The panel says 12,400 reached; the Instagram app says 12,910. Which one is right? Both — because they are not asking the same question. This guide lists, one by one, the rule behind every number in the panel and where the gap between the two sides comes from. When you have to defend a figure to your team or your client, this is the page to open.
Every day in the panel starts and ends at midnight in the account's time zone. The rule is the same everywhere: insight days, report days, chart bars and post dates.
That has two consequences:
If the account's time zone differs from your Instagram account's time zone on Meta, the day boundaries are offset and the daily figures on the two sides will not match exactly.
Day boundaries do not drift across a daylight-saving change; the panel shifts days by whole days, not by hours.
When the panel asks Meta for a day's metrics it supplies the day boundaries itself. That removes any ambiguity about which day a number belongs to.
Period windows
Periods are built with one small difference between the two sections:
Insights: the period is a closed interval — the first day and the last day are both inside it. Its unit is the calendar day, because Meta serves its metrics day by day.
Reports: the period is [start, end) — the start is inside, the end is outside. That way two consecutive periods neither overlap nor leave a gap. The screen still writes the last included day.
The comparison window is built the same way in both: the equally long number of days immediately before the selected period.
Where the archive begins
Every number you see in the Insights section is read from the panel's own archive. The archive fills under these rules:
The archive begins on the day you connected the account. The first sync backfills 30 days; nothing older ever arrives.
The archive is never backfilled further. Periods before you connected stay permanently empty.
A day written into the archive is never deleted. Only the last 3 days are rewritten on every run (Meta fills in delayed data afterwards).
Was this article helpful?
The archive's first day is printed on the screen: "Records kept since {day}".
Switching the section off does not delete the archive; it simply is not refreshed during that time, and syncing resumes where it left off when the feature is switched back on.
There is no archive in the Reports section: the numbers are counted afresh each time from the panel's own message, contact, conversation and flow records. That is why Reports is complete all the way back to the account's first day.
The difference between a dash (—) and a zero
This distinction holds throughout the panel and it is the key to reading a figure correctly:
0 — "It genuinely never happened during the period." It is a number and it enters averages and ratios.
— — "Meta never provided this metric." It is an unknown and it does not enter averages or ratios.
The panel never prints a zero for a metric it does not know. So if a card shows a dash, do not feed it into your own calculations as zero.
The same rule governs the change badges: if either period is unknown, no ratio is calculated and "Previous period unknown" is written instead. Treating an unknown as zero would manufacture a "100% increase".
How a period total is built
On the cards in the Insights section a period total is built like this:
The metric is summed across every day in the period.
A day that does not carry the metric does not enter the sum (it is not counted as zero).
If no day in the period carries the metric, the total is a dash.
If even one day carries it, the total is a number — but then it covers only the days with data, not the whole period.
The follower count is exempt from this rule: it cannot be summed, so the last known value of the period is printed.
Post metrics are lifetime figures
The numbers on the Posts tab and in the post analysis are not period totals but lifetime totals. That is how Meta provides them.
The consequences:
The period picker does not filter the post list. Changing the pill does not change the list.
A post's figures keep growing over time; a post from two years ago can still gain reach today.
When comparing two posts, look at their publish dates too: a new post has not matured yet.
Engagement rate is also calculated from lifetime figures: total interactions ÷ accounts reached.
Why the panel's figure can differ from Meta's
There are seven reasons, and all of them are explainable:
Meta's lag (up to 48 hours). After the panel fetches a day, Meta can raise that day's figure. The panel rewrites the last 3 days on every run, but if a day's lag exceeds three days, the gap stays.
Offset day boundaries. The panel uses the account's time zone; the Instagram app uses your account's time zone on Meta. If those differ, "yesterday" is a different 24 hours on each side.
Different window lengths. Instagram's own summaries are usually 7- and 30-day windows; the panel's default is 28 days. 28 and 30 days do not produce the same figure.
Where the archive begins. The panel does not know the days before you connected; Instagram does.
Metric names changed. In 2025 Meta retired the impressions counter and replaced it with Views. Comparing figures from older screenshots with today's is misleading.
Meta withholds some metrics. Newer metrics (such as Reposts) are not recognised on every account or API version; the panel leaves them as a dash. This is a restriction on Meta's side and applies to every tool that uses their interface.
Deleted posts. A deleted post drops out of the panel's list, so totals can look different.
None of these gaps is a fixable bug; each side is correct by its own rules. The rule to follow is: always compare like with like. If you compare month over month, read the panel every month; do not read the panel one month and the app the next.
Exactly what each Reports figure counts
These numbers come from the panel's own records; Meta is never asked.
Card
What is counted
Messages received
Inbound messages recorded during the period
Messages sent
Outbound messages recorded during the period
Sent by automations
Outbound messages whose source is an automation
Sent by agents
Outbound messages whose source is the inbox
New contacts
Contact records created during the period
Conversations opened
Conversations created during the period
Conversations closed
Closed conversations last changed during the period
Runs started / completed / failed
Live flow runs started during the period, and their statuses
Conversions
Conversion events recorded during the period
First response time
The median duration of the waits (seconds)
Three rules apply to the whole table:
Only live runs are counted. Draft tests and simulator runs never enter any figure.
The closing moment is not stored. The "closed" figure is an approximation; adding a note to a closed conversation moves it into the current period.
An automation's message does not count as a reply. First response time only stops when a teammate writes from the panel.
Percentages and rounding
Change percentages are rounded to one decimal place.
Audience shares are also rounded to one decimal place and divided by the whole breakdown (not by the ten buckets on screen), which is why they do not add up to 100%.
Engagement rate is rounded to one decimal place.
Watch times arrive from Meta in milliseconds and are printed in seconds with one decimal.
Report metrics that measure duration are written in seconds, minutes or hours depending on their size.
Amounts are stored to the cent and formatted by currency.
A concrete example
An agency prepares a September report for its client and puts the two screens side by side:
Insights › Last month: reach 84,200; the comparison period (August) 71,500; the badge reads "up 17.8%".
The client looks at "last 30 days" in the Instagram app and sees 79,000.
Three reasons for the gap: the app's window is not all of September but the last 30 days (so it includes the end of August), the day boundaries sit in a different time zone, and the last two days the panel fetched had not filled in yet. The agency makes the right call: it uses one source in the report and states the period explicitly ("1–30 September, panel metrics").
Common mistakes and how to fix them
Mixing two sources in one report. Do not put panel and Instagram figures in the same table.
Counting a dash as a zero. It breaks your arithmetic; a dash means "unknown".
Treating today's figure as final. The day is not over, and Meta's lag applies too.
Equating 28 days with 30 days. Two different windows give two different figures.
Tying post figures to a period. They are lifetime totals.
Reporting on a period before you connected. In Insights those days are permanently empty.
Reading an automated greeting as the "first response". Only a human reply stops the clock.