AdapData

Established 2015 Notes

Notes · I Strategy advisory

A data strategy that starts from decisions

In short

A data strategy states which decisions an organisation intends to make better, and what evidence those decisions require. Written that way it usually calls for less infrastructure than expected. For private-market firms and the companies they own, the most valuable material is already held: timesheets, job records, quotations, service tickets and the operational logs of the business. A platform built before the decisions are named tends to serve reporting rather than judgement.

Note

What is a data strategy, and what is it not?

A data strategy is a short document that names the decisions worth improving, the evidence each requires, the people accountable for them, and the smallest arrangement of systems that would supply that evidence reliably. It is a strategy in the ordinary sense: a choice about where effort is placed and, more importantly, where it is not.

It is not an inventory of systems, a warehouse design, or a maturity assessment. Those are outputs of a strategy rather than substitutes for one. A document that lists capabilities to be acquired without naming a decision to be improved has deferred the difficult part.

Which decisions are worth the effort?

Three tests are sufficient in most cases. The decision recurs, so that an improvement compounds. Somebody is accountable for it by name. And the person making it would act differently if the evidence changed.

The third test disposes of a surprising amount of proposed work. Many reporting requests concern decisions that are already made, or that are made on grounds the evidence would not alter. Those requests may still be worth meeting for governance reasons, but they should not be presented as improvements to judgement, and they should not justify a platform.

Why do timesheets and operational records matter more than they appear?

Firms often begin by seeking external data when the more informative material is already inside the business. Timesheets record where expert attention actually goes. Job records show which work overran and which client accepted a variation. Quotations, once matched to outcomes, reveal which assumptions were consistently optimistic. Service tickets describe the same failure recurring under different descriptions.

This material is unglamorous, incomplete and full of local convention. It has one property that purchased data rarely has: it describes the specific business rather than the market average. In our experience the first honest analysis of internal operational records changes more decisions than the first external dataset.

What governance is actually required?

Enough to answer four questions and no more. Who owns each material dataset. Who may see it. How long it is kept. And who is told when something goes wrong.

Governance regimes fail in two directions. Too little, and personal or commercially sensitive material accumulates in places nobody has assessed. Too much, and a committee is created that reviews requests slowly enough that the business routes around it, which produces worse practice than having no regime at all. The proportionate version is written on a few pages, is owned by a named executive, and is reviewed once a year.

Where should the data live?

Residency is now a commercial question as well as a legal one. Where the data sits determines which law applies, which supervisory authority has jurisdiction, and which suppliers may lawfully be used. For firms operating across the United Kingdom, the European Union and the United States, the answer is rarely uniform across all datasets.

The practical approach is to classify by sensitivity rather than by system, then place each class deliberately. Ordinary operational records can usually sit wherever the business is most efficient. Personal data, deal material and anything covered by client confidentiality obligations should have a residency decision recorded against it, and that decision should be revisited whenever a supplier changes its subprocessors.

How does this differ at firm level and at company level?

At firm level the recurring decisions concern origination, diligence, monitoring and reporting to investors. The evidence needed is mostly documentary and mostly internal, and the constraint is confidentiality rather than volume.

At company level the decisions are operational: pricing, scheduling, procurement, retention. The evidence sits in transactional systems, and the constraint is quality rather than confidentiality. A firm that applies its own template to a portfolio company usually finds it does not fit, and the attempt to make it fit consumes a year.

Questions

What is a data strategy?
A short document naming the decisions an organisation intends to improve, the evidence each requires, who is accountable, and the least infrastructure that would supply that evidence reliably. It is a set of choices, not an inventory of systems.
What does a data strategy consultant do?
Establishes which decisions an organisation intends to improve, what evidence each requires, and the least infrastructure that would supply it reliably. The useful output is a short document naming decisions and owners, rather than a technology selection.
Should a data platform come first?
Rarely. A platform built before the decisions are named tends to serve reporting rather than judgement, and its cost is committed before its purpose is known. Naming two or three decisions first usually reduces the infrastructure required.
What data matters most in a private-market context?
The operational records a business already holds: timesheets, job records, quotations matched to outcomes, and service tickets. They describe the specific business rather than a market average, and they are ordinarily available without procurement.
How much governance is proportionate?
Enough to state who owns each dataset, who may see it, how long it is retained and who is notified when something goes wrong. Regimes that grow beyond this are routed around by the business, which is worse than having none.
Does data residency still matter?
Yes, and increasingly for commercial reasons as well as legal ones. Residency determines applicable law, supervisory jurisdiction and which suppliers may be used. It should be recorded per class of data and revisited when a supplier changes its subprocessors.
What are the five pillars of data strategy?
No canonical list exists, though most published versions cover purpose, governance, architecture, quality and people. As a checklist the headings are reasonable. As a plan they are not, since none of them names a decision the organisation intends to make better.

Engagement

We are usually engaged by a firm or an owner who has been offered a platform and would prefer to begin with the decisions. Work is senior-led and conducted in confidence. We reply personally to every enquiry; write to [email protected].

Related