Read your reports

Open Reports. The section has six views: Now, Demand, Service, Team, Broadcasts and Ads. The period and channel you choose travel with the tab; a figure without those filters is no good for comparing.

Demo organization: Casa Nativa. Interface shown in Spanish.

NowCopy link

This is the live view. It answers four questions, each opened by the answer in one sentence:

• How long have they been waiting? How many people are waiting for a reply now, by how long, and for how many on WhatsApp the 24-hour window has closed. It lists those waiting longest.
• Who has them? How many wait with no one assigned and how many each person or AI agent has.
• Is anyone answering right now? When the last reply was, who has people waiting but hasn't replied for 2 hours or more, and how many replies each one sent today.
• Is today a normal day? How many people wrote today so far, against the same weekday at the same time over the 8 previous weeks. Within ±15% it's a normal day.

A person is waiting when their conversation is open and their last message, not counting reactions, has no reply. Closed conversations don't wait: closing one decides no reply was needed.

Use it to split up today's work. Since it changes as messages arrive, treat the figure as a snapshot of the moment, not as the period's result.

DemandCopy link

It answers one question: how many people are writing to you, when and through where. Each block opens with the answer in one sentence and the evidence below it. The sentences come from fixed rules over the numbers, not from an AI: something is "normal" when it is within 10% of its reference.

A person is a conversation with at least one customer message in the period; they are new if they wrote for the first time in the period.

• Summary. How many people wrote, how many were new, how many already knew you and how many came from an ad, against the previous period. When the period ends today, the day is still in progress and is never called below normal.
• Is it going up or down? The period next to the 8 equal periods before it, and each day against the average of the same weekday over the 8 weeks before (from 4 weeks of history).
• When do they arrive? The share that arrived in the morning, afternoon, evening and overnight by the hour of their first message, the hour with most arrivals, the busiest moment and the day furthest from normal. The map shows how many people wrote in each hour, Panama time.
• Where do they come from? The channel they write on and whether they came on their own or from an ad, with where the ad was seen (Instagram or Facebook) per the link Meta sends. What each ad cost is in Ads.
• What are they looking for? The most repeated first messages (many are prewritten by the ad) and how many people ask for prices in their first message. Greetings alone don't count.

ServiceCopy link

Answers one question: are we serving the people who write to us well? Each block opens with the answer in one sentence and the evidence below. Sentences come from fixed rules over the numbers, not from an AI.

A time is one customer message, or several in a row, until someone answers. The wait counts from the first of those messages. Answers come from a person in Ricky, an AI agent or the channel's app; automations, templates and reactions are not answers. A person is a conversation.

• Summary. How many people wrote and how many times, the typical wait (median) against the previous period, how many people waited more than 1 hour at least once and how many are still without a reply. One day is compared with the same weekday over the 8 previous weeks; a longer period, with the same number of days right before.
• How fast do we answer? The share answered within 5 minutes, 1 hour, 1 day or still unanswered; new people's first answer; and who answered: the AI, people in Ricky or the phone app.
• Whom did we leave without a reply? People whose conversation is still open and whose last message has no answer, by channel; how many WhatsApp ones are past 24 hours (only a template can reach them), how many came from an ad and how many never got an answer. Conversations someone closed are not counted: closing one means deciding no reply was needed. What someone wrote less than an hour ago does not count yet.
• When do we serve worst? The hour when most people waited more than 1 hour or got no reply, and the best one. For one day, bars per hour; for several days, a weekday-by-hour map; from 15 days, also the typical wait day by day.
• Whom do we serve worst? Compares channels, new people against people who already knew you, and people who came from an ad against those who came on their own. A group is served worse when its median is at least 25% longer.
• Do we resolve what they ask for? How many conversations were closed, how long they were open and how many people have their conversation closed today, with the closes of every 7 days.

A block can add a What you can do when the numbers call for it.

TeamCopy link

Answers how each person and the AI work. Each section opens with its answer in words: how the work is shared (times each one answered, to how many customers, and how many conversations they closed), how conversations customers open get assigned (how long until someone has them, how many are still unassigned and who assigns), who answers fast and who slowly, when each one starts and stops, who has customers without an answer, and how much the AI helps. A time is the answer to a customer message, or to several in a row, and counts for whoever sent it; automations and templates do not count. Fastest and slowest are chosen only among people with 20 answers or more. The What you can do suggestions come from fixed rules over those numbers.

How often someone answers shows activity, not quality. Before evaluating someone, consider whether they handle long inquiries, several channels or cases they received from someone else.

BroadcastsCopy link

Summarizes the period's broadcasts: broadcasts and recipients per day, the result per recipient and how many were read. A sent message doesn't confirm the customer received it; check the details before resending, so you don't duplicate the ones that did arrive.

AdsCopy link

Answers five questions about your Meta ads, each with a sentence and what shows it. A person comes from an ad when they wrote from that WhatsApp or Instagram ad; if they came back through another ad, they count for both, but once in the total.

• How did it go? What you spent, how many people wrote to you and what each one cost (spend divided by people), against the previous period and why it changed: whether you spent more or less and whether each person came out cheaper or dearer (from a 5% difference). Below, from ad to chat: how many people saw your ads (Meta's reach, without repeating people), how many tapped them and how many wrote; and spend, people and cost per day (up to 14 days) or per week, leaving out today, because Meta sends its spend later.
• Where does your money go? Each campaign with its share of the money next to its share of the people, and what each person cost there. It names your best campaign when it brought 20 people or more with at least 10% of the money and costs 10% less than your average, or less.
• Which ads work? Three groups of running ads, each with its image; tap one to see all its numbers. Working: they cost 25% less than your average, or less, with 10 people or more. Expensive: they cost 1.5 times your average or more, or spent what 5 average people cost with nobody writing. Getting dearer: each person cost 20% more in the second half of the period than in the first, with 10 people or more in each half. An ad is judged only with 5 people or more and spend of at least 5 average people. Expensive ads you already paused are mentioned apart.
• Who and where does it work for? A map by age, women and men, with what each person cost: cheap in green and dear in red, only where there are 10 people or more. And where your ads showed (feed, reels, stories, status…) with their spend and cost per person. The sentence names the audience that writes the most (with 50 people or more in total) and the cheapest placement against the dearest, when the gap is 20% or more.
• What should I do now? Up to three actions with their money: pause or change the expensive ads (with what they spent and how many people they would have brought at your average cost), give more budget to the ones that work and renew the one getting dearer.
• All your ads, sorted by spend, people or cost, downloadable as CSV.

When there are too few people to say something, the report says so instead of drawing conclusions. Age, sex and placement come from Meta: Ricky cannot know where someone saw the ad or their age, so those parts use Meta's conversation count, which can differ slightly from Ricky's. Images are Ricky's own copy of each ad, so they do not expire. Without Meta Ads connected the report shows how many people wrote from each ad, without spend, reach, placements or audience.

Spend uses your ad account's days and people use Panama time, so the edges of a period can differ by a few hours.

AppointmentsCopy link

Shown only to businesses that connected Odoo with their appointment agenda. It crosses Odoo's appointments with Ricky's conversations to answer how many appointments were booked, at which location and for which services, how many come from people who wrote through Ricky and from which ads, how many people who write end up booking, whether people come to their appointments, and who books them.

A booked appointment is one created in Odoo during the period, whatever its date. An appointment comes from Ricky when its phone wrote to Ricky on WhatsApp in the 14 days before it was booked; that says there was a conversation, not that the conversation caused the appointment. Do people come to their appointments? instead counts the appointments whose date fell in the period and has passed; unmarked means nobody marked in Odoo whether the person came or it was cancelled. A period is compared with the same number of days right before it; a single day, with the average of the same weekday in the 8 weeks before. Ricky only reads the Odoo agenda, it never changes it.

Review routineCopy link

1. Choose the period and the channel.
2. Start with Now to handle what's waiting.
3. In Demand, spot peaks and compare the team's coverage.
4. In Service, check who is still without a reply and when service is worst; a low median doesn't explain cases with no reply.
5. Review broadcasts with problems.
6. Adjust assignments, instructions or hours based on what you find.

If a figure looks wrong, note the tab, the period, the channel, the metric and a concrete example. Include those details when you ask for help.

To receive these results on a schedule, set up Email reports.