At 2 p.m., 90 grams of protein might be perfectly normal for someone who eats a large dinner. Later in the evening, the same number could mean they’re unlikely to reach their target. The number hasn’t changed. The context has.
Bodro handles this by forecasting where the day is likely to end. The forecast uses local time, personal history, and the way each person usually distributes food and activity across the day.
This article explains how that model works and how Bodro turns it into useful, timely feedback.
01
A daily target describes the desired result for a completed day. It doesn’t say when those calories, nutrients, or active calories need to accumulate.
People follow very different schedules:
- some barely eat breakfast;
- lunch or dinner may account for most of the day’s food;
- workouts happen before work, in the afternoon, or late at night;
- some people are done moving hours before midnight.
That’s why a simple comparison between “percent of target completed” and “percent of day elapsed” often gets the situation wrong. Completing 50% of a target after 70% of the day has passed doesn’t automatically mean someone is behind. The remaining 50% may normally happen in the evening.
Bodro treats the target and the path toward it as two separate things:
- the target is the result the user chose for the end of the day;
- the trajectory is how that result usually develops over time.
The current day is evaluated against the trajectory—not against the assumption that progress should be linear.
02
For nutrition, Bodro looks at the previous 28 completed local days. For each metric, it selects days that finished within the final target requirements and looks at how much progress usually happened after the current time.
From that history, Bodro builds two meaningful end-of-day forecasts.
THE TYPICAL FORECAST
The first forecast uses the median of successful days.
It answers a simple question: Where will today end if the rest of the day looks roughly like a typical successful day?
The median isn’t pulled around by one unusually large dinner or one exceptional day. It represents the person’s usual pattern.
THE BEST REALISTIC FORECAST
The second forecast uses the best 10% of successful outcomes.
This is not a theoretical best case. It represents something the person’s own history shows is still realistic.
“Best” depends on the metric:
- for protein, fiber, and activity, it means a stronger finish;
- for fat, it means a more restrained remainder of the day;
- for calories and carbs, it means the realistic path that can still finish inside the allowed range.
Under the hood, those cases use different sides of the statistical distribution. In product terms, there are still only two forecasts that matter: the usual outcome and the best realistic correction.
Bodro switches to a personal forecast after it has at least seven usable days for a metric. Until then, it uses a general time-of-day model based on common daily eating patterns. The personal model takes over as enough history becomes available.
03
In Bodro, the acceptable range describes how far the forecasted end-of-day result may be from the selected target. It does not define how much of the target should already be completed at a particular hour.
The range is very wide at the beginning of the day. Most people are asleep for much of that period, almost no current-day data is available, and many normal outcomes are still possible. A morning forecast therefore does not need to land precisely on the target. It can miss by a meaningful amount without triggering a warning.
After roughly the first third of the day, when the active part of the day usually begins, that range starts to tighten. As evening approaches, the typical forecast needs to land closer to the selected target. A forecast that is acceptable in the morning may signal risk in the afternoon because less time remains for the day to correct naturally.
During the final two to two and a half hours, the early-day flexibility is gone. The typical forecast needs to satisfy the final requirement. If only the best realistic forecast can still do that, the day is already considered at risk.
Final requirements also depend on the type of metric:
- protein and fiber need to reach a minimum;
- fat needs to stay below a maximum;
- calories and carbs need to finish within a range around the target;
- active energy needs to reach a minimum.
ACCEPTABLE FORECAST RANGE
04
How Bodro assigns a status
The two forecasts are checked against the acceptable range for the current time.
ON TRACK
If the typical forecast finishes inside the range, no correction is needed. Following the person’s usual pattern is likely to produce the selected result.
AT RISK
If the typical forecast misses the range but the best realistic forecast still reaches it, the day is at risk. The usual routine is no longer enough, but a correction that has happened before is still available.
OFF TRACK
If even the best realistic forecast misses the range, the day is off track. The same applies when a limit has already been irreversibly crossed—for example, when actual fat intake is already above the current maximum.
For one-sided goals, there is also a completed state. Once protein or fiber has already reached its target, Bodro no longer needs to forecast whether it can be reached.
This distinction matters because “needs attention” and “can no longer be fixed through a normal adjustment” are not the same situation.
05
Suppose someone has a daily protein target of 180 grams.
At 2 p.m., they have logged 90 grams. That is only half of the target, but on successful days they typically add another 95 grams after 2 p.m. The typical forecast is therefore 185 grams. There is no reason to flag a shortfall.
Now consider a different day. At 7:30 p.m., they have logged 117 grams. Their typical forecast ends at 129 grams, while their best realistic forecast reaches 191 grams.
7:30 P.M. · 117G LOGGED
129g
Typical forecast
191g
Best realistic forecast
The typical path no longer reaches the target, so the day is at risk. But a stronger finish that has appeared in their successful history can still get them there. If that second forecast were also below 180 grams, the day would be off track.
The current number only becomes meaningful when it is combined with time and personal history.
06
Activity follows the same overall logic—two forecasts and a range that tightens throughout the day—but its personal trajectory is built differently.
Bodro uses hourly active-energy data from connected health-data sources across the previous 28 completed days. That history shows how much activity usually remains after the current time.
The daily baseline is intentionally conservative. Bodro uses the lower quartile of historical daily totals as a personal “quiet day.” A few unusually active days therefore do not make the forecast unrealistically optimistic.
The hourly pattern becomes especially important late in the day. If someone almost always stops being active by 8 p.m., Bodro should not assume later in the evening that they will keep moving at an average rate until midnight. The typical and best realistic remainders come from their actual time-of-day history.
The remaining forecast also decreases smoothly within the current hour. It does not wait for the next hourly boundary or change in a sudden step.
07
Where this calculation is used
The model supports several parts of Bodro:
- nutrition and activity highlighting;
- the current day’s preliminary score;
- explanations of what affected that score;
- the live context provided to the agent;
- proactive cards about risks and realistic corrections.
All of these surfaces receive the same server-calculated facts, targets, forecasts, and status. The app and the agent should therefore interpret the same situation in the same way.
The forecast itself is statistical and deterministic. A language model does not decide whether someone is on track. The agent receives the calculated context and can explain it or help the user decide what to do next.
08
What the model is for
The model is not trying to predict every action for the rest of the day. Plans change. Someone may add an unplanned workout, move dinner later, or deliberately depart from their usual routine.
Its job is to provide a realistic answer to three questions:
Will the usual rest of the day lead to the selected targets?
Is a correction needed?
Is that correction still realistic based on personal history?
This avoids treating a normal morning or afternoon pattern as a problem. It also avoids waiting until the end of the day to point out something that could have been corrected earlier.
Targets in Bodro are set by the user. They are not medical standards, and the model is not a substitute for professional advice. It is a way to understand, while the day is still in progress, whether the user’s current trajectory still matches the strategy they chose.
OPEN BETA
Try Bodro early.
Request an invite to try it with your own data.
You’ll receive occasional product updates as Bodro evolves.