Ask YC: How do you project web traffic? | Hacker News Reader

Ask YC: How do you project web traffic?

11 pointsjon_dahl8 comments

As I raise funding for my startup, investors want to see pro forma financial projections. I think I can estimate our expenses fairly well for a year or two, and some of our revenue assumptions (e.g. subscription cost, ad CPMs). But our revenue depends on two unpredictable things:

-- traffic -- conversion rates

This is crystal ball territory, and everyone knows it. But we still need to make claims and substantiate them.

Has anyone seen a good model for estimating traffic based on assumptions? Any insight into what to give our investors?

View on HN

You can't predict traffic up or down meaning your conservative estimates might be 10X high or low.

Conversation rates depend on a whole bunch of parameters like the content of your site, traffic, etc. I have seen conversion rates as high as 5-15% and as low as .5%.

Take a look at Powerset's growth modeling: http://www.blognewcomb.com/blog/2007/04/how_is_powerset_pred...

That said, as antiismist mentions, model your traffic and conversion rates 10X higher & than lower starting from what your predict.

If you are making claims it should be easy to substantiate :-)

This is one of the topics I always touch upon when talking to web entrepreneurs who've launched sites. I usually ask it in terms of scaling and how fast someone should worry about it. What people tell me is that they were either surprised by how fast their site grew or they've been disappointed by the slow uptake. In either case, the models they've created were pointless.

So worry less about modeling and spend more time planning for the best (or worst) case when you have so much traffic coming in that you have to scale up. Modeling is useful IF you're paying for the traffic. If you're buying traffic and you have X amount of dollars to spend per month, then it's fairly easy to project numbers. I'm assuming most of us here are building community sites and are depending on either viral or SEO traffic so you're left to more guessing.

Edit: I just looked at the Powerset modeling they released and looked at their actual numbers on alexa/compete and you can see they're absolutely nothing like they projected.

Think of it less as a "projection" as much as a "what if" exercise.

Do a few different scenarios: Single-digit, Double-Digit, Exponential, Hockey-Stick, Shark-Fin, and explain how you would operate in each of these situations, and how you would perform financially. Then show them what you think are the most likely scenarios.

You must not be dealing with typical tech investors if they're asking for those sorts of things. Tech investors generally don't ask precisely because they are meaningless.

The best you can do is look at your nearest competitor. Pro forma numbers work great for the guy who ran one hotel and is opening another down the road. They can justify using the same occupancy rate, etc.

Using your competitor's historic traffic data is, of course, meaningless, but to someone who is asking you for numbers that are by definiton meaningless guesses, it probably counts as justification.

I just spend $125 on Adwords sending traffic to a barebones/not really launched site with a sales page and found out about 25% of visitors click to my registration page and 10% sign up for a free account. Once I figure out how to monetize those accounts (it's something health related) and get more than $5 in revenue each it'll be a gold mine as this is something that 50% of the population will deal with in their life (not gender based).

One of the beauties of Adwords and especially the content network is that you get an idea of how many people are interested in you topic/market.

So if you're doing something subscription bases, throw up a prelaunch sales page, spend $100 and see what you get. If that's the only data holding you back from you VCs, consider it money well spent.

I'm having a similar problem, mainly with revenue projections with advertising.

So I'm just building out a table - each row in the table is some level of traffic, and each column has some projection of yield (yield = CTR X CPC), like reasonable worst case, probable case, and reasonable best case.

Basically, identify the variables in the model. Come up with some reasonable ranges for what those could be, put those in a table. For example, for CPC advertisements, the CTR that I have heard about range from a minimum .05% to a maximum of 30%, and the CPC ranges from $.01 to $100 (e.g. for mesothelioma related ads at one point).

Honestly, you should use something like the Drake Equation:

http://en.wikipedia.org/wiki/Drake_equation

It's impossible don't get hung up on it. As an early stage startup you should have one concern--build the product. A working product is more impressive than artificial projections.

Reply on news.ycombinator.com