Managed ClickHouse Cloud FinOps: a dedicated desk that runs your cost discipline every day

Operating model

How our ClickHouse Cloud FinOps services operate

Exhibit 1 shows the model. Your estate provides three inputs, all read-only. The ChistaDATA FinOps desk runs four cadences. You receive four outputs, each with a named owner on your side.

Lifecycle

From first access to steady state

Each phase has an exit test, so progress is measured rather than reported. Mobilise and Baseline set the foundation once; Run is where ClickHouse Cloud spend optimization compounds month after month.

Daily control

ClickHouse Cloud budget alerts that arrive while there is still time to act

Most budget alerts arrive after the money is spent. Ours forecast to month end every morning and classify each service. BREACH means the forecast exceeds budget. WATCH means month-to-date has crossed the watch line. UNBUDGETED means a service spends with no budget at all. Exhibit 4 is the board the desk reviews, generated by the evaluator below from a synthetic month of cost data.

#!/usr/bin/env python3
"""
Managed ClickHouse Cloud FinOps: daily budget alerts per service, from the finops cost warehouse.

For each budgeted service: month-to-date spend, a straight-line forecast to month end, and a status.
  BREACH  forecast above budget, or month-to-date already at 100%
  WATCH   month-to-date above the watch line (default 80% of budget) but forecast within budget
  OK      neither
Services that spend but have no budget are reported as UNBUDGETED: they are the allocation gap.
Credentials and the webhook come from the environment; --dry-run prints instead of posting.
"""
import base64, calendar, json, os, sys, urllib.parse, urllib.request
from datetime import date

def query(sql: str) -> list[dict]:
    auth = base64.b64encode(f"{os.environ['CH_USER']}:{os.environ.get('CH_PASSWORD', '')}".encode()).decode()
    req = urllib.request.Request(f"{os.environ['CH_URL']}/?query={urllib.parse.quote(sql)}&default_format=JSONEachRow",
                                 headers={"Authorization": f"Basic {auth}"})
    return [json.loads(l) for l in urllib.request.urlopen(req, timeout=60).read().decode().splitlines() if l]

def evaluate(as_of: date, budgets: dict[str, float], watch: float = 0.8) -> list[dict]:
    first = as_of.replace(day=1)
    days_in_month = calendar.monthrange(as_of.year, as_of.month)[1]
    rows = query(f"""
        SELECT service_id, toFloat64(sum(chc)) AS mtd
        FROM finops.usage_cost_daily FINAL
        WHERE usage_date BETWEEN '{first}' AND '{as_of}' AND entity_type = 'service'
        GROUP BY service_id""")
    out = []
    for r in rows:
        mtd, budget = r["mtd"], budgets.get(r["service_id"])
        forecast = mtd / as_of.day * days_in_month               # straight line; replace with seasonality later
        if budget is None:
            status = "UNBUDGETED"
        elif forecast > budget or mtd >= budget:
            status = "BREACH"
        elif mtd >= watch * budget:
            status = "WATCH"
        else:
            status = "OK"
        out.append({"service": r["service_id"], "mtd_chc": round(mtd, 1), "forecast_chc": round(forecast, 1),
                    "budget_chc": budget, "pct_of_budget": round(100 * forecast / budget, 1) if budget else None,
                    "status": status})
    return sorted(out, key=lambda x: ["BREACH", "UNBUDGETED", "WATCH", "OK"].index(x["status"]))

if __name__ == "__main__":
    as_of = date.fromisoformat(os.environ.get("AS_OF", date.today().isoformat()))
    budgets = json.load(open(os.environ.get("FINOPS_BUDGETS", "budgets.json")))   # {"svc-prod": 2000, ...}
    result = evaluate(as_of, budgets)
    for r in result:
        print(f"{r['status']:10} {r['service']:12} mtd {r['mtd_chc']:>8} forecast {r['forecast_chc']:>8} "
              f"budget {r['budget_chc'] or '-':>6} ({r['pct_of_budget'] or '-'}%)")
    alerts = [r for r in result if r["status"] in ("BREACH", "UNBUDGETED")]
    if alerts and "--dry-run" not in sys.argv:
        req = urllib.request.Request(os.environ["FINOPS_WEBHOOK_URL"], method="POST",
                                     data=json.dumps({"text": f"ClickHouse Cloud budget alerts {as_of}", "alerts": alerts}).encode(),
                                     headers={"Content-Type": "application/json"})
        urllib.request.urlopen(req, timeout=30).read()

# Lab test: AS_OF=2026-10-15, budgets.json = {"svc-prod": 2000, "svc-bi": 1000}, synthetic cost data
# BREACH     svc-prod     mtd   1104.5 forecast   2282.5 budget   2000 (114.1%)
# UNBUDGETED svc-sandbox  mtd    120.6 forecast    249.2 budget      - (-%)
# OK         svc-bi       mtd    456.8 forecast    944.1 budget   1000 (94.4%)

On day 15 the production breach is still a forecast, with half a month to act on it. The unbudgeted sandbox matters more, because nobody was watching it. The desk’s first action on an UNBUDGETED line is to find its owner. Its first action on a BREACH is to trace the variance to services, meters and query fingerprints.

Inside the pack

What every monthly FinOps pack contains

The pack follows the same structure every month, so trends are visible and nothing is quietly dropped when the news is bad. A finance reader can stop after the first page; an engineering reader goes on to the backlog and the evidence behind it.

Section six is the one most reports leave out. A guardrail that blocks a legitimate month-end job is a cost as well, and so is an alert that arrives after the money is spent. Reporting our own misses next to the savings is what keeps ClickHouse Cloud cost management credible with the people who have to act on it.

Service levels

Service levels and the measures we report on ourselves

We also report on our own performance every month: the share of budget alerts triaged on the day they fire, allocation coverage against target, the share of backlog items with an owner and a date, and savings realised against savings proposed. If realised savings trail proposals for two consecutive months, the quarterly review starts there.

FAQ

Managed ClickHouse Cloud FinOps: questions we are asked

What do ClickHouse Cloud FinOps services from ChistaDATA include?

A daily cost pipeline and ClickHouse Cloud budget alerts, a weekly optimisation backlog, a monthly FinOps pack with review, and a quarterly commitment review, run by a named desk with read-only access.

Does the desk need access to our data?

No. It needs SELECT on system tables and on the finops database, and a billing API key that can read usage costs. It is refused on business tables and on all DDL.

Who approves changes in a FinOps managed service?

Your platform team, always. The desk proposes each change with blast radius, rollback and expected saving. Applying it is a separate, explicit step.

How is ClickHouse Cloud spend optimization measured?

Against the same metrics every month: allocation coverage, utilization, scan efficiency and commitment burn, plus savings realised against savings proposed.

Can we start with engineering and move to managed later?

Yes. Our ClickHouse Cloud FinOps engineering engagement builds the same platform the desk runs, so handover is a change of operator, not a rebuild.