< home

TikTok's Math

Assume TikTok acquires a cohort of NN users, with an average customer acquisition cost of CACCAC per user.

Then the total contribution profit generated by this cohort can be written as:

Profit=N×(LTVCAC)Profit = N \times (LTV - CAC)

Here, LTVLTV means the lifetime gross profit per user, excluding acquisition cost. It can be decomposed as:

LTV=t=1TRt×(ARPDAUtCTSt)LTV = \sum_{t=1}^{T} R_t \times (ARPDAU_t - CTS_t)

where RtR_t is the retention rate of this cohort on day tt, ARPDAUtARPDAU_t is the average revenue per daily active user on day tt, and CTStCTS_t is the cost to serve one active user on day tt.

For an advertising product like TikTok, ARPDAUtARPDAU_t can be further decomposed as:

ARPDAUt=MPUt×AdLoadt×CPMt1000ARPDAU_t = MPU_t \times AdLoad_t \times \frac{CPM_t}{1000}

where: MPUtMPU_t is minutes per active user on day tt, AdLoadtAdLoad_t is the number of ad impressions per minute, and CPMtCPM_t is the revenue per thousand ad impressions.

Therefore:

LTV=t=1TRt×(MPUt×AdLoadt×CPMt1000CTSt)LTV = \sum_{t=1}^{T} R_t \times \left( MPU_t \times AdLoad_t \times \frac{CPM_t}{1000} - CTS_t \right)

And the cohort-level profit becomes:

Profit=N×[t=1TRt×(MPUt×AdLoadt×CPMt1000CTSt)CAC]Profit = N \times \left[ \sum_{t=1}^{T} R_t \times \left( MPU_t \times AdLoad_t \times \frac{CPM_t}{1000} - CTS_t \right) - CAC \right]

Since the number of active users from this cohort on day tt is:

DAUt=N×RtDAU_t = N \times R_t

we can rewrite the same formula as:

Profit=t=1TDAUt×(ARPDAUtCTSt)N×CACProfit = \sum_{t=1}^{T} DAU_t \times (ARPDAU_t - CTS_t) - N \times CAC

Or fully expanded:

Profit=t=1TDAUt×(MPUt×AdLoadt×CPMt1000CTSt)N×CACProfit = \sum_{t=1}^{T} DAU_t \times \left( MPU_t \times AdLoad_t \times \frac{CPM_t}{1000} - CTS_t \right) - N \times CAC