How to do it
Start from the channel, not from SAM
SOM is a distribution question wearing a market-sizing costume. Ask how many of the right buyers one working channel puts in front of you per month — a search term with known volume, a community with a known size, a partner with a known client list, outbound with a known reply rate.
Apply a conversion rate you can source
Use a number from your own funnel if you have one, or a published benchmark for the same motion if you do not, and say which. Self-serve SaaS converting visitors to paid is low single digits; warm outbound to a qualified list is far higher. Using a blended "industry average" hides which of the two you are.
Cap it at what you can actually service
If onboarding takes four hours and there is one of you, your annual ceiling is a few hundred customers regardless of demand. A SOM above your delivery capacity is a forecast of a support failure.
State the period out loud
"SOM is ₹2 crore" means nothing. "₹2 crore of ARR by month 30, from one channel, at a 3% conversion rate" is a claim someone can argue with — which is what makes it worth writing.
Worked example
Conversion rates by motion — use the row that matches yours, not a blended average
A blended industry average sits somewhere in the middle of this table and describes none of these motions. Using it is how a plan built on cold ads inherits the conversion rate of a referral channel.
SOM is a distribution question wearing a market-sizing costume
Every credible SOM reduces to one sentence: this channel puts this many of the right people in front of us each month, this share of them buy, and we can service that many. If you cannot say that sentence, you have a share of SAM rather than a forecast.
The consequence is that improving SOM almost never means changing the market. It means making one channel work, then a second. Founders who want a bigger SOM usually need a distribution experiment, not a bigger spreadsheet.
The capacity ceiling nobody models
Onboarding time is the constraint that turns a good SOM into a support failure. If each customer needs four hours of setup and the team has one person doing it, the annual ceiling is a few hundred regardless of how many want to buy.
Model it explicitly: new customers per month × onboarding hours, against the hours actually available. When the plan exceeds capacity, the honest options are to shorten onboarding, hire ahead of the curve, or lower the forecast — and the first is usually cheaper than both the others.
What goes wrong
- Taking a percentage of TAM — the "we only need 1% of a huge market" claim is the clearest signal an idea has not been examined
- Assuming a channel exists before you have made one work once
- Using a conversion rate from a different motion than the one you will run
- Ignoring service capacity, so the plan needs more onboarding hours than the team has
- Leaving the time period unstated, which makes the number unfalsifiable
Is your number plausible?
Each check catches a different way the arithmetic can be right and the answer still wrong.
- Divide SOM by the period in months. If the implied new customers per month is a number you have never hit, the plan is a wish
- Multiply new customers per month by your onboarding hours. If the total exceeds the hours the team has, SOM is above capacity
- Check SOM against SAM. Above roughly 10% in three years from one channel is unusual and needs a reason you can state
Related answers
What period should SOM cover?
Two to three years is the usual window, and stating it is what makes the number meaningful. A SOM with no period attached cannot be right or wrong, which is why so many decks contain one.
Should churn be in the SOM calculation?
Yes, if you are expressing SOM as customers active at the end of the period rather than customers ever acquired. Say which you mean — the two can differ by half at realistic churn rates.
Is 1% of the market a reasonable SOM?
It might be, as an output. It is never reasonable as an input. If 1% is where your channel maths lands, that is a finding; if it is where you started, you have assumed the answer and learned nothing.