A goal without a baseline is a guess. You can't write "will improve reading fluency" if you don't know today's number, and you can't prove progress at the annual review without a starting line to measure from.
But baseline data collection has a reputation problem. It sounds like weeks of clipboards and stopwatches. It doesn't have to be. Here's the system I actually use.
What baseline data is (and isn't)
Baseline is the student's current performance on the exact skill the goal targets, measured the way the goal will measure it. That's it.
If the goal is "will read 120 words correct per minute," the baseline is today's wcpm, not a reading level, not a percentile, not "below grade level." This is the same-unit rule from how to write measurable IEP goals: baseline and mastery criterion must speak the same language, or you can't compare them.
Baseline isn't a full evaluation. It's a snapshot: small, current, and in the goal's unit.
How many data points you actually need
Three to five, collected across different days. That's the whole answer.
One data point is a bad day (or a good one) masquerading as a trend. Two is a coincidence. Three to five across a week or two gives you a median you can defend at a table when a parent or advocate asks "how do you know?"
More than five is usually procrastination dressed up as thoroughness. See mistake #3 below.
What to measure, by goal type
Match the method to the skill:
Fluency / rate goalstimed probes
One-minute reading probe, math fact probe. Three probes, take the median.
Accuracy goalswork samples
Five problems, ten spelling words: percent correct across 3–4 attempts.
Frequency behaviorstally marks
Count the behavior per period across 3–5 days. A sticky note works.
Duration behaviorsstopwatch
On-task minutes per 20-minute block, averaged across days.
Task completion / independencetask analysis
Percentage of steps completed independently, across 3–4 trials.
The pattern: simple counts, short windows, repeated a few times. Nothing here requires equipment beyond a timer and a sheet of paper.
The 10-minute data sheet
The data sheet that gets used is the one that takes ten minutes to make and thirty seconds to fill in. My rules:
- Student name, date, skill, across the top.If it needs more header than that, it’s too complicated.
- One skill per sheet.Combined sheets get abandoned by week two.
- Pre-fill the measurement column.If you’re counting frequency, the sheet is just dates with blank boxes for tallies.
- Keep it where the behavior happens.The sheet taped to the kidney table beats the perfect sheet in a binder across the room.
If building even that feels like one more thing, use the free data sheet builder: 29 data sheet types, printable, no account. Pick the type that matches your goal's unit and you're collecting tomorrow.
Defining behavior before you count it
For behavior baselines, the definition comes before the data. "Disruptive" isn't countable. Neither is "off-task" until you say what it looks like.
Write the definition so a substitute could count it: "talking to peers without permission during independent work": observable, countable, no judgment words. Then count that exact thing for 3–5 days. The definition goes in the present levels; the counts become the baseline.
The 4 mistakes I see most
1. One data point.
"I gave one probe and he got 45." That's not a baseline, that's a Tuesday. Median of three, minimum.
2. A different unit than the goal.
Baseline in wcpm, goal in "reading level." You can't subtract a level from a wcpm. Same unit both sides, always.
3. Collecting forever, never writing the goal.
Data collection becomes a way to delay the hard part. If you have three solid points, write the goal. You can keep collecting while the draft circulates.
4. Baseline collected after instruction started.
The moment you start teaching the skill, your data is progress monitoring, not baseline. Collect first, then intervene. For annual reviews this matters legally: the baseline has to describe performance before the goal period.
Quick checklist
- Baseline measured in the exact unit the goal will use?
- At least 3 data points across different days?
- Behavior defined in observable, countable terms (if applicable)?
- Data collected before instruction on the goal began?
- Baseline recorded in present levels with dates?
Where this fits
Baseline feeds everything downstream: it's the opening number in your present levels, the starting point of every measurable goal, and the first point on every progress graph. Get it right and the rest of the IEP is arithmetic.
One thing software is genuinely good at: carrying the baseline through. When I built IEP Architect, I made sure goals derive from the student data you paste in. The baseline you measured becomes the goal's starting point automatically, in matching units, instead of something you retype (and mistype) three sections later. Draft the full IEP in about 90 seconds, review it as the professional in the room. First one's free, no credit card.