Techbanq

Engagement models · governance

Governance is one honest hour a month, not a thicker report

The day-to-day management of an offshore team is one job; governing the relationship is another, and conflating them produces executives reading sprint reports. Governance asks three questions monthly — is working software arriving, what could break this, are the people stable — and holds two things in writing: who owns architecture, and what would trigger intervention.

Written for
CTO, CIO, COO
Decision
Govern an offshore engineering relationship at board and executive level
Languages
EN · AR

Information security for software delivery

  • ISO

    ISO/IEC 27001

    ISO/IEC 27001 certified

Certifications & registrations

The instrument

Three feeds into one monthly decision about the relationship

Everything the review needs arrives from three directions and leaves as one of three postures. The hour is short because the evidence is prepared; it is honest because the evidence is checkable.

A surface diagram: three evidence feeds — outcomes, risks and people — converge into a monthly review, which issues one of three postures: continue, probe, or intervene against pre-written triggers.

  • Outcomes

    What shipped to users this month, shown working — against what was said last month.

  • Risks

    A live register with owners and dates — technical debt, dependencies, the single points of failure.

  • People

    Who joined, who left, who is overloaded, where knowledge is concentrating — the slowest signal and the most predictive.

What we do

The monthly review

One hour, senior people from both sides, evidence read in advance. Its output is a posture, recorded with reasons.

  • Continue

    Evidence healthy; the leadership's job this month is to stay out of the way.

  • Probe

    One signal is off; a named person investigates and reports before the next review.

  • Intervene

    A written trigger was crossed; the pre-agreed response begins, with no improvised drama.

The postures are the point: a review that can only ever conclude "continue" is a ceremony, and one that intervenes without written triggers is a mood.

Measurement

Choose metrics by what it costs to fake them

Every metric becomes a target the moment a relationship depends on it, so the governance question is not which metrics are accurate but which are expensive to fake. Activity metrics fail this test cheaply: hours are gamed by logging, ticket counts by splitting, velocity by inflating estimates — each fakeable at a desk, without deceiving anyone in particular. Working software fails it only at the price of building working software: a demonstrated feature in front of the people who asked for it, defects found by users rather than by the team, and time from request to production are all hard to counterfeit because counterfeiting them requires most of the real work.

The second discipline is provenance: read the numbers from systems of record you can open yourself — the repository, the pipeline, the incident tracker — rather than from a slide the vendor prepared about its own performance. A vendor curating its own scorecard is not necessarily dishonest; it is simply doing what every measured party does, which is why governance never outsources the measuring. Three or four hard-to-fake numbers, read from the source, reviewed against last month's, beat any dashboard of twenty soft ones.

Authority

Who decides what, settled before the first dispute

Architecture disputes are decided by whoever holds authority at that altitude — so the altitudes are written down while everyone still agrees.

01

Direction stays with the buyer

Target architecture, technology standards, build-versus-buy and anything that outlives the engagement are yours — a vendor may argue, in writing, and then must build to the decision.

02

Implementation belongs to the team

Within the standards, the how — patterns, libraries, internal design — belongs to the engineers doing the work; governing at this altitude is how executives become the bottleneck they fear.

03

The boundary cases get a written decision

A new database, a new external dependency, anything crossing a security boundary — pre-listed as decisions that must surface, each resolved through a short written proposal rather than discovered in a code review.

04

Authority is a register, not a memory

The decisions and who made them live in a log both sides read — so a year later, the question is answered by a record rather than by whoever remembers loudest.

The annual audit

Once a year, examine the relationship as if it were new

Monthly reviews catch drift in delivery; they do not catch drift in the relationship's foundations. Once a year, ask the questions you asked before signing — of the relationship you now have.

STAGE 01

Recertify access and the contract chain

Every account against every current person, every grant against current need, and the contractual protections against how the engagement has actually evolved. No access or clause surviving on inertia.

STAGE 02

Map knowledge concentration

For each critical system, name who could work on it tomorrow. Single names are findings; the year's documentation and rotation debt gets paid here. A bus-factor map with owners for the red entries.

STAGE 03

Test exit readiness

Could you leave in ninety days without losing the operation — repositories, environments, documentation, the knowledge to run what exists? Answer by inspection, not by assumption. An exit that stays hypothetical because it stays possible.

STAGE 04

Re-decide the relationship

With the evidence assembled, decide deliberately: renew as is, reshape the model, or begin an exit. The audit's value is that the decision is made rather than inherited. Another year chosen on evidence, or an exit begun early.

Our role

We equip the governance that could fire us

We are an offshore vendor describing how to govern offshore vendors, and the conflict resolves the same way as elsewhere on this page: in writing. Everything above is a commitment we accept about ourselves, checkable from your side of the table.

  1. 01

    The pack, from your systems

    The monthly evidence — demonstrations, metrics, the risk register, people movements — is assembled from systems you can open, so the review never depends on taking our word.

  2. 02

    Risks reported against ourselves

    Knowledge concentrating in one of our engineers, debt we created, a delay we can see coming — raised by us in the register before you find them, because a governance body surprised twice stops believing.

  3. 03

    The audit, welcomed annually

    Access recertification, the knowledge map and the exit-readiness test run with our cooperation on a named date — a vendor that resists the audit has answered it.

The triggers, the postures and the annual decision are yours to hold; our part is evidence honest enough that trusting us stays a conclusion you re-reach monthly, never a habit.

Common questions

What buyers ask

What belongs in the monthly governance review of an offshore team?+

Three subjects and nothing else: outcomes, risks and people. Outcomes means working software demonstrated against what was promised a month earlier — not activity, not percent-complete. Risks means a live register with owners and dates, including the ones the vendor raised against itself. People means joins, leaves, overload and knowledge concentration, because team instability predicts delivery failure months ahead of any delivery metric. What does not belong: sprint mechanics, task lists and anything the working level already manages daily — governance that descends to that altitude duplicates management and misses its own job, which is deciding monthly whether to continue, probe or intervene.

Which delivery metrics can a vendor not easily game?+

The ones whose cheapest counterfeit is the real thing. A feature demonstrated working in front of the people who requested it can only be faked by building the feature. Defects that reached users, read from your incident tracker, cannot be edited by the vendor at all. Time from request to production, read from your pipeline, resists inflation because both endpoints are recorded by systems. Contrast the cheap ones: hours logged, tickets closed, story points — each adjustable at a desk without touching the product. Two rules follow. Prefer metrics measured by your systems over metrics reported by the vendor, and prefer few over many: three hard numbers reviewed with attention beat twenty that nobody interrogates.

When should leadership intervene in an offshore engineering relationship?+

When a trigger written in calm times is crossed — and, with rare exceptions, only then. Workable triggers are observable: two consecutive reviews without demonstrable working software, a risk with a named owner unaddressed past its date, senior departures the vendor cannot explain, or evidence arriving curated where it used to arrive raw. Writing them down in advance does two jobs: it stops the anxious executive intervening on mood, which teaches a team to manage impressions instead of software, and it stops the optimistic one waiting past the point where intervention is cheap. Between triggers, the discipline is symmetrical: trust is not a feeling either — it is the monthly conclusion from evidence, re-earned and re-granted each review.

Who should hold architecture authority when the engineering is offshore?+

Split it by altitude and write the split down. Direction — target architecture, technology standards, build-versus-buy, anything that outlives the engagement — belongs to the buyer, because you will live with these choices after any vendor is gone; a good vendor argues its case in writing and then builds to the decision. Implementation detail within those standards belongs to the engineers doing the work, wherever they sit — pulling it onshore out of unease creates a bottleneck and tells the team it is not trusted with its own craft. The load-bearing part is the boundary list: name in advance the decisions that must surface — a new datastore, a new external dependency, anything crossing a security boundary — so neither side discovers, mid-dispute, that it assumed authority the other believed it held.

Shape the review with us

A governance design session producing your monthly review agenda, the metrics it reads, and the written intervention triggers.

Start the conversation