The declaration
What the people doing the work say is left.
- Asks
- The team, for remaining time.
- Computes
logged + remaining- Fails when
- Remaining estimates are missing or stale.
- Blind to
- A tool that tidies the field for the team.
Your project has two forecasts. One of them is lying.
Freeze a won estimate into a commitment, measure it against the hours your team logs in Jira or ClickUp, and read the outcome twice: what the team declares, and what the finished work proves.
No credit card. Connect one project that worries you and let a night of worklogs answer the question.
Nightsky Commerce · baseline v2 · fixed priceDemonstration data
Healthy
Inside the commitment, or under the first threshold.
At risk
First threshold crossed. One notification, once.
High risk
Second threshold crossed, with the money figure.
Insufficient data
Nothing measurable came back. A state, not an error.
The most confident number in your delivery tool is the least maintained one.
Jira ships with Adjust automatically switched on, so every worklog is subtracted from the remaining estimate.
A team that logs hours and never re-estimates therefore produces exactly this, for the entire life of the project:
remaining = original − spentlogged + remaining = the original estimateDeviation reads 0.00%. Severity reads healthy. The margin stays green until the tasks start closing and the truth arrives all at once — usually after the last invoice anyone could still change.
What a single-forecast report shows — same project, same hoursDemonstration data
So we ask twice, and put the answers in front of each other.
One forecast trusts your team. The other trusts nothing but the part of the plan that is already closed. Neither is authoritative alone, and that is the point.
The evidence
What the finished work implies about the rest.
- Asks
- Nobody.
- Computes
logged ÷ closed share of plan- Fails when
- Too little is closed, or statuses are unmapped.
- Blind to
- Work spread unevenly across the plan.
They agree.
Confidence you earned rather than assumed, because two methods with different failure modes landed in the same place.
They diverge.
A finding in itself. The screen names which forecast is thin, what the gap is worth in money, and what to look at first.
What we never do is average them into one health score. A blended number is one nobody can challenge and nobody can act on, and it buries the very disagreement you needed to see.
Estimating is step one of five. The other four decide the margin.
Each step is only trustworthy because the one before it was written down.
Estimate
Structured scope in a keyboard-first grid: modules, tasks, roles and hours, with your rate card behind every figure.
Freeze
The won version becomes a commitment, with hours, cost, margin and rates copied into it. A rate rise tomorrow never moves a deal closed last spring.
Measure
Worklogs arrive from Jira or ClickUp each night. Every hour keeps the role and rate it was worked at, and lands on the line it belongs to.
Detect
Both forecasts, the deviation trend, data readiness and out-of-plan work, against thresholds you set per workspace or per project.
5
Decide
An overrun has three exits. The register remembers which one you took, so a cost somebody accepted can be told from one nobody noticed.
Annex
Scope taken from the delivery tool, with hours measured rather than guessed.
Cut scope
Drop part of the work to stay inside the commitment you made.
Absorb it
Take the cost knowingly, and on the record.
What the screens actually say.
Every figure names where it came from and how complete the data behind it is. Nothing here is inferred to fill a gap.
Projects · Nightsky CommerceHigh risk
Margin sold41.0%baseline v2 · 11 Mar
Margin declared39.4%forecast A
Margin proven23.2%forecast B
Hours done against committed1 204 / 1 840 h
Closed share of plan46%
Data readiness84% · 3 unmapped authors
Out-of-plan work detected96 h on 4 tasks
Gap between forecasts, in money18.2 pp · 61 400 PLN
Baseline in forcev2 · frozen 11 Mar
Last sync from Jiratoday, 03:12
Risk thresholds · workspace defaults
| Pricing model | At risk | High risk |
|---|---|---|
| Fixed price | −3 pp | −6 pp |
| Time & materials | +10% | +20% |
Overridable per project, and every change is an audit event.
Decision register
Annex won and folded in+184 h
Baseline v3. Measured hours reattached to the plan.
Cost absorbed−22 400
Our own regression. Recorded, not hidden.
Alert dismissednoise
Dismissal is not an exit — it tunes the rules.
Demonstration data · not a customer's project
A commitment has to come from somewhere.
Everything the estimation workspace did before is still here. It is simply no longer the end of the story.
Estimation grid
Keyboard-first, modules and tasks, roles and hours, fixed-price or T&M per line. Totals recalculate as you type.
Versions and history
Every change to scope, effort and pricing is a tracked state you can read back and compare.
Estimation library
Reusable modules and tasks, so the third e-commerce project starts from the first two.
Client-ready offers
A branded offer and PDF generated from a version, with a delivery timeline derived from the plan.
Clients and pipeline
Estimates grouped by client, with statuses from draft through review to won or lost.
Review and approval
Who reviewed the estimate and when, without introducing approval bureaucracy.
Jira, ClickUp, Teamwork
Export the scope as structured tasks, or link work that already exists. Measurement works with Jira and ClickUp alike.
Roles, audit, workspaces
Permissions per role, an audit trail on every state change, and the whole product in Polish, English and German.
Four things this product will not do, on purpose.
A tool that watches money has to be believable on the days it says something you do not want to hear.
No single health score
Two forecasts stay two forecasts. A blended percentage would be unfalsifiable, and it would bury the disagreement that carries the information.
No alert without a change of state
Mail goes out when severity moves, at most once per project per day, and a dismissed finding never re-fires until it genuinely worsens. Daily noise is the one failure this product would not survive.
No rewriting of history
A commitment is a copy, not a reference. Change happens by creating the next version, so the original stays readable after two annexes and a rate rise.
No invented numbers
Every figure names its source and its readiness. When nothing can be computed, the screen says so instead of showing a confident zero.
Questions a delivery lead asks first.
That is precisely the case forecast B exists for. It ignores remaining estimates entirely and derives the total from the share of the plan already closed, so a team that only logs hours still gets a measured answer — and the gap between the two forecasts tells you how much the declared one is worth.
Start with the one project you already worry about.
Freeze its commitment, connect Jira or ClickUp, and let one night of worklogs answer the question you have been answering from memory.