Reading Progress Monitoring: What to Track and What to Ignore

Reading generates more data than any other subject and most of it goes unused. Here is what genuinely tells you whether a reader is improving.

Published 2026-08-31 · 4 min read

Reading assessment produces an enormous amount of data — levels, ages, fluency scores, comprehension percentages, phonics checks, running records.

Very little of it changes what happens in a lesson. This is about the part that does.

What you are actually trying to find out

Three questions. Everything worth measuring answers one.

Can they decode the words? The mechanics — phonics, blending, recognising words automatically.

Can they read at a pace that allows understanding? Fluency. A child decoding every word accurately but slowly cannot hold a sentence in mind long enough to understand it.

Do they understand what they read? Comprehension — and this splits into literal understanding and inference, which are different skills that develop separately.

A child can be strong in one and weak in another. Data that gives you a single reading number hides exactly the information you need.

The measures worth taking

Words correct per minute. A pupil reads an unseen passage at their level for one minute. Count words read correctly.

Crude-sounding, genuinely useful. It is quick, repeatable, and sensitive enough to show change over weeks rather than terms. It is the single best indicator of whether an intervention is working.

Take it monthly for pupils you are monitoring. Same procedure, similar difficulty passage, same time of day if possible.

Accuracy percentage. Words correct divided by total words attempted.

Tells you whether the text is at the right level. Below about ninety-five percent accuracy, the text is too hard for independent reading — the child is working so hard on decoding that comprehension is not available.

Retelling quality. After reading, the pupil retells what happened. Rate it against a simple scale you define in advance — for example: no recall, some events out of order, main events in order, main events plus detail and inference.

Quick, and it catches the child who reads fluently and aloud beautifully while understanding very little. That child is easy to miss.

Error analysis, occasionally. Note what kind of errors occur. Substituting visually similar words points at one thing. Guessing from the first letter points at another. Skipping unfamiliar words points at a third.

You do not need this monthly. Once a term per monitored pupil, and it tells you what to teach rather than just how they are doing.

The measures worth deprioritising

Reading age as a single figure. Useful for reporting, poor for teaching. It compresses decoding, fluency and comprehension into one number, hiding which one is the problem.

Levelled book bands as a progress measure. Bands vary between schemes and moving up a band is a judgement rather than a measurement. Use them to choose books, not to track progress.

Comprehension question scores in isolation. A low score might be comprehension, or vocabulary, or working memory, or not understanding the question format. Without knowing which, you cannot act.

Anything assessed once a term. Too infrequent to show whether what you are doing is working. By the time the data arrives the term is over.

How often

Whole class: three times a year is enough. Termly, for reporting and to catch anyone slipping.

Pupils you are monitoring: monthly. Frequent enough to see whether an intervention is working within a half term.

Pupils in intensive intervention: fortnightly. You need to know quickly whether to continue or change approach.

More frequent than fortnightly produces noise rather than signal. Reading fluency varies day to day for reasons that have nothing to do with reading skill.

The structure

One row per pupil per assessment. Date, pupil, words per minute, accuracy percentage, retelling rating, passage level, notes.

A summary showing trend. Latest score, change since last, change since baseline, number of data points.

The trend column is the one you look at. A child at seventy words per minute might be doing well or badly depending entirely on whether they were at fifty or ninety last month.

For a simple trend indicator:

=IF(C2>D2,"Up",IF(C2<D2,"Down","Flat"))

Where C is the latest score and D the previous.

And distance from a target:

=C2-TargetScore

Negative numbers tell you who needs attention. Sort by that column and your intervention list writes itself.

Charting it

One line chart per monitored pupil. Date across, words per minute up. Add a horizontal line at your target.

Four data points on a chart communicate a trend instantly. The same four numbers in a table take a minute to interpret and are easy to misread.

Print it for parent meetings and SENCO conversations. A chart showing a line climbing is the clearest evidence you can offer that something is working.

Reading the data properly

A flat line over six weeks means the intervention is not working. Not that the child cannot learn. Change something — the approach, the intensity, the level of text.

A sudden drop is usually not reading. Illness, a difficult week at home, an unusually hard passage. Do not act on one data point. Look at the next one before concluding anything.

High fluency with low retelling is a comprehension problem masked by good decoding. This child is frequently described as a good reader and is not one. The retelling measure exists specifically to catch them.

Low fluency with good retelling is a decoding problem. They understand when they get there. Work on the mechanics — the comprehension is already present.

Those last two are the most useful distinctions in the whole system, and neither is visible in a reading age.

What changes because of it

After each round, one decision per monitored pupil.

Continue as is. Increase intensity. Change approach. Refer on. Stop the intervention because it has worked.

Write the decision down next to the data. At review you then have not just what the numbers were but what you did about them, which is what anyone reading it actually wants to know.