Decision matrix: which WhatsApp conversations get answered after hours, which wait until the next day, and which can't wait
What this matrix is for
Almost every business has a rule for business hours and none for what happens afterward. The result is that each person on the team decides differently: one replies at eleven at night because a price inquiry came in, another leaves everything for the next day, and a third forwards anything to the morning. The customer gets three different standards depending on who they got.
A restaurant that closes at 23:00 receives a message at 23:40 that says "do you have a table for four tomorrow at 21:00?". If no one replies until 10:00, the customer has already booked somewhere else. On the other hand, the same restaurant receives at midnight a "how much is the set menu?": that answer works just as well the next morning, because the customer isn't going to go eat at that hour either. They are two messages that arrived twenty minutes apart and are not answered at the same speed.
This matrix doesn't say what to answer. It says how fast each thing is answered. They are two separate decisions, and it's best to resolve the second one first, because it defines which messages have to be resolved before the day starts and which can wait sitting in the inbox.
The matrix classifies every conversation that comes in outside business hours into three categories: NOW (answered on the spot), TOMORROW (waits until the next support shift) and ESCALATE (not resolved by an automatic reply or a template: a person has to see it, even if it's the next day). The category is decided by two variables: whether the customer can resolve something with just the answer and whether there is a time window that is closing.
This page assumes you already know what your business hours are and that there is a record of what comes in after those hours. If you don't measure it yet, it's best to start by building that record before applying the matrix.
The three categories, with the criterion that separates them
NOW groups the messages where the answer unlocks an immediate action by the customer and the cost of answering late is losing that action. It's not 'everything urgent': it's what the customer can do only if you answer on the spot, and can't do if you answer at nine in the morning. The typical case at a salon: someone writes at 22:30 to confirm the 9:00 appointment the next day. If no one confirms, at 9:00 that person doesn't show up and the slot stays empty.
TOMORROW groups messages where the answer works just as well at nine in the morning as at eleven at night, because the customer won't do anything with that information until the business opens anyway. This is where most of the volume falls: price inquiries, general availability, how to get there, what a service includes. An e-commerce store that receives "do you ship to Rosario?" at 2 in the morning doesn't lose that sale by answering at 9.
ESCALATE groups messages that need human judgment: a complaint, a problem with an order in progress, a negotiation, something that went off script. An automated reply doesn't resolve them, and answering them wrong makes the situation worse. They are escalated to a person, not answered with a template. A real estate agency that receives a message at 23:00 from someone who already visited an apartment and says they found dampness can't resolve that with a business hours template.
The difference between NOW and ESCALATE is the one that gets confused the most. A message can be urgent and not resolvable by a standard reply: that's ESCALATE, not NOW. And a message can be resolvable by a standard reply and not be urgent: that's TOMORROW.
| Category | What defines it | Who responds | When |
|---|---|---|---|
| NOW | The customer can do something concrete with the reply and the window closes within hours | Automated reply or template, with no human intervention | Within the same time slot in which it arrived |
| TOMORROW | The reply works just as well later because the customer can't act until the business opens anyway | Any team member, on the first shift | First service block of the following day |
| ESCALATE | Needs human judgment: a complaint, an ongoing problem, a negotiation, or something off script | A designated person, not a template | The following day, unless the exceptions section says otherwise |
The category is assigned based on the content of the message, not on the time it arrived or who sent it.
The table for classifying, conversation by conversation
This is the table used in the moment. It is read from top to bottom: the first row that matches the message defines the category and stops there. The order matters, because the same message can meet the conditions of two rows and the first one wins.
| # | If the message... | Category | Why |
|---|---|---|---|
| 1 | Reports a cancellation, a delay or a change in a service that is currently in progress | ESCALATE | There is a person affected right now and the automated reply can't resolve the case |
| 2 | Asks to confirm or change an appointment, a reservation or a delivery that occurs within the next 24 hours | NOW | The window closes before the next service shift: if it waits, the appointment has already passed |
| 3 | Reports a problem with a payment, a duplicate charge or a refund | ESCALATE | It involves money and requires verifying the case; a template can make it worse |
| 4 | Asks whether there is availability or stock for something they want to use or buy in the next few hours | NOW | The reply enables or rules out an immediate action by the customer |
| 5 | Asks for a price, a quote or the details of a service with no stated urgency | TOMORROW | The customer can't move forward until the business opens anyway |
| 6 | Asks about hours, address, payment methods or how to get there | TOMORROW | It's information that doesn't change between night and morning |
| 7 | Sends a transfer receipt or asks for details to transfer | NOW | The reply unlocks the payment; if it waits, the customer cools off or pays somewhere else |
| 8 | Writes for the second or third time about the same topic without having received a reply | ESCALATE | The insistence is a sign that the previous reply wasn't enough; it needs a person |
| 9 | It's an inquiry that is already answered in the history of that same conversation | TOMORROW | It's answered with the same information, with no urgency |
| 10 | Mentions an explicit deadline: 'by tomorrow', 'today is the last day', 'expires on Friday' | NOW | The customer set the clock themselves; if the date is within the next 24 hours, it can't wait |
| 11 | It's unclear what they're asking for, or it's a long audio with no transcription available | TOMORROW | Classifying wrong is worse than waiting; in the morning it's read with more context |
| 12 | It's a message from a supplier, a bank or someone who is not a customer | TOMORROW | There's no commercial action that gets lost by waiting |
The first row that matches defines the category. Rows are not averaged or accumulated.
How to apply the matrix without debating every case
- 1Define who is the designated person for the DERIVAR of each time slot. If no one is designated, the category does not exist in practice: everything ends up waiting until the next day.
- 2Write the responses for the AHORA column once and save them as templates. If each person rewrites them at eleven at night, the category becomes slow and inconsistent.
- 3Leave a record in the conversation of which category was assigned. Without that record, you cannot audit later whether the matrix is being applied or whether everyone is doing whatever they want.
- 4Review once a week the DERIVAR that went unanswered. Those are the ones most often lost, because they have no natural owner in the daily flow.
- 5Adjust the criteria with the real cases from the week, not with the ones you imagined when building the matrix. Rows that are never used get removed.
The most common mistake when applying a matrix like this is treating it as a quick-response rule. It is not: it is an assignment rule. What it defines is who each message goes to and with what level of urgency, not what text is sent.
The exceptions that break the matrix
There are four situations where the matrix does not apply, and it is worth having them written down before they come up, because in the moment no one wants to debate the criteria.
- Active campaign: if the business sent a campaign that same afternoon, the messages that arrive are responses to that campaign. A good portion of those responses are MAÑANA, but the volume goes up and it is worth having someone look at the inbox before the day starts.
- Holiday or extended closure: if the business does not open the next day, the AHORA become the ones answered before closing, and the MAÑANA shift to the first business day. The matrix is recalculated, not stretched.
- One-off business event: a promotion that expires that night, a launch, a special date. Messages that mention that date are AHORA even if the table classifies them as MAÑANA, because the clock was set by the business, not the customer.
- Recurring customer with a history of open complaints: if the conversation has an unresolved problem from previous days, any new message from that person is DERIVAR, even if the text looks like a simple inquiry.
Exceptions are documented once and reviewed when they change. If they are improvised every time, they stop being exceptions and become the real criteria, which is what the matrix was meant to prevent.
Common mistakes when building and using the matrix
These are the problems that come up once the matrix is written but still doesn't quite work. Almost none of them are problems with the table itself: they are problems with how it was put into practice.
- Classifying by time instead of by content. A message that arrives at three in the morning is not more urgent than one that arrives at eleven at night. Time is not the variable.
- Putting everything that seems important into NOW. If the NOW category grows until it includes half the messages, it stops being an exception and becomes the rule. The matrix loses its purpose, which is precisely to separate.
- Not assigning anyone to FORWARD. It is the category that falls apart the most because it has no natural moment in the routine. If no one looks at them in the morning, they pile up.
- Answering FORWARD with a generic template to 'not leave them without a response.' It is worse than not answering: the customer understands that there is someone on the other side and expects a resolution that will never come.
- Not logging the assigned category. Without a log there is no way to know whether the matrix is being applied or whether each person interprets it differently.
- Reviewing the matrix only when there is a problem. The criteria become outdated on their own: the schedule changes, the volume changes, the team changes.
How to know if the matrix is working
The matrix is not measured by how many messages are answered outside business hours, but by how many ended up in the wrong category. A useful indicator is the number of conversations that someone has to reclassify in the morning: if that number is high, the criteria are not clear or the rows do not cover the real cases.
The second indicator is the time to first response for NOW messages. If that category exists but the response goes out the same way the next morning, the matrix is written but not applied.
| What is measured | How it is calculated | What a high number indicates |
|---|---|---|
| Conversations reclassified in the morning | Conversations whose category changed between the night and the next day | The criteria do not cover the real cases or are not written in enough detail |
| NOW answered within the window | Messages classified as NOW that received a response before the window ended | The category exists on paper but no one is assigned to cover it |
| FORWARD without an owner | Messages classified as FORWARD that do not have an assigned person | The assignment step is missing; it is the most common mistake when starting |
| TOMORROW answered in the first shift | Messages classified as TOMORROW that received a response in the first service block | The first block of the day is not enough for the volume that accumulated overnight |
| Exceptions applied | Times one of the four exceptions was used during the week | If it is used every week, it is not an exception: it is a row missing from the table |
The five indicators are calculated from the conversation log, not from impressions or estimates.
Where this fits in an inbox with suggestions
The matrix works the same with or without a tool. What changes is how long it takes to apply it. In an inbox where responses are suggested and someone approves them, the NOW category is resolved faster because the response is already written and only needs to be confirmed; the DERIVE ones still need a person, because the suggestion does not replace judgment.
Wando suggests responses over the official WhatsApp API and the team approves them with one click, and it learns from the real responses the business has already given. That helps the NOW and TOMORROW rows have a consistent response without rewriting it every night. The classification into three categories, on the other hand, is a business decision: the team defines the matrix, not the tool.
If you want to see how it applies to a real inbox, you can try Wando for 14 days without a card and load your own conversations to classify them with this table. Current plans and prices are at wando.online/pricing.
Frequently asked questions
What's the difference between an urgent message and a message that gets answered after hours+
Urgent is a perception; the matrix works with a verifiable condition: whether the customer can do something concrete with the answer and whether the window to do it closes before the next support shift. A message can feel urgent and meet neither condition, and in that case it goes to TOMORROW.
How many categories should I use+
Three is enough for most businesses: answer now, wait until the next day, and route to a person. Adding a fourth intermediate category usually generates more discussion than decisions, because in the moment nobody knows precisely which of the two intermediate ones to send a message to.
What do I do if I don't have anyone to cover the NOW category+
If no one is assigned, the category doesn't exist in practice. In that case it's best to reduce it to a minimum: leave in NOW only the messages where the window closes within hours and send the rest to TOMORROW. A small matrix that gets followed is better than a complete one that doesn't get applied.
Are messages that come in from a campaign classified the same way+
Content rules, so most replies to a campaign go to TOMORROW. What changes is the volume: if the business sent a campaign that afternoon, it's best for someone to check the inbox before the first shift so the day doesn't start with a queue that doesn't fit into normal hours.
How do I record the category assigned to each conversation+
With a label per category on the conversation. What matters is that it stays in the same place where the conversation lives and not in a separate spreadsheet, because if you have to open two systems to reclassify, the record gets abandoned in the first busy week.
How often should the matrix be reviewed+
Once a week, looking at the real cases from that week. The criteria go out of date on their own: business hours change, volume changes, the team changes. A matrix that isn't reviewed ends up reflecting the operation from six months ago.
What happens if a message meets the conditions of two rows in the table+
The first row that matches wins, reading the table from top to bottom. That order is there on purpose: cases involving a person affected in the moment or involving money are at the top, and informational inquiries stay at the bottom.
Answer WhatsApp with AI
Try Wando free. No credit card required.