Engagement heatmaps for X posts

An engagement heatmap puts days of the week down the side, hours across the top, and colors each cell by how well posts published in that slot performed. Done carefully, it answers "when do my posts do best?" in one glance. Done carelessly, it crowns a 3 a.m. Sunday slot because one post happened to take off there.

The difference comes down to four choices: which metric you color by, which time zone you use, how many posts sit behind each cell, and how you word the conclusion. This page goes through each one. For the general mechanics of the chart type, see the heatmap page.

The basic layout

Seven rows, one per day, starting with Monday (or Sunday, if your readers expect it). Columns for hours. Twenty-four hourly columns look precise, but most accounts don't post often enough to fill them, so start with 3-hour blocks and narrow them only if the counts allow.

Use a single-hue sequential scale, light for low and dark for high. A rainbow scale makes readers hunt for meaning in the hue changes, and a diverging scale implies a meaningful midpoint you probably don't have. The color guide has more on picking the ramp. Add a small legend with three or four labeled steps; readers need to know whether dark means 3% or 30%.

Engagement rate vs raw engagements

Raw engagements (likes, replies, reposts, clicks added up) mostly follow impressions. Posts shown to more people collect more engagements, so a raw-count heatmap mostly maps when your audience is scrolling. That is useful, but it's an audience-activity chart, and it shouldn't be read as a performance chart.

Engagement rate divides engagements by impressions. It asks how well a post did for each person who saw it, which makes slots with very different reach comparable.

Both views can be worth showing, side by side and clearly titled. Just don't let a raw-count chart be read as "posts at 6 p.m. are better." They reach more people; whether they work better is the rate chart's question.

Inside each cell, summarize with the median rate rather than the mean. One viral post can drag a cell's mean far above what a typical post there achieves. The tweet performance guide covers that outlier problem in depth.

Time zones

X exports timestamps in a fixed time zone, which has often been UTC rather than your local time. Check the raw values before you build anything: if your morning posts appear at 2 p.m., you are looking at UTC.

Then decide whose clock the chart uses. There are two reasonable answers:

• The account owner's time zone, if the goal is "when should I schedule posts?"
• The audience's main time zone, if most followers live somewhere else. X's audience section may show top locations; use it as a rough guide.

Write the choice in the subtitle ("times in US Eastern"). Also watch daylight saving time. A six-month window that crosses a clock change shifts every post by an hour unless you convert with a proper time zone library instead of a fixed offset.

Minimum sample size per cell

This is where most posting-time heatmaps go wrong. If you post 300 times in six months, that is about six posts per cell in a 56-cell grid on average, and the real distribution is lumpy. Your usual slots have 15 posts; the odd hours have one or two.

Pick a minimum, such as five posts, and grey out or hatch any cell below it. Print the count in small text inside sparse cells, or offer it in a tooltip in interactive versions. If too many cells go grey, use wider blocks (6 hours, or weekday vs weekend) or a longer date range.

Rule of thumb: if you can't say how many posts are behind the darkest cell, you aren't ready to publish the chart.

Reading "best time to post" honestly

Even a clean heatmap shows correlation. Your Wednesday morning posts may do well because that is when you post your weekly data thread, your best format. Move a weak post to Wednesday at 9 and it stays weak.

A few habits keep the reading honest:

• Compare like with like. Build the heatmap for one post type at a time if your formats cluster at certain hours.
• Look for broad bands, such as "weekday mornings," rather than a single winning cell. Bands are more likely to hold up next quarter.
• Report the size of the gap. "3.1% vs 2.4% on a typical weekday afternoon" is honest; "the best time to post is Wednesday at 10" overstates it.
• Test it. Schedule similar posts in two slots for a few weeks and compare.

Where the data comes from

Everything here needs a post-level export: one row per post with its timestamp, impressions and engagements. On X, that comes from the content section of the account's own analytics, which can be exported to CSV from the desktop view. Access to the full analytics dashboard and its export has been tied to a paid Premium subscription and the details change, so check your plan. For several accounts or longer history, use the official X API.

Don't scrape X to fill gaps. It breaks the platform's rules, and scraped data has no impressions, which is the one number an engagement rate can't do without.

Common mistakes