Estimating work: T shirt sizing
I haven’t used the phrase “t-shirt sizing” for a while but it recently popped up in a dinner conversation with friends. It was one of those tech buzzwords that we glossed over, but thankfully one friend stopped us and asked “could you explain tshirt sizing?”, which led to a fun discussion of what it meant in our work environments (places at different sizes/growth stages).
Simplified definition
Tshirt sizing is used to estimate if a workload is “S — Small”, “M — Medium”, “L — Large”, or “XL — Extra Large”. It is said to be part of the agile project management movement, and attribution to the term has been given to Larry Osterman [1].
Implications of t shirt sizing
Sizing alone is a start, but what’s crucial are its implications. So for example, teams will need to agree on what each t shirt size looks like for workload — by weeks/months and engineers. So MAYBE something like the below bullets, where it would also depend on company/team/product/eng environments. Assuming X engineers, t shirt sizes could mean:
- S — 2–4 weeks of work
- M— 1–3 months
- L — 3–6 months (or longer??)
- XL — I don’t dare estimate this! 😆
Benefits of t shirt sizing
The primary benefit of t shirt sizing is in its simplicity for communications — communicating to leadership/cross functions about product plans, timelines, and resourcing needs. It’s imprecise nature could mean some fudging (e.g., no product team wants to be the one to propose a XL feature 🛑 — how would one even galvanize resources???). However, the relative sizing of S vs M can be incredibly helpful in deciding what initiatives to prioritize.
Slight segue — a common follow up question to sizing is whether more people allocated would mean the work can be completed faster (e.g., work is parallelized), or whether the work has dependencies and so more people does not help. Mythical Man Month has a fantastic essay discussing “Brooks Law”, which argues that when tasks cannot be partitioned, the “application of effort has no effect on scheduling (pg17), and in fact adding more people delays a project further, given coordination costs [2]. T shirt sizing cannot answer this question— it’ll need a more precise engineering manager estimate of type of work, eng weeks, eng skill etc.
Simplicity vs precision
Ultimately “t shirt sizing” is t shirt sizing for a reason — it’s an estimate, is not precise, but a good way to start the conversation on workload.
I recognize the irony of writing this given I have never been in product, eng, or technical program management! But sizing affects teams beyond those. Having been in cross functional roles interacting with product and eng to (a) write product strategy and plan resource asks (b) develop a partnerships strategy to support a product (c) find a partner to meet timelines that would fulfil that strategy (d) prioritize partnerships work according to product timelines, t shirt sizing and its implications on timelines helps effectively accomplish all of the above. I’m glad for this heuristic.
[1] More here and the definition as“making extremely rough estimates in terms of a small number of predefined categories”.
[2] Gemini had this helpful analogy — Imagine digging a ditch. You can definitely assign one person and it will take a certain amount of time.But if you add a second person, they both can’t dig in exactly the same spot without getting in each other’s way. They need to coordinate, decide who digs where, and potentially move dirt that the other person has already piled up. This coordination adds extra time to the project.