Trigpoint
“The finer you slice it, the more decoupled it is”: they are two different things.

Microservices

Microservices · Lewis & Fowler (2014)
Checked against the sources ✓ Software architecture · service boundaries Read first: Domain-driven design (DDD) Read first: Conway's law Tell the editor privately ▸
In three sentences Split one application into a set of services that are released independently, and the unit of deployment and scaling gets smaller. What you buy is autonomy and local scaling. What you pay is the full cost of a distributed system. And the cost most easily underestimated is that consistency across services has stopped being free.

↳ Seen a counter-example?

Holds when
When load or changes concentrate in a few parts, the boundaries are clear, and deployment and monitoring are in place.
Evidence
Conceptual framework; Checked against the sources
Breaks down when
It fails if you split before deployment and monitoring are cheap, cut finer than business boundaries, or expect cross-service transactions.
Microservices Lewis & Fowler (2014) What splits is the deployment unit, not the coupling Monolith Feature one Feature two Feature three change one place, release the lot scaling means duplicating all of it Split into services Service one own store Service two own store Service three own store released and scaled alone Gain: one team owns one service organised around business capability Pay: distributed operations network, observability, debugging, deploys Most underestimated: consistency strong cross-service guarantees are gone The order: find a stable boundary with a monolith first, then split over-splitting is easy to spot: more time goes on the joins than on the content Figure: trigpoint.pub
Figure 1 · On the left the monolith: change one place and the whole package has to be released again, and scaling means duplicating all of it. On the right, after splitting: the unit of deployment and scaling becomes the service, at the cost of three things arriving at once, of which the most underestimated is the loss of consistency across services. Along the bottom is the order of splitting: use a monolith to find a stable boundary first, then split. The comparison is conceptual: no service counts, latencies or measured numbers. View full size ▸

1. What happened

The problem with a monolith is not that it is large. It is that it is coupled at the level of deployment: any small change anywhere means releasing the whole package again, and scaling means duplicating all of it even when the load falls on one feature.

This 2014 article by Lewis and Fowler pinned down the alternative: build a single application as a set of small services, each running in its own process and communicating through lightweight mechanisms[1]Lewis, J. & Fowler, M. (2014-03-25). Microservices: a definition of this new architectural term. martinfowler.com/articles/microservices.html. Bibliographic details checked against the publisher's record. Claims matched against the full text. The article's definition: building a single application as a suite of small services, each running in its own process and communicating with lightweight mechanisms, often an HTTP resource API. The same article lists nine common characteristics and notes that not every microservice architecture has all of them.. The article also lists nine common characteristics and notes that not every microservice architecture has all nine[1]Lewis, J. & Fowler, M. (2014-03-25). Microservices: a definition of this new architectural term. martinfowler.com/articles/microservices.html. Bibliographic details checked against the publisher's record. Claims matched against the full text. The article's definition: building a single application as a suite of small services, each running in its own process and communicating with lightweight mechanisms, often an HTTP resource API. The same article lists nine common characteristics and notes that not every microservice architecture has all of them.. This page dwells on three of them: organised around business capabilities, componentised as independently deployable services, and decentralised in both data and governance.

Organised around business capabilities buckles straight back to Conway's Law: service boundaries mirror team communication boundaries, so the ideal shape is one cross-functional small team owning one service entirely. That is not a coincidence; the article cites that law explicitly.

Decentralised data is the item underestimated most. Each service manages its own storage, so consistency across services stops being free: the guarantee you used to get from a single database transaction disappears, and what replaces it is eventual consistency, compensating flows and distributed debugging.

One last thing. The same author gave a piece of counsel elsewhere that gets needed more often than the whole list of characteristics: use a monolith to find a stable boundary first, and split afterwards[5]Fowler, M. (2015-06-03). MonolithFirst. martinfowler.com/bliki/MonolithFirst.html. Bibliographic details checked against the publisher's record. Claims matched against the full text. A short piece by the same author in 2015 argues that a new project should not start with microservices, and should use a monolith to find stable boundaries between services first; microservices work well only when those boundaries are stable, and pay off only for more complex systems..

2. Why this matters to you

What this page is useful for is not services. It is the timing of a split.

You will badly want to break your workflow into small pieces: one that writes outlines, one that looks things up, one that polishes, one that checks citations. Each piece is simple on its own and easy to test on its own, which sounds entirely right. What this page says is that the return on splitting depends on a set of preconditions, and without them a split is pure overhead.

What are the preconditions? Another short piece by the same author lists three: rapid provisioning, basic monitoring and rapid deployment[6]Fowler, M. (2014-08-28). MicroservicePrerequisites. martinfowler.com/bliki/MicroservicePrerequisites.html. Bibliographic details checked against the publisher's record. Claims matched against the full text. A short piece by the same author in August 2014 lists three baseline capabilities to have before adopting: rapid provisioning, basic monitoring and rapid application deployment; without these baseline competencies, the microservice style should not be considered.. Translated to you: do you have a cheap way to connect those pieces up, to see what state they are in, and to find out which one went wrong when something breaks in the middle. If running the whole flow means moving output by hand three times, and finding a fault means walking back through each piece, the coordination cost you pay already exceeds what the split saved.

The second thing is the finer is better misreading. Small is a contrast with the monolith, not a target. The symptom of over-splitting is easy to recognise: you spend more time on the joins than on the content.

The third, and the most practical, is the counsel to use a monolith to find a stable boundary first, translated directly: do not split before the boundary is clear. Get the whole thing working in one place first, until you can say which stretches are stable and which are different every time, then split along the stable seams. Do it the other way round, splitting before you have felt the boundary, and you will build interfaces in the wrong places, and those interfaces are hard to remove afterwards. That is exactly what the criterion in Modularity & Thin Crossing Points is describing.

Verification layer · Members
The guided layer of this node is open to all readers.
Verification details and failure boundaries are for members.
US$8.25 a month on average · or subscribe monthly ▸

Lab cases Written only where there is evidence

No field record for this node yet.

Common misreadings

Page verification: Checked against the sources ✓ · Content last revised 2026-11-14

  • A common retelling, microservices were proposed by those two authors: the article itself says the term was discussed at a workshop in May 2011 and settled on by that group in May 2012. This site marks the two authors as the ones who crystallised and popularised the definition. Secondary accounts report a related coinage as early as 2005, from a different technical line. #
  • A common retelling, small is the point and smaller is better: small is a contrast with the monolith, not an order to cut as fine as possible. The article notes that remote interfaces have to be coarser than in-process calls; the same author's other piece argues for starting with a monolith and splitting once the boundaries are stable. #
  • A common retelling, this is the authoritative definition of the term: the article itself says there is no precise definition, describes common characteristics only, and offers its definition as a leading candidate. #
References

How to read the source notes “Bibliographic details checked” means author, year, title and venue were confirmed. It does not mean the source's claims were checked. “Matched against the full text” means the claim was compared with the text of the source itself. Definitions as of 2026-09-20.

  1. [1] Lewis, J. & Fowler, M. (2014). Microservices: a definition of this new architectural term. martinfowler.com/articles/microservices.html. Published 25 March 2014. ↩
  2. [5] Fowler, M. (2015). MonolithFirst. martinfowler.com/bliki/MonolithFirst.html. Published 3 June 2015. ↩
  3. [6] Fowler, M. (2014). MicroservicePrerequisites. martinfowler.com/bliki/MicroservicePrerequisites.html. Published 28 August 2014. ↩

↳ Is a citation wrong?

Tell me when a new shelf opens

Tell me when a new shelf opens

Shelf notices are sent only when a new shelf opens or when the reading rules change.

Trial

Tell the editor privately: a correction, a counter-example or where it applies

Submissions are read only by the editor, and nothing is published before they are handled. Once a note is handled, a summary and response written by the editor are published. No sign-up is needed, and the nickname may be left blank.

What are you submitting?
Set a public name to be credited by name ▸
Submit anonymously, without signing in

Signing in lets you follow the status on your page. What you have written is kept.

You are signed in: this note is linked to your account, visible only to you and the editor, so you can follow its status on your page. The text is never published. Full privacy policy ▸

Submit anonymously instead, not linked to your account ▸

No email is collected. Your original text and nickname are read only by the editor and are never published; the site publishes only a summary the editor writes. The human check is provided by Cloudflare Turnstile. To limit abuse, the system keeps a hash of your IP address, not the address itself. Your original text is stored with our cloud provider and can be deleted on request. Full privacy policy ▸