Post by Hadley Harris with model by Kristin McDonald
- Make sure your portfolio construction aligns with power law shaped distribution returns
- Your initial strategy is just a starting point. It must be measured and adjusted over the life of the fund.
- Getting recycling right at seed is challenging. Use a holistic model that pulls in everything to optimize your chances. 🔥We’ve included a downloadable version of ours below🔥
I called this seed stage portfolio construction for dummies because when I look back at our strategy over the years I’m amazed at how many dumb mistakes we made. I’m not saying we’ve totally figured it out but we have learned a ton since we started investing in seed rounds back in 2010. Portfolio construction is one of those areas of venture that is difficult to fully understand until you’ve been through an entire fund cycle so hopefully this helps some people avoid making some of the same mistakes we did.
Choosing your initial strategy
The first thing you need to do is layout your initial portfolio construction. As I’ll explain in the next section this should not be the construction you end up with but it’s still very important to have a solid starting point to iterate off of.
Number of initial investments
The most common mistake I see here for new seed managers is having too few portfolio companies. For example I meet many preseed funds that are targeting 15–18 portfolio companies in their fund. This sounds reasonable but in practice is not a good idea. Even at a solid conversion rate of 60% between preseed→seed, seed→A and A→B, that fund ends up with a post series B portfolio of 3 companies. I hope it goes without saying that very few institutional LP’s would sign up for a B focused fund whose strategy was to only invest in 3 companies. In practice we think you want in the neighborhood of 20–25 post series A or 10–12 series B companies. Given venture returns generally follow a distribution similar to the power law this gives you a good number of “shots on goal” for exits that can be fund returners. From there you can work your way back to the stage where you plan to enter using an estimated conversion rate.
Pacing/initial investment period
Pacing is a hard thing to project but you’ll still want to layout over how many quarters you’ll make your initial investments. You need to think deeply about the quality of your deal flow as well as how many investment processes and new portfolio companies your team can manage. Our pacing is to do 35 new investments over 3.5 years. That initial investment period is quite a bit longer than average for a seed fund but we don’t feel comfortable doing more than about 10 investments per year across 4 GP given our level of involvement with our portfolio companies and as I explained above we feel strongly in having a relatively broad initial portfolio at the seed stage.
We’ll get into recycling below but one thing to note is the longer period you spread your initial investments, the more time you’ll have to get liquidity for recycling.
Average size of initial investments
The classic mistake here is new managers pick an amount like 250K or 750K when they should be pegging initial checks to a target ownership. At the end of the day, the percentage you own of a portfolio company when it exits is what’s important, not the amount you put in. Pick a target ownership percentage you’ll try to stay close to. Then based on the average terms you project investing at, approximate the average allocation you’ll need to invest to buy that ownership. As an example for our 4th fund, which is $100M, our target average ownership is 12.5% (all are 10%-15%) which for our investments equates to an average initial check of $1.25M. Part of the rationale for this ownership range is that if we maintain ownership in the 10% range when the company is worth $1B, an exit at that time can return the whole fund. Every successful VC fund I’ve heard of has at least one company return at least the entire fund (remember the power law) so you want to make sure you’re well aware of what that will take.
The size of your initial checks will have a strong impact on your strategy. At $1.25M we pretty much always have to lead or colead, which we’re built to do. In today’s seed environment you can probably do preseed checks up to around $300K and seed checks of up to $600K on average as a follower. Above that and you’ll likely need to start leading. Unless you feel confident you’re ready to lead and take the responsibilities involved with doing that (be the primary fund to drive diligence, negotiate terms, drive legal, be the hands-on investor of record, often take a board seat, be the first one founders look to if an extension is needed, etc), it’s probably best to set your ownership targets (and analogous initial check size) to a level where you’ll be able to get allocations as a follower.
Follow on strategy
To what extent do you want to follow on in order to maintain your ownership through subsequent rounds? You’ll get tons of conflicting advice here between people who say you need to optimize for “doubling down in your winners” and those who say you get the best value early. The optimal answer here depends on the size of your fund. The larger your seed fund, the more likely you’ll optimize results by reserving a sizable amount for follow on investments. A common mistake is to save money for follows at the cost of having too few portfolio companies, as discussed above. When you have a small seed fund it’s more important to have a broad portfolio than to protect your ownership post initial investment because in our experience, the ownership you buy at seed dwarfs the ownership you buy in later rounds for the portfolio companies that drive strong fund level returns. Like everything, there are exceptions to this but broadly we’ve seen this hold true. When you do follow on in subsequent rounds, the common practice is to “take your pro rata”, which means to maintain your ownership percentage. In the early rounds the allocation size to take your pro rata is often in the neighborhood of your initial check. Some funds have a strategy to do “super pro rata” in subsequent rounds but managers need to understand this strategy is dependent on who leads the following round and may be hard to execute consistently.
As seed funds grow, their optimal strategy is to maintain their ownership deeper into a portfolio company’s lifecycle as they need to put that extra money to work. To project this out you’ll need to estimate the conversion rates between each round. This can be difficult especially if you don’t have much historical data. For example, in Eniac III ($55M) we had a strategy of maintaining our ownership through the series B but our seed→A conversion rate ended up being about 50% higher than we projected, which resulted in that fund only being able to do half of its overall series B prorata. As we’ll discuss in the next section, had we had a more dynamic strategy at the time we could have made adjustments earlier to optimize based on the results we were seeing.
The end result of laying out your follow on strategy is that you should have projections for how your fund will be allocated over the rounds you expect to participate in. For example for Eniac IV (100M) we projected:
Portfolio construction must be dynamic
Over time we’ve come to appreciate that portfolio construction needs to be optimized throughout a fund’s lifecycle as projections are replaced by real data. If you’re planning on fully recycling (which almost every sophisticated LP wants to see since it effectively allows you to recycle your fees away) having a dynamic strategy is pivotal. As I mentioned before, one way to build your preseed/seed portfolio is to pick a later stage portfolio that gives you a good number of potential fund returners and work backward. To do this you’ll project a conversion rate between each round, but these are educated guesses so you’ll want to update these as you see real world results.
Take the example below where you’re building a seed portfolio with a target of having 12 active portfolio companies post series B. You initially project 75% conversion rate between each round but in practice, you see only 50%. A relatively small change in conversion rate results in a very different portfolio construction strategy.
Note that I’m using a consistent conversion rate for each round for simplicity — in practice you should project and track each conversion rate separately.
In this case, if you started seeing this lower conversion rate, you’d want to adjust in order to do more initial investments. In order to do this, you’d take money allocated for follow on rounds and reallocate those funds to doing more new deals. You need more “shots on goal” as having only 4 series B companies is not a well-diversified portfolio and thus very risky. In other words, you need more initial investments to achieve your target portfolio diversification.
One thing almost every seed manager struggles with is recycling. Recycling is especially challenging for seed investors compared to later-stage investors because the average hold time per investment is 2–3 years longer than series A.
The tension here is that you want to invest around 100% of your fund (i.e. fully recycle) but 20% (sometimes more, especially for very small seed funds) is generally earmarked for fees and costs over the life of the fund so you need to get that 20% back in the form of exits in order to fully recycle. Sometimes you get lucky and have a sizable exit early in a fund’s lifecycle like we did with Anchor but overall most exits in the first few years tend to be pretty small. Therefore most seed funds invest that 80% of their fund before they get back 20% worth of liquidity. If this happens you’ll have to either stop investing (which means possibly missing out on some exciting opportunities) or you can invest funds earmarked for future fees. Since most funds collect fees over a 10 year period you can invest funds earmarked for those later year fees betting on the fact you’ll have liquidity before those fees are needed. This is what most sophisticated seed funds do. Having visibility into this bet is where having a strong model is pivotal, which we’ll get into more in the next section.
One note, it’s really hard to hit 100% right on the nose. Generally, the investments that get you to full recycling are later follow on rounds which can be hard to predict well in advance. Our advice is to aim above full recycling, in the 105%-115% range, to ensure you get over 100%. A lot of great performing seed funds go well over 100%. When you raise your fund you’ll want to include provisions that allow for this although most sophisticated LP’s will be happy to amend your docs to help make this happen given it is in their best interest.
Driving with a single holistic model is key
For years we used a handful of models to maintain our portfolio construction. Generally at least one to lay out and track the strategy and another to project future cash needs over time. When Kristin joined us from Point72 last fall we finally had some serious modeling experience on the team. She and I worked together (99.9% Kristin, .1% Hadley) to put together this model which does a great job of tying everything together.
You’ll need Excel 2019 or Excel for Microsoft 365 for the model to work. If you don’t have those (and don’t feel like updating Excel), you can also upload the file onto Google Sheets and use it there.
I’ll let Kristin take it over from here to walk you through how to use our model.
The model is built to take your existing data and show you potential cash outcomes (i.e. how quickly you’ll move through your fund) depending on how you change certain preset variables. It is color-coded to make it easier to know what is an input (where you can change a variable) and what is an output (which should not be edited).
There are three tabs — the Summary and Assumptions tab, the Model tab, and the Portfolio tab.
- On the Summary and Assumptions tab, you can change the investment variables and view a summary of the fund’s current and predicted metrics.
- The Portfolio tab pulls in these variables to create a hypothetical portfolio, which can be updated with real data as you make investments.
- The Model tab then takes this hypothetical portfolio and creates a cash flow model, the results of which are summarized on the Summary and Assumptions tab.
Data and Variable Inputs
Most of the variables you will want to flex for your fund are located on the Summary and Assumptions tab. These include:
- Fund name, size, and start date (cells D9:D11)
- Expected number of investments (cell D102)
- Average size of expected investments (cell D101 for the initial check, cells K102:K105 for follow-ons)
- Predicted conversion rate between each round (cells K111:K114)
- Frequency of investments and follow-on rounds (cells K118:K123)
- Fund fees and expenses (cells C141:C151)
The assumptions around investment size, frequency, and conversion rates get pulled into the Portfolio tab to create a hypothetical portfolio (with each round size weighted by conversion rates). As you complete investments, you can manually update the information for each deal (replacing the hypothetical investment with a real one). The model will continue to forecast future investments and follow-on rounds based on your initial assumptions.
Below is an example of a hypothetical portfolio based on the initial assumptions, and an example of “real” data entered for several investments.
You’ll see that while we did build in the ability to make assumptions around future returns (more on that below), we chose not to model in expected returns here (only returns that have already been realized).
Cash Flow Model
The Model tab then pulls data from the Portfolio tab into each fiscal quarter, to create a cash flow model based on your investment assumptions. It then subtracts out any management fees and expenses to forecast your available capital and investable capital over the life of the fund. This is then graphed and summarized on the Summary and Assumptions tab under the Summary section. So, as you change variables like check size, frequency of investments, and conversion rates, you can see how it impacts your available capital.
One more tool we found helpful is a return scenario model. Instead of modeling expected returns per company, we use the return scenarios to see how long the fund’s capital will last depending on its annual return performance.
On the Summary and Assumptions tab (beginning in cell C123), there is a table for three different potential scenarios where you can enter returns you might expect in different years. These rates of return are then pulled into the Model tab to model out each scenario. The results are then graphed on the Summary and Assumptions tab with each scenario’s expected cash max out date.
This model was built as an internal tool. While we’ve tried to make it as user-friendly as possible (including explanations in comments throughout the model), there are likely still some areas that are clunky or unintuitive. As with all models, we recommend you spend time understanding how it works before trusting its outputs, both to make sure you agree with our methods and in case there are any errors we haven’t caught yet.