a campaign three weeks ago is a snapshot. Somebody who withdrew last Tuesday is still on it, and in a system that checks at build time they receive the message and are entirely right to complain.
Consent is also held per channel, because agreeing to a service email is not agreeing to a newsletter. Treating them as one permission is how organisations end up sending marketing to someone who asked for an invoice, and then losing the ability to send them an invoice when they unsubscribe from everything.
Replies
A reply stops the follow up, immediately
When somebody answers, they come out of any sequence they are in. That single behaviour is the difference between a follow up and something that makes you look automated, and it is the one thing organisations most often get wrong when they build this themselves.
Everyone has received the message that begins by wondering whether they saw the last one, two days after replying to it. It is a small thing and it is unusually damaging, because it tells the recipient precisely how much attention is being paid.
Because messages live against the record, the reply is also visible to whoever picks the relationship up next, in order, with what preceded it.
Deliverability
Authenticated properly, in both directions
Everything sent from here leaves from one domain we control, so it can be signed and aligned for the checks receiving servers make. Your shared inbox replies come from an address on that same domain rather than from your own, which is the arrangement that lets them be signed at all, and the code refuses to send from anywhere else instead of sending a message that would arrive unauthenticated. The security page says which half of that is code and which half is DNS.
Inbound mail that becomes part of a record is matched to your organisation through a reply address that is yours rather than shared. A message intended for one customer's system cannot be delivered into another organisation's records by guessing an address.
Verification and reset messages carry single use tokens with short expiries, bound to the address they were sent to. A token that has been used is spent, and presenting it again does nothing rather than working a second time.
Parts of a thread
Where a conversation lives
Each part, and the reason it is shaped that way.
- Against the record
- Messages sit on the relationship they concern. The thread is part of the customer history rather than in one person's mailbox, which is the whole point.
- Shared, not forwarded
- A colleague picking up a customer reads the conversation. No one has to forward anything, and no one has to remember to.
- Timeline entries
- A message sent or received is an event on the timeline like any other, in order with the stage changes, the notes and the appointments.
- Consent aware
- The product knows whether someone has agreed to be contacted and about what, checked at the moment of sending rather than at the moment a list was built.
- Reply detection
- A reply stops any sequence that person is in. A follow up arriving after somebody has already answered is the fastest way to look automated.
- Per organisation reply address
- Inbound mail is matched to your organisation and to the relationship it concerns. A reply address is not shared between customers.
- Authenticated sending
- Everything leaves from one domain, and the code refuses to send from any other, which is what makes signing and alignment possible. Whether the records that complete it are published is a matter of DNS rather than of code, and the security page says so.
- Attachments
- Files on a message become files against the record, so the document someone sent you is where the customer is rather than in an inbox.
When the person who agreed something is on leave, the answer to a simple question should take a search rather than two days and a guess.
Which is the test the whole timeline is built to pass, and the one it fails without the conversation.
The judgement
Not everything belongs on a shared record
It is worth being explicit about this, because a page arguing that conversations should be shared can easily read as arguing that everything should be.
Correspondence about a customer belongs on the customer. A colleague's private view of a difficult client, a negotiation with a supplier that is genuinely confidential, a conversation with an employee that happens to involve a customer name: those are different, and a system that pulls everything in indiscriminately produces two outcomes, both bad.
The first is that people start using a different address for anything sensitive, which moves the important conversations further out of reach rather than closer. The second is that the record fills with material that should not be visible to everyone who can see the record, at which point sensitivity is doing work it should not have to.
So the judgement stays with the person. What lands on the record is correspondence with the people in your records, and there is no obligation to put everything there. That is a deliberate limit instead of a missing feature, and it is the one that keeps the record trustworthy enough to be shared widely.
Email, and nothing beside it
What email here does not cover
No chat and no text messaging
Email and the record of conversations. There is no live chat widget for your website and no SMS. If either is central to how you talk to customers, this does not cover it, and the roadmap names them rather than implying they are close.
Not every mailbox provider
Microsoft 365 and Google Workspace are the two that matter for this audience. The product's health endpoint reports honestly which connections are configured rather than assuming them. If you are on something else, ask, because the answer today is probably no.
Not a replacement for your email
Nobody should be reading their mail inside a CRM. This is not an inbox and is not trying to become one. People keep working where they work, and the record gets the correspondence that concerns it.
Where the substance leaks
Five ways this goes wrong
The real conversation moves somewhere else
The symptom is a record full of confirmations and no substance. The cause is people deciding, reasonably, that they would rather not have something on the record, so the decisions happen on the telephone and only the summary is written down.
Occasionally that is the correct outcome. Usually it means the boundary was never agreed and everyone set their own, which produces a record that is systematically missing exactly the messages that mattered.
Everything is attached to everything
The symptom is a timeline where the correspondence drowns the events. The cause is connecting a mailbox that handles a high volume of routine traffic, most of which is not about any particular relationship.
Somebody connected a shared inbox no one owns
The symptom is no one being sure whether a message has been dealt with. A shared address with no owner is a queue pretending to be a mailbox, and the thing that fixes it is an owner instead of a feature.
A customer is answered twice, differently
The symptom is exactly what it sounds like. The cause is two people looking at the same thread with no indication that either has replied.
The record shows what has been sent, which is most of the answer. The rest is a habit: take the thing before answering it, so someone else can see it has been taken.
Somebody leaves and their correspondence goes with them
The symptom is a customer relationship that has to be rebuilt from the customer's own memory. The cause is a mailbox that was never connected, which is the ordinary condition of most small organisations and the reason this exists.
This is the failure the whole feature is aimed at, and it is worth naming plainly because it is the one that costs the most and produces no incident at the time it happens.
The record in detail
Three addresses on one message, and why it needs all three
What is actually stored, and what appears on screen when someone opens it.
There are two things here, not one. A conversation is the unit people work with. A message is one email inside it. Keeping them separate is what lets a thread be assigned, closed and attached to a relationship as a single object, while each message keeps the facts that only belong to it.
What a conversation holds
A subject, taken from the first message. The relationship it concerns, which may be empty. The piece of work it concerns, if somebody has filed it against one: a job, a deal, a matter or a catalogue entry, so the correspondence about a thing sits with the thing. Whether it is open or closed. Who it is assigned to, if anybody. Whether it is sensitive. When the last message arrived, whether that message came in or went out, and how many there have been. Nothing else, because everything else is a property of one message rather than of the exchange.
The direction of the last message is the one field that earns its place twice. It is what the list uses to say that something needs an answer, and it is the reason the inbox can be read at a glance rather than opened one row at a time.
What a message holds
Which way it went. The identifier the mail system gave it. The identifier of whatever it answers. Three addresses. The subject. The plain text body and the markup body. A note of whether it arrived carrying attachments. The verdict the receiving system reached about whether it was spam. And when it happened, which is when it was sent rather than when it was taken in.
Why three addresses
The first version of this stored one. That was the envelope sender: the address the sending server declares, which is the right thing to trust, because it is the address the authentication checks are made against and it cannot be freely chosen by whoever wrote the message.
It is also the wrong thing to reply to. Real mail relayed through a sending service arrives with an envelope of the form bounces at somewhere, which exists to collect delivery failures. A reply addressed there reaches no one at all. That was found by sending a real message through a real relay, which is the only way it could have been found, and it is the reason there are now three fields with three jobs: the envelope is what is trusted, the From header is who the person is and what is shown on screen, and the reply address is where an answer is sent. When a sender gives no reply address, the From header is used. The envelope never is.
What is on screen
The list is conversations in order of what happened most recently, filtered to open or closed, showing the subject, the number of messages, when the last one was, and who has it or the word anyone. Anything whose last message came inbound is marked as needing an answer. Two hundred rows at a time, which is more than any organisation of this size has waiting.
Opening one gives the subject, whether it is open or closed, and either a link to the relationship it belongs to or a line saying that it is not yet matched to anybody you hold a record for. Then the messages in order, each marked as received or sent with the address it came from or went to, and the body shown as plain text.
As plain text deliberately. An email is the one place where a stranger can put markup in front of your staff, and rendering it faithfully is a courtesy extended to the wrong person. The markup body is stored, so nothing is lost, and it is not what is shown.
One thing is recorded and not displayed anywhere today: the spam verdict. It is written down rather than acted on, because deciding what is spam is a judgement an organisation should be able to see and disagree with, and at present it can do neither. What a message arrived carrying is no longer in that category: attachments are listed on the conversation and can be downloaded from it, as files against the record.
Threading and its failures
Threading is a claim, and the claim is checked
Almost everything difficult about holding mail against a record is a case where the obvious behaviour is also the dangerous one.
The stranger who has seen one of your identifiers
A reply says which message it answers, and it says so in a header written by whoever sent it. Anyone who has ever been copied into a thread, or been forwarded one message out of it, knows that identifier and can write a message claiming to answer it.
If that claim were taken on trust, they would land inside someone else's conversation. That is not merely untidy. A reply goes to the most recent message that came in, so the organisation's next answer, quoting everything that came before it, would be addressed to them.
So a message joins an existing conversation only if its sender is already part of that conversation, as someone who was written to or someone who wrote in. Anyone else starts a conversation of their own. That is the safe answer and it is also the truthful one: a new person writing in is a new matter until somebody says otherwise. The chain of ancestors is read from the end, because the nearest one is the most likely parent, and it is bounded to ten, because a genuine chain is a handful of identifiers and a thousand is someone finding out how long they can hold your database open.
The same message arriving twice, and by two routes
Mail systems deliver twice. A conversation that grows a duplicate every time a relay retries is worse than one that misses a message, so receiving is idempotent: a message whose identifier has already been stored returns the conversation it is already in and writes nothing.
Arriving by two routes is the commoner case and needs a second answer. A reply sent from Consonas also appears in the sender's own sent folder, so a connected mailbox would offer it back. Both identifiers are therefore checked: the provider's own, which catches the same message seen twice by the same connection, and the mail identifier, which catches the same message reaching the record by two different paths. Within a single batch too, because a provider can return one message under two identifiers in one page of results.
Things arriving out of order
A message is recorded at the time it was sent rather than the time it was taken in, so a mailbox connected on Friday puts Tuesday's exchange in Tuesday's place. The conversation's own clock moves to the latest message it holds, which means a backfill can quietly change the order of the inbox. That is the correct behaviour and it is still surprising the first time.
Someone you have no record for
The conversation exists, with no relationship attached, and the screen says so in those words instead of guessing. Attaching it afterwards is a deliberate act, and it checks two things rather than one: the conversation, and the relationship it is being attached to. Putting correspondence on a record is a claim on that record, so someone who cannot reach the record cannot put anything on it.
A relationship that is marked sensitive
Correspondence with a sensitive relationship is itself sensitive, from the first message. Without that, marking a relationship hid the name and left the entire body of the correspondence readable by everyone, which is a worse outcome than not marking it at all, because it looks handled. Attaching a conversation to a sensitive relationship after the fact carries the marking across at the moment it is attached.
Machinery, and people who leave
A message where everyone involved is a machine address is a notification rather than correspondence, and it is not taken. When someone exercises a right to be erased, conversations are detached rather than deleted: the exchange is the organisation's own record of what it agreed, and it stops pointing at anyone.
One week, one enquiry
One enquiry at an imagined joinery workshop, followed the whole way
Six people, no customer named here is real, and every step below is something the product writes down.
Consider a six person joinery workshop. They have a shared address that belongs to their organisation, and two of the six have connected their own mailboxes. A builder they worked with last year is already a record, because somebody made one at the time.
Wednesday morning: the message arrives
The builder writes to the shared address. The address is looked up to find which organisation it belongs to, which is the same check a request from a browser goes through, and nothing in the message itself is allowed to name a database. It is under the size ceiling, so it is accepted. Had it been over, it would have been refused at the door with a reason, so the sender is told, rather than accepted and quietly dropped.
It answers nothing, so it opens a conversation. His address matches a record, so the conversation belongs to that record from its first message. Three things then happen that no one asked for: an entry appears on the builder's timeline saying an email was received and naming the subject, the follow up sequence he was in stops, because he has answered and chasing him now is the fastest way to look automated, and an event is raised that a workflow could act on.
Wednesday afternoon: someone answers
The inbox shows one conversation marked as needing an answer, with nobody assigned. A colleague who has never dealt with this builder opens it, follows the link to the record, reads what happened last year, and replies.
The reply is written into the record first and sent second, and the screen says which of those two things happened. If the sending fails, there is a message sitting in the conversation that visibly did not go, which is something to chase. The alternative, sending first and recording on success, produces a reply that nobody can see and that the organisation therefore never sent either. It goes out from the shared address, so the answer comes back to the inbox rather than into one person's mailbox, and the builder's timeline gains a line saying an email was sent.
Friday: the builder replies and copies his client in
The reply threads, because the builder is already part of the conversation. His client is copied in, and the workshop holds no record for the client. That does not matter for this message: it is taken because somebody known is involved in it.
On Monday the client writes separately, about the same job, from the same address. He is not part of the existing conversation and there is no record for him, so his message opens a conversation of its own with nothing attached to it. Somebody notices, creates a record for him, links the two, and the correspondence appears on his timeline from that moment. Nobody was told to notice. That is the honest half of this example.
The following week: the joiner who handled it is away
A question comes in that needs an answer today. Someone who was not involved opens the builder's record, reads down the timeline, and finds the exchange in order with the appointment and the quotation. They answer without ringing anyone and without asking a colleague on leave to search their own mailbox.
That is the whole return on this feature, and it arrives in one moment rather than continuously. When the job finishes the conversation is closed. Three months later the builder writes again on the same thread, which reopens it, because someone writing back is the clearest possible signal that a matter is not finished.
What we turned down
What was rejected on the way to this
Each of these had an easier alternative, and the easier one was worse.
Reading and sending, and nothing else
A connected mailbox grants three things: permission to read mail, permission to send from that address, and permission to learn which address gave the permission. Sending is the largest of the three and it was added deliberately, so it is worth saying what bought it: a reply or a sequence step belongs in the thread the customer started, from the person they wrote to, and a reply that arrives from a service they have never heard of is one their filter and their memory both treat as a stranger.
What was turned down is everything past those three. The consent does not ask to delete, move or alter anything in the mailbox. The read half deserves an exact sentence, because it is the one somebody standing at the consent screen will want: the read permission both providers offer does extend to message bodies, and neither offers a narrower one that still allows sending. What the product asks the mailbox for is metadata, every time, which is a decision in our code rather than a limit written into the consent. The next section is that decision. A consent screen that lists full control of somebody's mail is a consent screen a sensible person refuses, and it would buy nothing this product does.
Marketing campaigns go the other way and always have. They leave through our own sending service, from a Consonas address rather than one of yours, because a thousand messages in one afternoon from one person's mailbox is how a mailbox stops being trusted. There is no per organisation sending domain today, and the terms say the same thing in the same words.
Asking for headers rather than bodies
What is requested from a connected mailbox is the identifiers, who a message was from and to, the subject and when it was sent. The body is not asked for, because it is not needed to decide whether a message belongs on a customer record, and asking for less is the whole posture here.
The cost is real and worth stating plainly. A message taken from a connected mailbox is a line in the history rather than something you can read in full. Mail sent to the organisation's own shared address is complete, because it was sent to the organisation. Mail from somebody's personal mailbox is a record that the exchange happened. If you want the substance, use the shared address.
Only correspondence with people you already hold a record for
The rule that decides what to take is a single line: a message must involve someone the organisation already has a record for. Everything else is skipped and counted, so you can see how much was left alone.
The alternative was taking everything and letting people delete what they did not want. A salesperson's mailbox contains their doctor, their bank, their colleagues and their family. Copying all of it into a system every colleague can read is a thing some products do and it is indefensible. The price of the rule is that correspondence with someone you have not made a record for yet is not taken at all, and the answer to that is to make the record.
Thirty days, not the whole history
Connecting a mailbox looks back thirty days by default. Connecting one should not drag ten years of correspondence into a record no one asked to have filled, and an organisation that genuinely wants the older material can say so. A first pass over a long history is a series of bounded passes rather than one that runs until it finishes, and each pass is asked for rather than continuous.
Credentials in their own table, excluded from the export by name
The tokens for a connected mailbox are the most dangerous thing this product will ever hold, because they are keys to someone's entire mailbox rather than to their records here. They live in a table of their own, so that everything which handles mailboxes does not automatically handle secrets, and the complete export excludes that table by name. Exports get forwarded, opened on other machines and left in folders nobody audits, which is the whole of the argument for keeping a live mailbox credential out of one.
Disconnecting a mailbox deletes the credentials and leaves the messages already taken. They are correspondence with customers and they belong to the organisation, not to the mailbox they happened to arrive through.
What is not searchable, and why that is a compromise
Search covers the subject of a conversation instead of the text of every message. Indexing entire bodies would multiply the storage cost of the largest family of records in the database, and every organisation pays for its own database. The consequence is that a phrase you remember from the middle of an email will not find it, and the subject, the person and the date will. That is a compromise rather than a design, and it is stated here so that nobody discovers it while looking for something.
Who may read the mail
Where the inbox begins, and what a restricted colleague can already do
Three permissions, deliberately separated
Reading a conversation, replying to one, and managing one are three different permissions. Managing covers assigning, closing, reopening and attaching a conversation to a relationship. Templates carry their own pair, viewing and editing, and connecting a mailbox is a permission of its own again.
In the roles the product starts with, a manager and an ordinary member hold all three across the organisation. A restricted member reads and replies but cannot assign, close or attach, which is the right shape for someone who should answer customers without deciding what the queue means. A read only member reads and does nothing. All of these are defaults, and an organisation may change every one of them.
A connected mailbox is always somebody's own
Connecting is granted at the scope of the person's own account, and the list of connected mailboxes is filtered to whoever is asking, always. Disconnecting someone else's is refused, and a mailbox that is not yours answers exactly as a mailbox that does not exist. That is not a permission an organisation can widen its way around, because the filter is applied whatever the role says: a connected mailbox is a personal account, and the list of them is not organisation information.
Sensitive conversations follow the relationship
A conversation on a relationship marked sensitive is only visible to people who may reach sensitive records, and that applies to the list as well as to the conversation itself. A search will not return it either. The point of marking a relationship is defeated entirely if the correspondence about it stays in plain sight.
Which plan
The shared inbox and email integration both begin at Standard, and that is stated on the pricing page rather than discovered. Below Standard, mail sent to your organisation is still received and still recorded. It is reading and replying that are refused. Bouncing a customer's message to punish an organisation's plan would cost that organisation something nobody can give back, so the history accumulates and is there the day they buy the inbox.
A connected mailbox counts against the integrations allowance, which is one on Free, two on Starter, ten on Standard and uncounted above that. Both the feature and the count were once unchecked, which meant a Free organisation could connect a mailbox and as many more as it liked. One person may connect one mailbox per provider per address, because two connections to the same mailbox would take everything twice.
What stays on every plan, including the free one
Every reply sent, every conversation closed, every mailbox connected and every mailbox disconnected is an entry in the audit trail, on every plan, permanently. Consent records and their whole history are on every plan. The complete export carries your conversations and your messages out with everything else. What it leaves behind is credentials: the mailbox tokens, the calendar connection with the tokens that sit on it, and the signing secret on a webhook, which is emptied out of a row that is otherwise carried in full. It also carries what the person running it may reach, so an export run by somebody who holds the export permission without the grant for sensitive records comes out short by exactly the number its manifest declares.
Retention, if you want it
An organisation may set a rule that conversations older than a stated age are cleared: the subject is replaced with a line saying it was removed under the retention policy, the messages are deleted, the relationship is detached and the search entry goes. The shortest rule the product will accept is thirty days, because a rule shorter than that is almost always a mistyped figure and the consequence of a mistyped retention rule cannot be undone.
What the audit trail records is a count rather than a list. An entry naming every record removed under a retention policy would preserve the very identifiers the policy existed to remove, which would make the trail the thing that defeats it.
Asked about mailboxes
Asked about email and conversations
Which mailboxes does it connect to?
Microsoft 365 and Google Workspace are the two that matter for this audience, on the plans that carry integrations. Which are configured is reported honestly by the product's health endpoint rather than assumed, and the roadmap names what is not connected instead of describing it as coming soon.
Does it read all of my email?
No. What comes into the record is correspondence with the people in your organisation's records, not the contents of somebody's personal mailbox. An email from your accountant about your own tax return is not customer correspondence and has no business being in a customer database.
Can I keep some emails private?
Yes. Not everything belongs on a shared record, and a system that forces everything into the open makes people use a different address for the conversations that matter. What sits on the record should be the correspondence about the customer, and that judgement stays with the person.
What happens to attachments?
They become files against the record, so a document someone sent you lives where the customer is rather than in a mailbox. They count toward your file storage allowance, which is on the pricing page.
Can we send from a shared address?
Yes. A reply leaves from an address on the one domain everything here sends from, which is what allows it to be signed; sending from anywhere else is refused rather than sent unauthenticated. The records that complete the arrangement are published in DNS, which is a deployment step and not something the product can check for you.
Does it do live chat or SMS?
Neither today. It is email and the record of conversations, and the roadmap says so by name. If chat on your website or text messaging is central to how you talk to customers, this does not do it.
How does it know which customer a message belongs to?
By matching the address against your records, and by the reply address, which is per organisation rather than shared. A message from somebody you do not have a record for can create one or be left alone, which is your choice rather than an automatic behaviour.
Is this a replacement for our email?
No, and it should not be. People keep using their normal mail application. What changes is that correspondence about a customer also lives on that customer, so the record is complete without anybody working in a second inbox.
Where the thread is the contract
Trades where what was agreed is only ever in an email
The feature pays for itself in one moment: someone covering for a colleague opens a record and knows what was promised. That moment arrives most often in construction, where a variation is agreed in a reply and priced from it months later, and in professional services, where the caveat someone attached to advice is the part that matters afterwards.
It matters for a different reason in travel, where one booking is a long exchange of amendments and the person who took it is frequently the person who is away, and in maintenance and facilities, where a problem is reported by email and then reported again by somebody else.
In all four the correspondence is the record, and the ordinary condition is that it leaves with whoever wrote it.
Put one customer's correspondence on the record
Then ask someone who has never dealt with them to tell you where things stand. That is the whole test.
Three people, a thousand relationships, no card and no time limit.