Celerity Data Studio
Precision that moves at the speed of people.
Overview
Celerity Data Studio is a web-based platform for recording, charting, and analysing behavioural data using the Standard Celeration Chart (SCC) methodology. Data is stored in real time via Firebase — everything you enter is saved automatically and accessible from any device.
Getting Started
The platform opens on the Home screen. The data hierarchy is structured as follows:
Organising Your Library
Everything in Celerity hangs off four levels: a person, the programs that belong to them, the folders inside those, and the charts inside those. A chart cannot exist without a folder, a folder without a program, or a program without the person at the top — so setting up a new case means walking down that path once, and after that you are only ever adding charts.
The person at the top is whoever the charts are about — someone you work with, or yourself if you keep charts on your own behaviour.
Each of the three create forms asks for a name and nothing else that is required. Creating does not open what you created; it appears in the library and you go into it. If you press Create with the name blank, the form will now say so rather than doing nothing.
Renaming is not done in these forms. Any of the four levels is renamed from its row in the library, through a simple prompt. Cancelling it, or leaving the name as it was, writes nothing.
Moving an item copies it to the new location and then removes it from the old one. If the removal fails, Celerity tells you and names both locations, so you can finish the move yourself rather than discovering later that the item is in two places.
Deleting
Nothing Celerity deletes goes to a bin. There is no trash and nothing to restore from, so the only copy of anything you delete is the one you exported beforehand. The three kinds of deletion ask for different things because they destroy different amounts.
Deleted from its row in the library. The confirmation is a single Delete? with Cancel and Delete — it does not name what you picked, so read the row before you click it.
Everything inside goes too. Deleting a folder deletes the charts in it; deleting a program deletes its folders and their charts; deleting a person deletes every program, folder and chart beneath them. There is no separate warning for this and no way back.
Asks for three things: the option to export your data first, your password, and the word DELETE typed out.
It removes your Personal workspace and your sign-in. Charts you made inside an organisation belong to that organisation and stay there — what you lose is your access to them, not the charts.
Asks for the export option, your password, and the organisation’s own name typed out. The name rather than the word DELETE, because an organisation is not only your data.
It removes the whole organisation and everything in it — every member’s charts, every person charted under it — and every member loses access at once. Only an owner can do it, and it cannot be undone.
Administration
Four places where you change your own account, or what other people can reach: account settings, organisation settings, grants, and the data-sharing terms. Two of them use the word role for different things, so it is worth being clear which is which before you use either.
Your own account: profile photo, name, email address and password.
My Role here describes your relationship to the people whose charts you view — Parent / Guardian, Family Member, Client / Patient and so on. It grants nothing. What you are allowed to do is set by your role in the organisation, which only an owner or admin can change, in Org Settings. Nothing in Account Settings affects your access.
The organisation’s members and its shared defaults. Each member row carries their role — Owner, Admin, Member, Viewer or Stakeholder — and that role is what decides what they can do. Remove takes a member out of the organisation and their access with it.
Below the members are settings that apply to everyone: chart defaults, default stat labels, default gridlines, the stakeholder help nudge, and the home banner.
If your organisation has bought a dashboard, its owner also sees a Dashboard section here, with a button to turn it on. It turns on at the newest published version; there is nothing to choose. Turning it on is permanent: your consent is recorded in the organisation’s consent log and cannot be removed. Only the owner sees this section, and only when the organisation has bought a dashboard.
Opened from a member’s row in Org Settings, and headed with that person’s name — grants are set one member at a time.
Grants apply to stakeholders only. A stakeholder with no grants sees nothing at all. Giving a grant to someone whose role is not Stakeholder changes nothing, because grants are only read for stakeholders — the role decides the mode, the grants decide the scope.
A grant names a level: the entire organisation, all programs belonging to one person, a single program, a folder, or one chart. A grant covers everything beneath it. Granting a program grants every folder in it and every chart in those folders, so grant at the narrowest level that does the job.
The terms you agree to if you opt in to data sharing: what is shared, how it is used, how to stop, and how it is secured.
Opting out stops future sharing. It does not withdraw what has already been shared, which is why the decision is worth making deliberately rather than reversing later.
Creating a Chart
Every chart in Celerity lives inside a folder, and folders live inside programs. So the first thing to do is not to create the chart — it is to be standing in the folder where the chart belongs. If the folder does not exist yet, make it first; a chart cannot be created without one.
From there, + Chart opens a short form. Only the name is required. Everything else either has a working default or can be added later, and nothing on this form is difficult to change afterwards.
Chart Type is the one field worth pausing on. It defaults to Daily/minute, and it determines how the chart counts and displays what you record, so it is easier to choose correctly now than to revisit later.
Pinpoint is optional on this form and important to the work. It is the behaviour you are measuring, written precisely enough that another person counting the same thing would produce the same number. You can leave it blank and add it later, but a chart whose pinpoint lives only in someone’s memory is a chart that stops being interpretable the moment that person is unavailable.
Interval Length and Unit appear only if you choose Flex-6 or Flex-5. For every other type they stay hidden, and you can ignore them.
Press Create and the chart is written. Where you end up depends on where you were standing. Create into the folder you are already viewing and the list refreshes with the new chart in it. Create into that same folder while reading one of its charts and you are returned to the folder — the chart you were reading closes. Create from the sidebar while working somewhere else and nothing moves at all: the chart exists, in the folder you chose, and the only sign is the sidebar. In none of these cases does the new chart open; that is always a separate click.
A new chart is empty by design. No data points, no aim, no phase lines. Those come next, and each has its own form.
Chart Types
The platform supports ten chart types. Switch at any time from the toolbar — all data stays intact; the chart simply re-renders on the new scale.
The New Chart form offers nine of them. Monthly/minute is not on that form and is reached by creating the chart as another type and then switching. Every type stays available from the toolbar unless you hide it under Setup → Defaults.
| Type | Y-axis | Typical use |
|---|---|---|
| Daily/min | Per minute | Standard SCC — most PT work |
| Daily | Per day | Low-frequency or count-per-day data |
| Weekly/min | Per minute | Zoom-out view of per-minute data |
| Weekly | Per week | Zoom-out of daily count data |
| Monthly/min | Per minute | Long-term per-minute trends |
| Monthly | Per month | Long-term count trends |
| Yearly | Per year | Multi-year overview |
| Flex-6 | Per interval | Custom interval, 120 periods |
| Flex-5 | Per interval | Custom interval, 100 periods |
| Timings | Per trial | Trial-by-trial timing / latency data |
Equal Interval is not a chart type. It is a view of any chart: use Scale Type in the toolbar to switch between Standard and Equal Interval. Equal Interval draws the same data on an equal-interval axis, for audiences who do not read Standard Celeration Charts.
Data Paths
A data path is one series on a chart — corrects and errors are the usual pair, but a chart can carry up to six of whatever you are counting separately.
Paths matter beyond the chart’s appearance: aims are set against a particular path, so a chart tracking two things can carry a different aim on each.
The path editor appears in two places — its own dialog, and inside Setup under Data Paths. They are the same editor, so a change made in either is simply the change.
Removing a path is not a cosmetic act. The points recorded against it go with it.
Set each path’s Path type. Frequency, the default, records a count and the time observed. Duration records how long one behaviour lasts, and Latency how long it takes to start after the cue. A Duration or Latency path records one event per session: enter its time, and the count is one. The type sets the symbol: \ for Duration, / for Latency.
The chart shows a Duration or Latency point as a frequency, one divided by the time in minutes, so a shorter time charts higher. Use a timed chart type (Daily/minute, Weekly/minute or Timings); a count-per-day chart does not use the time.
Paths made before path types existed are Frequency paths. Nothing about them changes.
Entering Data
- Set the date (defaults to today). Press a tap button once per occurrence.
- Enter the observation time in the duration fields. Tap entry and manual entry read it the same way.
- Click Record Interval to store the record.
- Switch to the Manual tab. Enter date, count(s), and observation duration, then click Add Point.
- The form stays open after each point, so several sessions can be entered in one sitting.
- To record more than one session for the same day, add each as its own point — the chart combines same-date records according to the aggregation setting.
- Data table — a scrollable, inline-editable log of all records. A break marker can be toggled on any row. + Row adds a blank row to this table and Save All writes the edits; both act on the table, not on the Manual tab.
- Paste data — paste CSV text directly. Format: date, count, obs_min per row. For a Duration or Latency path, put its time (30s, 2m, 1.5h or minutes) in the path’s column; the count is set to one.
- Choose file (.csv, .xlsx) — load a CSV or Excel file directly.
- Import Data and Export Data — two buttons: one reads the paste field, the other downloads what is on the chart as CSV.
Import takes comma- or tab-separated text, pasted or loaded from a file. Excel workbooks are read as well; only the first sheet is used.
The first column is the date. A row with an empty date cell, or a date that cannot be read, is skipped. The message at the end tells you how many rows were skipped, so a partial import never reports itself as a complete one.
Slash dates are settled by the whole column, not row by row. 05/06/2026 could be the fifth of June or the sixth of May. Celerity reads the entire date column before parsing any of it: the first value with a component above 12 can only be a day, and that settles the order for the whole file. If nothing in the column settles it, the file is read month-first. Two-digit years from 33 to 99 are read as 1933–1999, and 00 to 32 as 2000–2032.
The first row is treated as headings if its first cell is date, or if any cell in it begins with a letter. With headings, columns named obs, obs_min, min, minutes, observation, session_min or session_minutes are read as observation time and the rest as counts. Without headings the columns are taken by position: the date, then one column per path — or, on charts that record observation time, a count and a duration for each path in turn.
Importing the same dates again keeps both. Every imported row is stored as its own record, so re-importing a range that overlaps what is already there lands the overlapping days a second time. Both rows appear in the data table. On the chart, records sharing a date are combined according to the aggregation setting — under Sum the counts and the observation time are both added, so an exact duplicate plots at the same frequency it did before while the recorded counting time doubles.
Each chart tracks up to six behaviours (paths), each with its own name, colour, and symbol. Paths with no data are hidden automatically.
Sum adds all counts and observation time; First / Middle / Last pick one session by position; Highest / Lowest pick by frequency.
Trend Lines
A trend line is a straight line fitted to data on the log-scale chart, showing the rate of change (celeration) over a selected period. Trend lines are managed in Analysis → Trends.
Celerity automatically detects trend lines from 7 or more consecutive data points moving consistently in the same direction. These appear in the Auto-detected section of the Trends tab. They cannot be deleted but their chart visibility can be toggled with the eye icon.
- Trend — the fitted celeration value. ×1.25/wk means frequency multiplies by 1.25 each week. Acceleration paths show ×, deceleration paths show ÷. A significance label (e.g. Acceptable growth, Robust growth) appears for acceleration paths.
- Bounce — total width of the bounce envelope (see Bounce Envelopes).
- Med — median frequency of data points within the trend range.
| Rate | Label |
|---|---|
| ×1.00–×1.24 | Unacceptable growth |
| ×1.25–×1.39 | Acceptable growth |
| ×1.40–×1.79 | Robust growth |
| ×1.80–×1.99 | Exceptional growth |
| ×2.00–×2.99 | Massive growth |
| ×3.00+ | Super-massive growth |
Theil-Sen (outlier resistant)
Fits the line using the median slope of all data-point pairs. Highly resistant to atypical sessions, and the standard for PT and SCC work. It is the method every trend line uses.
Bounce Envelopes
The bounce envelope shows the typical variability around a trend line — two parallel lines placed symmetrically above and below the fitted line on the log scale. The envelope width is fixed at ±2 standard deviations of log-residuals, the SCC standard.
- Log-residuals are computed for each data point: actual log₁₀(freq) minus predicted log₁₀(freq).
- The standard deviation of those residuals is computed.
- The envelope edge is placed at ±2 SD from the trend line on the log scale.
- At ±2 SD, approximately 95% of data from a normal distribution falls inside the envelope.
Points outside the envelope are considered statistical outliers. The further outside the envelope a point falls, the rarer it is. Distances are expressed in course widths (CW) — the full envelope width from edge to edge.
| Distance beyond envelope edge | Approximate probability |
|---|---|
| 0.5 CW (4 SD from line) | 1 in 1,000 |
| 1.0 CW (6 SD from line) | 1 in 1,000,000 |
| 1.5 CW (8 SD from line) | 1 in 1,000,000,000 |
| 2.0 CW (10 SD from line) | 1 in 1,000,000,000,000 |
The bounce value (×N) displayed in the stat box and hover tooltip is accompanied by a smoothness rating:
| Bounce | Rating |
|---|---|
| ×1.00–×3.00 | Smooth and consistent |
| ×3.01–×6.00 | Bumpy and moderately variable |
| ×6.01–×9.99 | Choppy and inconsistent |
| ×10.00+ | Exceptionally erratic |
Aims & Aim Bands
An aim is the standard a chart is read against. Without one, a chart shows what happened; with one, it shows whether what happened is enough.
Aims are reached from the Aims panel or from Setup → Aims — the same editor in both places. Celerity offers two ways to state the standard, and they sit together without depending on each other.
An Aim Star is a single point: this frequency, by this date. It needs those two things and nothing else — direction defaults to acceleration, and you would change it when the aim is to reduce a behaviour rather than build one. If you turn on the minimum trend line, Celerity draws the rate of change required to reach the aim from where the learner started — a dashed line to the star. It is a standard to read performance against, not a summary of it: it stays where it was drawn while the data moves. The star can be dragged to reposition it whenever the chart is unlocked. If the chart carries several data paths, the aim belongs to whichever one you choose, so corrects and errors can each have their own.
An Aim Band is a range instead of a point: an upper and a lower frequency, with performance expected to stay between them. The upper must be greater than the lower, and the dates are optional — leave them blank and the band runs the length of the chart. A band carries its own colour, and the eye beside it in the list hides or shows it on the chart without deleting it.
Aim Stars are also read by Learning Pictures, and one picture depends on them. Aim — performance maintained at goal — is only reachable when a visible aim star is set on the acceleration path of a linked pair and the median frequency has reached it. Without an aim star, that same flat pattern is reported as Get Truckin’ instead. An aim you have hidden does not count for this.
Both keep you in the form after saving, so a chart can be set up in one sitting, and the chart behind updates as you go. Editing an existing aim loads it back into the form and the save button changes its wording; that wording is how you know whether the next save adds a new aim or replaces the one you opened.
Condition Lines
A condition line marks the point where something about the conditions changed — a new procedure, a different arrangement, a change in how the skill is being taught, a return from a break. It is what turns a chart from one long trend into a record of what you did and what followed.
The way in is the Phases button in the toolbar, and the panel it opens is headed Phases; the form inside it calls what you are adding a condition line. They are the same thing under two names.
The form asks for five things, of which only the date is required. Give it the date the change happened rather than the date you are entering it; the line’s whole value is that it sits where the change did.
The label is optional and worth filling in every time. An unlabelled line records that something changed without recording what, and the chart then tells you less than the person who drew it knew. Once the line is on the chart, its label can be dragged clear of the data whenever the chart is unlocked.
Line Style offers Dashed, Solid and No Line. No Line keeps the label and hides the line — a way to record a change without breaking the visual flow of the chart. Label Orient. sets the label horizontal or vertical, and Line Color sets its colour.
Your own lines are listed at the top of the panel, and that list is where you edit or delete them. Below it, Auto-detected Phases is collapsed by default. These are condition lines Celerity believes it can see in the data, offered as suggestions rather than as your record. The eye on that heading hides or shows all of them together, and it is the same setting that governs the auto-detected trend lines in Analysis → Trends — hidden in one place, hidden in the other.
The word phase appears in two unrelated places in Celerity. A condition line marks a change in what you did — it lives on the chart. Analysis separately names the periods between its own trend lines, and it is those periods that Change Across Trends compares. Adding or moving a condition line does not change that table.
Adding a line keeps you in the panel, so several can be entered in one sitting. The chart behind updates as each is saved.
Notes
A note is an annotation on the chart: a reason, a caveat, something a reader needs that the data does not say — an intervention, an absence, an observation. Only the text is required; everything else is appearance.
Open them with Notes in the toolbar. Enter the text, then set Text Color, Outline Color, and a Background of White or Transparent.
Direction and Alignment exist so a note can sit alongside a condition line or run across a period without covering the data it is explaining. Direction offers Horizontal, Vert Up and Vert Down; Alignment sets the text Left, Center or Right inside the note.
+ Add Note places it. Every new note lands in the same spot — near the top left of the chart — and you drag it from there to where it belongs, which you can do whenever the chart is unlocked. That starting position is worked out from the chart’s current dimensions, so a note added before the chart has drawn can land somewhere unintended; open the chart first and it will land where you expect.
Existing notes are listed above the form. The eye beside each one hides or shows it on the chart without deleting it, and ✎ and ✕ edit and delete it. A practitioner has a second control that the stakeholder view does not show: it marks an individual note as hidden from stakeholders, so a note can stay on your chart and off theirs.
Analysis Modal
The Analysis modal is the hub for all statistical analysis. Open it with the Analysis toolbar button. It has four tabs, each independently controlled.
- Add manual trend lines, view auto-detected trend lines, and manage Improvement Index (I.I.) pairings.
- The eye icon on each trend line controls on-chart visibility only — it does not affect whether the line feeds into the other tabs.
- + Add I.I. Pairing — pairs one acceleration path and one deceleration path to compute an overall accuracy index. The I.I. stat box appears on the chart.
Compares consecutive trend line periods, showing how rate of change, frequency level, and variability shifted between phases. Each comparison generates a stat box on the chart.
- Celeration turn — did the rate of change speed up or slow down?
- Frequency jump — did the level shift up or down at the phase boundary?
- Level change — the ratio of median frequencies between phases.
- Bounce change — did variability increase or decrease?
- I.I. Change (I.I.C.) — how did overall accuracy shift between phases?
Eye icons control on-chart visibility of each comparison's stat box only — they do not affect the Summary Table.
Detects 7-point runs — seven or more consecutive data points all moving in the same direction. These are statistically significant trends that may indicate a real change in performance. Runs are listed by path with date range and direction.
A structured table of all stats across all trend line periods — trend values, bounce, median frequency, I.I., phase comparisons, and more.
- Use the eye icons to show/hide individual periods on the on-chart overlay.
- Export as SVG (best for print), PNG (screen image), or CSV (spreadsheet data).
| I.I. Value | Label |
|---|---|
| Any ÷ value | Worsening — accuracy declining |
| ×1.01–×1.20 | Very slight improvement |
| ×1.21–×1.30 | Small improvement |
| ×1.31–×1.50 | Adequate improvement |
| ×1.51–×2.00 | Substantial improvement |
| ×2.01–×3.00 | Exceptional improvement |
| ×3.01+ | Extraordinary improvement |
Hover Tooltips
Hovering over the chart reveals live statistics. Two tooltip types are available on desktop (not mobile).
Shown when the cursor is within 14 pixels of a data point’s centre. Displays:
- Path name (in the path colour)
- Date, count, and time observed
- For outliers: distance beyond the bounce envelope in course widths, and the associated probability
Shown when the cursor is within the bounce envelope band. Displays:
- Path name (in the path colour)
- Trend value with Acceleration or Deceleration label, and significance label (acceleration paths only)
- Bounce value and smoothness rating
Data point tooltips take priority when the cursor is near a point.
Decision Support
The Decision Support panel automatically analyses your chart data and surfaces performance flags. Open it with the Decision toolbar button.
- The aim has been met in 2 of the last 3 sessions — consider moving to the next step.
- Performance has decreased in each of the last 4 or more sessions in a row
- Performance has increased in each of the last 4 or more sessions in a row (for deceleration paths)
- Performance has stayed flat across 4 or more consecutive sessions
- Acceleration rate below minimum (×1.25) — current rate shown
- Deceleration rate below minimum (÷1.25) — current rate shown
- Performance has fallen below the trend line in each of the last 3 sessions
Learning Pictures
Learning Pictures are named performance patterns detected automatically when paired paths are present — one acceleration path and one deceleration path linked together in Setup → Data Paths. Detection uses the last two trend line periods of data and fitted trend lines (not individual data points).
| Picture | Category | What it means |
|---|---|---|
| Crossover | Improving | Corrects accelerating past errors — trend lines are intersecting |
| Jaws | Improving | Both paths moving toward aim and diverging — not yet intersecting |
| Take-off | Improving | Corrects accelerating while errors hold flat |
| Climb | Improving | Corrects accelerating while errors are eliminated (at floor) |
| Uphill | Improving | Both accelerating with corrects dominating — increasing % correct |
| Dive | Improving | Corrects hold steady while errors reduce |
| Aim | Maintaining | Both paths flat and at aim level — performance maintained at goal |
| Get Truckin' | Maintaining | Both flat and consistent but below target |
| Rock Bottom | Maintaining | Both paths stalled at floor frequency |
| Snowplow | Worsening | Both moving wrong direction and lines converging |
| Landing | Worsening | Corrects declining while errors remain neutral |
| Surface | Worsening | Errors growing while corrects stay flat |
| Downhill | Worsening | Both decelerating with errors dominating — decreasing % correct |
When a Learning Picture is detected and the Decision Support panel is open, thick solid trend line highlights draw on the chart for each paired path — covering the most recent trend period plus one period forward. Use the eye icon next to each picture name to show or hide its highlight.
Decision Guide
The Decision Guide walks practitioners through a step-by-step decision tree for either Skill Acquisition or Generalization. Access it from the Decision Support panel by clicking Open Decision Guide.
Completed guides are listed in the Decision Guide modal under their tree (Skill Acquisition or Generalization), sorted newest first. Each row shows the date, recommendation, and chart type. Click View to review the full Q&A path in read-only mode.
Stakeholder Mode
Stakeholders are members granted read-access to specific charts. When logged in as a stakeholder, the interface is simplified to show only what is relevant for reviewing progress.
- View the chart and all its data
- Open the Notes modal — add notes, edit and delete their own notes
- Open the Decision Support panel and view Learning Pictures
- Change the chart scale (current type) and timespan for their own viewing session
- Download the chart as SVG or PNG
- Enter, edit, or delete data points
- Add or edit trend lines, aims, or phases
- Open Setup or Analysis
- Edit or delete notes created by the practitioner
- Toggle the chart lock state
Chart Display Options
- Chart scale — switch chart type and start date from Row 2 of the toolbar.
- Aggregation — how same-session records are combined, from Row 2.
- Gridlines — All / Major / None, set per-chart in Setup → Defaults.
- Dark/Light mode — toggle from the Account panel.
- Download SVG — scalable vector format, best for print and reports.
- Download PNG — raster image, best for embedding in slides or documents.
Landing Page & Navigation
The landing page is the entry point to Celerity Data Studio. It has four tabs — Home, Platform, About, and Contact — before and after signing in.
Shows the sign-in form when logged out. Once signed in, the form panel is hidden and the full brand panel is displayed with an Open Platform button. Click it — or click the Celerity logo — to return to the app.
Click Home, About or Contact to open that page of the Celerity website. The Contact page gives the email address for help and questions. To sign in again, click Platform in the menu on the website.
- Clicking the Celerity logo on the About or Contact page of the website opens the Home page.
- Clicking the Celerity logo in the app's home screen hero returns to the landing page.
- Clicking the logo in the top bar of the app returns to the landing page.
The platform is fully responsive. In portrait orientation the landing page displays the brand panel full-screen. When logged in on mobile the login form is hidden and the brand panel fills the screen. In landscape orientation the chart area and toolbar are accessible with a compact hero panel.
Calculation Methods
This section documents the exact statistical methods Celerity Data Studio uses to compute celeration, bounce, median, and data aggregation. All calculations operate on log₁₀-transformed frequency values, consistent with the ratio-scale nature of the Standard Celeration Chart.
Aggregation
Pre-processingAll statistics are computed on aggregated data — raw records are first bucketed into time slots matching the chart type. This ensures that stats reflect exactly what is displayed on the chart.
bucket = round((pointMs − t₀) / 86,400,000)
// Weekly — point snapped to Sunday first
bucket = round((sundayMs − t₀) / (7 × 86,400,000))
// Monthly
bucket = (year − startYear) × 12 + (month − startMonth)
// Yearly
bucket = year − startYear
// Flex (custom interval)
bucket = round((pointMs − t₀) / intervalMs)
freq = count / obs
// Count charts (Daily, Weekly, Monthly, Yearly, Flex-5/6)
freq = count
Median Frequency
DescriptiveThe median is computed on the aggregated frequency values for each path within the selected celeration range.
median = freqs[n/2] // middle value if odd count
median = (freqs[n/2 − 1] + freqs[n/2]) / 2 // mean of two middle if even
Median is reported in the chart's native frequency units (e.g. /min, /wk).
Theil-Sen Celeration
RecommendedTheil-Sen is a robust, non-parametric line-fitting method. It is highly resistant to outliers — a single unusual data point cannot disproportionately pull the celeration estimate. It is the recommended method for all PT and SCC work.
slopeMs = median(slopes) // middle value if odd; mean of two middle if even
intercept = median(intercepts)
Bounce (Envelope)
VariabilityBounce quantifies the typical spread of data around the trend line. The envelope is placed at ±2 standard deviations of log-residuals — the SCC standard. The bounce value reported in the stat box is the total envelope width as a multiplicative factor.
SD = √variance
bounceVal = 10^(2 × bounceOffset) // total width as ×N multiplier
| Chart Type | Celeration Period | Label |
|---|---|---|
| Daily/min, Daily | 7 days | /wk |
| Weekly/min, Weekly | 5 weeks | /5wk |
| Monthly/min, Monthly | 6 months | /6mo |
| Yearly | 5 years | /5yr |
| Flex-5, Flex-6 | 5 intervals (Flex-5) or 6 intervals (Flex-6) | /5d, /6wk … built from the interval |
Best Practices
- Measure daily. The SCC is designed for daily data. Gaps reduce the reliability of celeration estimates.
- Keep observations consistent. Use the same duration each day when possible.
- Use fair pairs. Track both the target behaviour and an incompatible alternative.
- Use clear pinpoints. "Reads words correctly" is better than "engages with text".
- Start a new cel line at each phase change.
- If frequency is not moving, change something. A flat celeration after 3–5 data points is a signal to adjust.
- Use Daily/min as your primary chart. The traditional SCC runs Mon–Sun for 20 weeks.
- Celeration targets: ×1.25 to ×2.0/wk is common for skill-building. Above ×2.0/wk is a steep learning rate.
- Bounce benchmarks, as the field reads them: below ×2 is tight, ×2–4 is typical, above ×6 suggests the programme needs attention. These are the practice conventions, not the bands Celerity labels with — the app’s own thresholds are in the Quick Decision Reference below.
- Leave atypical days in. PT data frequently includes them, and the trend line is fitted with Theil-Sen — one unusual session moves the line very little. An outlier is information about what happened, not noise to remove.
- Read the chart first before checking the stats box.
- Download SVG for reports.
These are the app’s own thresholds — what Celerity will actually label and flag on your screen, so nothing here should surprise you when you see it. The field’s benchmarks above are read differently and deliberately kept separate.
| Situation | Action |
|---|---|
| Trend right direction at target rate (×1.25+) | Continue — consider Decision Guide to confirm |
| Trend right direction but too slow | Increase intensity or try a better technique |
| Trend flat (below ×1.25 — what Decision Support flags) | Change something — materials, prompting, timing, or reinforcement |
| Trend wrong direction | Stop and analyse before continuing — run the Decision Guide |
| Bounce above ×3.00 (Celerity labels ×3.00–×6.00 bumpy, ×6.01–×9.99 choppy, above that exceptionally erratic) | Look for environmental variability (inconsistent sessions, distractions, illness) |
| Aim met 2 of last 3 sessions | Decision Support panel shows Advance flag — move to next step |
| Review flag in Decision Support | Open Decision Guide for a structured recommendation |
- Open the Decision panel regularly. It checks all paths for advance and review criteria automatically.
- Pair your paths. Link acceleration and deceleration paths in Setup → Data Paths to enable Learning Picture detection.
- Name the picture. Identifying the current Learning Picture (Crossover, Jaws, Dive, etc.) helps communicate progress to teams and families.
- Use the Decision Guide. When a review flag appears, run the guided decision tree — it uses your chart data to auto-suggest answers and produces a saved recommendation.
Help AI
The Help AI is an in-app assistant accessible from the 💬 Help button in the top bar. It can answer questions about Celerity Data Studio features, explain Standard Celeration Chart concepts, and guide you through interpreting your stat box values.
- Click 💬 Help in the top bar. A panel opens in the bottom-right corner of the screen.
- The panel shows suggestion chips — topic categories to help you get started quickly.
- You can also type any question directly in the input field and press Enter or click Send.
- Close the panel with the ✕ button. The conversation is preserved while the chart is open.
When the panel first opens, five category chips are shown. Tap a category to see specific question options:
- 📈 Reading my numbers — interpret your stat box values, understand significance labels, read the stat box
- 🔧 Setting up — add data, add trend lines, set aims, add phase lines, set up paired paths
- 🔍 Analysis tools — Change Across Trends, Patterns, Summary Table, Improvement Index
- 🧭 Decision tools — Decision flags, Learning Pictures, Decision Guide
- 📖 Concepts — celeration, bounce, the Standard Celeration Chart, fair pairs, Precision Teaching
Each category shows a ← Back chip to return to the top level. Chips disappear once you send any message. Click ↺ Clear to reset the conversation and restore the chips.
The Help AI does not read your chart values automatically — you share them, and it interprets them. To get an interpretation:
Even without reading stat values, the Help AI receives context when a chart is open:
- Chart type — used to convert celeration periods into real time (e.g. each period = 1 week on a Daily chart)
- Data path names and roles — used to tailor language to your domain
- Pinpoint — the behaviour description from Setup
- Data domain — set in Setup → Data Paths; enables domain-adaptive language for Learning Pictures and interpretation
- Detected Learning Picture — when the Decision panel is open, the AI knows which pattern was detected
Do not type names, birth dates or ID numbers into Help. As a safeguard, before your question is sent:
- The name of anyone in this organisation's library, or named in the open chart's Setup, is replaced with a placeholder such as [Learner 1]. The same person keeps the same placeholder for the whole conversation, and the real name is put back into the answer you see. A line under your question says how many names were replaced.
- A name is caught written in full, as a first name or surname on its own, as an initial with a surname, or in lower case. A misspelling is caught only in a name part of eight letters or more. A name that is also an ordinary word, such as Will or Grace, is caught only next to the person's other name.
- Email addresses, phone numbers, birth dates and ID numbers are removed.
- The placeholders are forgotten when you click ↺ Clear or switch organisation.
- Each Help AI conversation is limited to 10 exchanges — shown as a counter in the panel header (e.g. 3/10 exchanges).
- When the limit is reached, click ↺ Clear to start a fresh conversation.