How to Build a Defensible Comp Set
A comp set is an argument, not a search result. Pulling 30 SaaS multiples and taking the median is what gets your valuation discounted. The defence is the criteria — and partners read both the comparables you included and the ones you left out.
The short answer
A defensible comp set is a deliberately narrow group of 8 to 12 transactions or trading multiples chosen on disclosed criteria — sector, growth profile, revenue scale, geography, capital structure — with each comparable individually justified. Indefensible comp sets do one of two things. Either they pull a wide median ("SaaS at 8x ARR") with no business-specific reasoning, or they cherry-pick favourable outliers without defending the exclusion of the others. Partners reject both. The defence is the criteria, not the multiple.
Key Takeaway: Investors price the case you constructed, not the database screen you ran. The comp set is the case. A median taken from 30 unfiltered comparables tells the partner you didn't think about which businesses are actually like yours — and the discount follows.
Why most founders get this wrong
The instinct is to maximise the number of comparables on the assumption that more data produces a tighter range. The opposite is true at scaleup stage. A comp set of 30 SaaS companies dilutes the signal: the median absorbs verticals you don't compete in, growth profiles you don't share, and capital structures that change the multiple by more than the operating model does.
The three failure modes that follow:
The screen-and-paste. The founder pulls every SaaS company in the database between £5M and £50M ARR and takes the median EV/Revenue multiple. The partner reads this and discounts because the screen included sub-scale companies, decelerating companies, and businesses with materially different gross margins. The median means nothing — and the founder cannot defend any specific comparable without admitting they didn't think about it.
The cherry-pick. The founder selects the four highest-multiple comparables and presents them as the comp set. Partners notice immediately. The exclusion of the other reasonable comparables is the tell — a defensible comp set should be the population you would have included before you knew the multiples, not after.
The over-narrow set. The founder cuts the comp set to three companies on the grounds of perfect comparability. This produces a different problem: a sample of three doesn't carry statistical weight, and partners discount narrow sets on the assumption the founder filtered out unfavourable comparables. Eight to twelve is the band that holds.
Example: A vertical SaaS founder building the comp set for their Series B looked at 28 SaaS companies with similar revenue. Tightening the screen to vertical SaaS (5 of the 28), with NRR above 110% (3 of the 5), gross margin above 80% (2 of the 3), and primary growth from net new logos rather than expansion (both qualified) produced a 2-comparable set. Adding the next-best 6 with explicit "imperfect comparability — included for these reasons" annotations produced an 8-comparable set the partner could underwrite.
What "good" looks like
A defensible comp set passes through three screens, each progressively narrowing the population. The screens are disclosed; the rationale per comparable is explicit; the exclusions are defended.
1. Sector screen — the obvious filter, applied carefully
Start narrow. Vertical SaaS for vertical SaaS, B2B marketplace for B2B marketplace, infrastructure software for infrastructure software. Cross-sector medians ("all SaaS") are useful as a sanity check but never as the primary comp set. The sector screen also captures business-model nuance: subscription versus usage-based versus transactional, which materially changes the multiple.
2. Stage and size screen — match the round, not the headline revenue
Filter by ARR band (typically half to twice your current ARR), by growth rate band (within 15 percentage points of your forward growth), and by gross margin band (within 5 to 10 points of your gross margin). At Series A the band can be wider; at Series B the bands tighten because investors price the trajectory more precisely.
3. Transaction type screen — match the financing event you're running
Trading multiples (public companies) carry implicit liquidity premiums. Precedent transactions (private financings) carry implicit synergy or strategic premiums depending on buyer type. A Series B raise should be primarily benchmarked against precedent Series B transactions, not against listed SaaS at scale. See using precedent transactions at Series A and B for the data sources and adjustments.
4. Defend each inclusion in one sentence
For every comparable in the final set, write one sentence explaining why it belongs. "Included: similar vertical, similar ARR scale, NRR within band." For excluded comparables that the partner might raise, write the rejection rationale: "Excluded: gross margin 12 points lower, primary growth from acquisitions not organic."
5. Triangulate with intangibles, not with averages
The comp set produces a multiple range. Where you sit inside that range is governed by the intangible asset base — customer-capital depth, switching costs, brand recognition, data assets, technology defensibility. See why 70% of your valuation is intangible. The comp set sets the boundaries; the intangibles set the position inside them.
How to apply it to your round
The work happens before the partner conversation, not during it. A founder who arrives at the second meeting with a pre-built, annotated comp set has reframed the conversation: the partner now reacts to the founder's analysis rather than the founder reacting to the partner's database.
Practically, the sequence is:
Build the set early. Construct the comp set during the first two weeks of round preparation, not in response to a term sheet. The work is meaningful — three to five days for a scaleup raise — and it produces a working document the team can debate internally before any partner sees it.
Document the screens explicitly. The first slide of the comp-set section should disclose the three screens and the population reduction at each stage. "Started with 87 SaaS companies in the database; sector filter to vertical SaaS, 14 remained; stage and size filter, 9 remained; transaction-type filter, 8 remained, plus 4 included as imperfect comparables with annotation." Partners trust disclosed methodology more than they trust a single multiple.
Let the partner challenge specific comparables. Partners almost always raise one or two comparables they expected to see and didn't, or one or two that look out of place. The annotated rationale answers most of these in advance. The remainder can be discussed substantively rather than defensively.
Refresh quarterly. Comp sets decay. Public-company multiples move; private-financing data accumulates. A comp set built six months before the round may need a refresh before the partner conversation — and the refresh itself is a useful signal of operator discipline.
Comp set that fails
- 30+ companies with no disclosed screens
- Median multiple presented without rationale
- No annotation per comparable
- No defence of exclusions
- Trading multiples used to anchor a private financing
Comp set that works
- 8-12 comparables, three disclosed screens
- Population reduction shown at each stage
- One-sentence inclusion rationale per comparable
- Explicit exclusion rationale for the obvious omissions
- Precedent transactions matched to your round type
Trading multiples vs precedent transactions in the comp set
The comp set is rarely homogeneous. Most defensible Series A and B comp sets contain a mix of trading multiples (public companies, where you have continuous price discovery but with a liquidity premium) and precedent transactions (private financings, where the price is from an actual closed deal but the data is older). The discipline is to disclose the mix and weight it deliberately. A 50:50 split is common at growth stage; a higher proportion of precedent transactions is appropriate at earlier stages where public-company comparables don't exist. See using precedent transactions at Series A and B for the discipline that complements trading-comp analysis.
What partners do with the comp set internally
The partner takes your comp set into their committee. The committee will do one of two things: accept your set as the working comparable group and price the round inside the range it produces, or reject it and substitute the comp set the firm's pricing model would have produced. The substitution is what kills the round economics for founders — the firm's default comp set is broader, less defensible, and consistently produces a lower median than a well-constructed founder-led set. The work the founder does on comp-set construction directly determines whether the partner takes your set into committee or substitutes their own. That single committee-room dynamic is where most of the valuation gap actually opens up.
The Bottom Line
A comp set is a written argument the founder constructs before the partner conversation. The defence is the criteria, not the multiple. Eight to twelve comparables, three disclosed screens, one-sentence rationale per inclusion. The work is small relative to the valuation impact — and partners price the discipline of a well-constructed set materially higher than a screened median.
Related reading
Comp sets sit inside a wider methodology stack. For the full menu of valuation methods that partners triangulate against the comp set, see valuation methods for scaleups: DCF, comps, RFR & MPEEM. For the precedent-transaction discipline that complements trading multiples, see using precedent transactions at Series A and B. For the asset base that determines where you sit inside the multiple range, see why 70% of your valuation is intangible. For the framing that re-segments the comp set entirely, see narrative arbitrage: same business, different number. For the partner-facing benchmark by sector, see Series A valuation benchmarks by sector. For the underlying value-driver framework, see The Opagio 12.
Build the comp set before the round opens
Eight minutes. Twelve drivers. The starting frame for the asset base that determines where you sit inside the comp range.