Accounting for client Life Time Value (LTV) in growth experiements

Seeing the planning of the referral program and then modeling Return On Ad Spend (ROAS) got me thinking we should have a discussion about calculating client LTV and then codify a framework around it.

The referral program’s rewards of between 5%-10% of the DPI asset value driven given out in INDEX mean that with DPI streaming fees to Coop of 0.60%/year we’re accounting that the pay back will be between 8.33-16.67 years of holding DPI (assuming no other Index Coop products bought and held).

As the simple ROAS model shows, even if we achieve a solid Click Through Rate (CTR) on our ads and conversion rate on our landing page, if we account that the client only holds DPI (according to some median behaviors) for one year, it’s hard to make ads pay. But, if we assume holding periods of more than one year and clients buying more Index Coop products, ROAS becomes positive.

Given we’re building an investment funds platform - and noting how other platforms like Amazon and Facebook have financed their growth - I think we need to codify a strategic LTV framework and use this in our ad, content, referral, social, and other channels. I think the major inputs which go into this, which we have some control over, are:

  1. A number of years of holding period: quality of our product hopefully drives long holding periods. I was thinking 3-5 could be reasonable.

  2. Purchase of more Index Coop products: quality of our products hopefully also drives purchase of other products after DPI. I was thinking 50% of clients might buy non-DPI products, maybe buying one more each year after the DPI purchase.

  3. Do we use fund assets and successfully earn revenues doing productive things with them, similar to TradFi index managers?

There’s obviously other inputs into the simple ROAS model we cannot control, such as price of DPI. And, other inputs we can discuss, such as average DPI unit holding size (I used median, which is potentially conservative. Thanks to @jdcook for help here) and streaming fee % of other products to the Coop.

Ultimately, with @dylan and @ztcrypto’s purchase event tracking infra (nearly built) we can see what asset value and streaming revenue is specifically driven from which campaign, but we still need an LTV framework to evaluate that against.

Owls: what do you think:

A) We should use as inputs for 1 and 2?
B) We should assume for revenue generation of fund assets via productive uses?
C) Other considerations are for the Index Coop for codifying LTV?

Will be great to discuss and this and align on a framework.


UPDATE:
Here’s a strawman for reference using some average (median) client behaviors:

DPI:

  • DPI holding of average client (median): 2.5 units
  • DPI price: $400 (this is subject to quite rapid change)
  • USD value of DPI held by average client: $1,000
  • DPI streaming fee to Coop: 0.6%/year
  • Streaming fees from average client - Year 1: $6
  • 3 Years DPI streaming fees/client: $18, 5 Years: $30

CROSS SELL INTO OTHER INDEX COOP PRODUCTS:

  • Rate average client buys more Index Coop products: 50%
  • No. of extra products available during holding period: 4 (this is conservative, IMO - could be quite a bit more)
  • Other product holding size: same dollar value of DPI holding - $1,000
  • Other product streaming fees to Coop: 0.6%/year (could be more of less, depending on products being harder/simpler to execute and their streaming fees)
  • Other product streaming fees from average client - Year 1: $6
  • 3 Years other product streaming fees/client: $18, 5 Years: $30

CLIENT LTV:

  • 3 Year holding period: $36, 5 Year: $60

NON-STREAMING FEE REVENUE:

  • How much could the Coop make leveraging fund assets as TradFi index managers do?
  • Assumption: 0.6% (other Owls might have stronger ideas here. Looking at you especially @overanalyser)
    • 3 Year holding period: $36, 5 Year: $60

COMBINED LTV:

  • 3 Year holding period: $72, 5 Year: $120

(If the price of DPI (/other products) goes up 1x, 2x, 3x, these numbers change drastically).

3 Likes

Yes we will discuss about this and updatre.
@0x_Dev you can tag me by @ztcrypto. :slight_smile:

2 Likes

Thanks @ztcrypto and congrats on the first post!

1 Like

Big time post! As always, awesome stuff.

I think we should be taking a look at this as an example of how to get rough spend justification!

Of course, can be tuned more and adapted on a persona basis. I think this kind of modeling, especially when we’re discussing 6-7 figure outlays, is a good template.

On thing that strikes me is accounting for churn and unit retention could help harden this.

Also curious to hear @dylan and @puniaviision thoughts on this as a model for spend justification and how we can improve.

hey @0x_Dev, thanks for putting this together and the number of assumptions in here makes my head spin :upside_down_face:

Some thoughts from me:

  • Holding period of 5 years seems rather high. I personally think we might see a lot of churn during the bear market. 2 or 3 years makes more sense imo. At least until we go through a bear market and have some data on user behaviour.

  • I do think that DPI holders will buy more Index products. 50% of customers buying 1 more product each year is reasonable in a bull market, but not in a bear market imo (unless we come up with products that offer short exposure or hedging). Across cycles, I would maybe say 33% not 50%, but also depends on how many products we have on offer. Going from 1 to 2 to 3 is reasonable, but at some point we might reach saturation per customer.

  • I think intrinsic productivity is a must. 0.6% is rather low, I see maybe 1% or 1.5% as being conservative enough.

2 Likes

Hi Dev, Apologies for taking so long to reply.
This isn’t really my area of expertise so will be brief. It’s a great initiative so I look forward to seeing the outcomes :+1:

Just to echo verto, I think a shorter time frame is the best approach given the unknowns. 1-2 Additional products I think is also a fair assumption with diminishing returns to be expected after that.
Everyone is onboard with intrinsic! Will obviously vary with product and market conditions. Again agree with AG between 1-1.5% Maybe 2%. Let be optimistic :crossed_fingers:

1 Like

This is a suuuper interesting conversation to kick off…

First,+1 to @verto0912’s takes

Second, there is a lot here, so I’m going to lob some Qs over the fence:

1. What is the desired outcome of this post? (my “i think this is right” answer: to begin making data-informed decisions around big-ticket spending)
2. What analogous frameworks or models exist? Amazon/Facebook are mentioned, though i’d imagine financial product providers might offer better examples
3. What is needed to get an MVP version of this off the ground?

I won’t claim expertise on this topic, so I’m ^ pushing more to figure out what it’ll take to put a quick/dirty & accurate-enough model into practice so we can learn!

Many thanks for the comments Owls,

Replying in one large post @'ing folks along the way.

@LemonadeAlpha: thanks for the comments. Agreed we should use this as an example, which can can strategically revisit over time too. As you say, different personas will behave differently - and we can dig into that later too. Hopefully keeping holding period to 3 years and cross sell rate to 35% - more humble than my initial numbers - helps respect churn.

@verto0912: thanks for the feedback. Really helpful. I’ve settled on holding period of 3 years and 35% cross sell rate in the revised V2 model. Great to learn your thoughts re intrinsic productivity yield too - that’s encouraging!

@MrMadila: thanks also for the feedback. In the V2 model I’ve (conservatively) used 1% yield from intrinsic productivity, which results in an implied LTVROAS of 239%. But, if we yield 1.5% LTVROAS increases to 275% - good to know.

@anon10525910: thanks for the questions.

  1. The desired outcome of this post is to receive feedback (which has very helpfull come in), before I make a new post with a revised V2 ROAS model and a poll. Hopefully the poll vote is sufficiently positive to settle on this update model and use it in the trenches for activities in ads, content, biz dev, etc
  2. Reference point frameworks would exist most closely in TradFi index providers, as you suggest. The main thing here is to account for revenue driven from client fees and the asset managers trading/lending fees - so we’re covering both bases here. See below for comments on these reveue breakdown too
  3. I think we need the abovementioned 2nd post and a poll vote to get the Coop’s reasonable input, then proceed with the agreed (or further revised) model in growth experiments generally

All Owls:

  • Revised V2 ROAS model can be viewed here, in Tab 1
  • This feels like a solid, reasonable basis to appraise our growth work against for the next few months before we revisit this with more data
  • I like the fact that we’ve made holding period and cross sell rate conservative
  • The implied revenue splits which flow out of this model have a quite balanced, diversified appearance (screenshot below). This seems reassuring to me, though if intrinsic productivity yield is 1.5% or more, the split does start to skew a bit

Screen Shot 2021-03-14 at 14.19.53

  • I will leave this a few more hours to let others see it during the weekend, then create a second post on this subject with a poll at the bottom