SOV Advertiser Retention

Executive Summary

SOV keeps few advertisers month to month. Of advertisers whose first charged SOV day was in a given month, 26% are active again the next month, 11% in month 3, 8% in month 6 and 7% in month 12. The curve flattens after month 6 into a small core of about 5–7% that keeps buying.
Revenue holds up better than the advertiser count. Month-12 spend is 19% of the cohort’s first-month spend (month 1: 44%). The advertisers who stay spend about 2.7× as much as the cohort average did in its first month.
Use is short and on/off, not a subscription. The median advertiser buys 5 charged days in total, and 79% of first campaigns last 3 days or less. Still, 48% come back after a gap of more than 7 days (median 55 days later), and reactivated advertisers make up about 30% of each month’s active base.
Owners churn about twice as fast as brokers after the first month. Owners are advertisers whose first SOV listing went through Aqar’s owner licensing flow (licensing_requests). They are a third of advertisers. They return in month 1 at the same rate as brokers (25% vs 27%), but by month 12 only 4.6% are still buying, against 8.5% of brokers. Owner revenue retention at month 12 is 13%, against 22% for brokers. Owners’ share of SOV spend fell from about 35% in mid-2024 to 14% in September 2026.
Retention has not improved. Month-1 retention has stayed at 24–31% for every quarterly cohort since 2024. Month-12 retention dropped from 11% (2024 Q1 cohort) to about 5% (2025 cohorts).

Definitions

Advertiser = user_id in sadb_sov_daily_biddings.
Charged SOV day = a bidding row with status = 4 and non_chargeable = false, the same filter as the SOV pricing simulator. This drops campaigns that were created but never ran: 22.1k users created a campaign, 14.6k were ever charged.
Cohort = the calendar month of an advertiser’s first charged day (from 2024-03, the first full month of SOV).
Month-k logo retention = share of the cohort with at least one charged day in calendar month k after the cohort month. Calendar months flatter month 1 slightly, because a campaign that starts on the 30th and runs into the next month counts as retained.
Revenue retention = cohort spend in month k ÷ cohort spend in month 0 (budget is in points; 100 points = 1 SAR).
Only cohorts old enough to observe month k are included. October 2026 (partial) is excluded.

Retention Curve (All Cohorts Pooled)

Month k	Eligible advertisers	Active in month k	Logo retention	Revenue retention	Came back at any point by month k
1	14,229	3,755	26.4%	43.5%	26.4%
2	13,847	1,974	14.3%	27.2%	31.8%
3	13,487	1,496	11.1%	22.5%	35.4%
6	12,484	1,012	8.1%	19.4%	41.5%
12	10,634	758	7.1%	19.1%	48.0%
18	8,278	488	5.9%	17.9%	52.7%
24	5,526	299	5.4%	17.8%	57.4%

The gap between the last two columns shows the on/off pattern. By month 12 about half the cohort has bought SOV again at some point, but only 7% are buying in any given month.

SQL Query

WITH
  um AS (
    SELECT
      user_id,
      toStartOfMonth(start_datetime) m,
      sum(budget) / 100 sar
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
      AND start_datetime < '2026-10-01'
    GROUP BY
      user_id,
      m
  ),
  f AS (
    SELECT
      user_id AS uid,
      argMin(m, m) cohort,
      argMin(sar, m) sar0
    FROM
      um
    GROUP BY
      user_id
  ),
  j AS (
    SELECT
      f.uid AS uid,
      f.cohort AS cohort,
      dateDiff('month', f.cohort, um.m) k,
      um.sar AS sar
    FROM
      um
      JOIN f ON um.user_id = f.uid
  ),
  ks AS (
    SELECT
      arrayJoin([1, 2, 3, 6, 12, 18, 24]) kk
  ),
  base AS (
    SELECT
      kk,
      uniqExact(uid) eligible,
      sum(sar0) sar0
    FROM
      f
      CROSS JOIN ks
    WHERE
      cohort <= addMonths(toDate('2026-09-01'), - kk)
    GROUP BY
      kk
  ),
  act AS (
    SELECT
      kk,
      uniqExactIf (uid, k = kk) active,
      sumIf (sar, k = kk) sar_k,
      uniqExactIf (
        uid,
        k >= 1
        AND k <= kk
      ) ever_back
    FROM
      j
      CROSS JOIN ks
    WHERE
      cohort <= addMonths(toDate('2026-09-01'), - kk)
    GROUP BY
      kk
  )
SELECT
  kk,
  eligible,
  active,
  round(active / eligible * 100, 1) logo_ret,
  round(sar_k / sar0 * 100, 1) rev_ret,
  round(ever_back / eligible * 100, 1) returned_within_k
FROM
  base
  JOIN act USING kk
ORDER BY
  kk

Owners vs Brokers

An advertiser is an owner if the listing on their first charged SOV day has a row in licensing_requests, the table behind Aqar’s flow for issuing ad licenses to owners. Everyone else is a broker: their listing carries a license they got through REGA themselves. licensing_requests is only in MySQL, so the queries in this section run on Metabase database 2 against the MySQL sov_daily_biddings. With the same filters, the MySQL counts match ClickHouse exactly (14,229 advertisers eligible for month 1).

Owners make up 4,791 of the 14,229 advertisers (34%). Only 121 of the 7.6k owner-flow listings promoted with SOV never got a license issued, so including unissued requests makes no practical difference.

Retention Curve

Month k	Broker eligible	Broker logo	Broker revenue	Broker came back by k	Owner eligible	Owner logo	Owner revenue	Owner came back by k
1	9,438	26.9%	43.6%	26.9%	4,791	25.3%	43.3%	25.3%
2	9,179	15.9%	29.6%	33.4%	4,668	11.0%	22.2%	28.8%
3	8,934	12.5%	24.5%	37.6%	4,553	8.3%	18.3%	31.1%
6	8,270	9.3%	22.1%	44.6%	4,214	5.8%	13.6%	35.4%
12	6,966	8.5%	22.2%	52.5%	3,668	4.6%	12.7%	39.5%
18	5,346	7.2%	19.7%	57.4%	2,932	3.4%	14.2%	44.0%
24	3,604	6.5%	21.1%	62.2%	1,922	3.3%	11.0%	48.4%

Both groups look the same in month 1, then split. An owner usually has one property to sell or rent, so once it is gone there is nothing left to promote. Brokers keep getting new inventory, and 62% of them come back at some point within two years.

Usage

Advertisers with a first charged day before 2026-09:

Metric	Broker	Owner
Advertisers	9,438	4,791
Lifetime spend (SAR)	6.92 M	2.21 M
Average lifetime spend per advertiser (SAR)	733	462
Average charged days	22.2	13.3
Lifetime ≤ 7 charged days	61.0%	71.3%
Ran 2+ campaigns	62.4%	48.3%
Average listings promoted	4.21	1.77
Promoted 2+ listings	55.6%	25.9%

One in four owners promoted more than one listing. These are likely owners with several properties, or small brokers who list through the owner flow.

By Category

The category is the one targeted on the advertiser’s first charged day. Only segments with at least 150 advertisers are shown.

Category	Segment	n (M1)	Month 1	Month 3	n (M12)	Month 12
Apartment for Rent	Broker	1,416	27.7%	14.0%	1,056	11.9%
Apartment for Rent	Owner	1,800	24.7%	7.7%	1,342	4.4%
Land for Sale	Broker	1,698	25.6%	10.9%	1,292	8.0%
Land for Sale	Owner	780	26.8%	8.9%	614	4.4%
Villa for Sale	Broker	1,880	28.7%	13.8%	1,386	7.0%
Villa for Sale	Owner	547	28.7%	10.1%	433	3.7%
Apartment for Sale	Broker	1,691	29.7%	14.5%	1,237	8.7%
Apartment for Sale	Owner	362	29.3%	9.0%	275	6.2%
Villa for Rent	Broker	245	23.7%	13.7%	181	9.4%
Villa for Rent	Owner	293	22.2%	6.1%	225	3.6%
Building for Sale	Broker	526	25.5%	8.5%	367	8.7%
Building for Sale	Owner	176	19.9%	6.5%	131	4.6%
Big Flat for Rent	Owner	166	23.5%	4.4%	122	5.7%
Big Flat for Sale	Broker	451	29.0%	13.4%	293	6.1%
Building for Rent	Broker	306	27.1%	10.2%	219	9.6%
Apartment for booking	Broker	291	15.1%	6.2%	260	2.7%

The owner gap holds in every category. It is not caused by owners being concentrated in low-retention categories. Apartment for Rent shows the biggest gap: it is the only category with more owner advertisers than broker advertisers, and brokers keep 11.9% at month 12 against 4.4% for owners.

Month	Owner advertisers	Broker advertisers	Owner share of spend
2024-04	373	583	37.6%
2024-08	530	1,106	27.4%
2024-12	419	881	24.3%
2025-04	285	766	21.5%
2025-08	286	977	20.6%
2025-12	275	838	21.6%
2026-04	292	988	18.2%
2026-08	275	972	14.7%
2026-09	287	933	14.1%

Here each charged day is classified by its own listing, not by the advertiser’s first listing. The number of owner advertisers per month has held at about 250–300 since 2025, but their share of spend keeps falling. Owners now spend less per advertiser, while broker spend grows.

SQL Query

-- Metabase database 2 (MySQL): retention curve by owner/broker
WITH
  b AS (
    SELECT
      user_id,
      listing_id,
      start_datetime,
      budget
    FROM
      sov_daily_biddings
    WHERE
      status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
      AND start_datetime < '2026-10-01'
  ),
  fl AS (
    SELECT
      user_id,
      listing_id
    FROM
      (
        SELECT
          user_id,
          listing_id,
          ROW_NUMBER() OVER (
            PARTITION BY
              user_id
            ORDER BY
              start_datetime,
              listing_id
          ) rn
        FROM
          b
      ) t
    WHERE
      rn = 1
  ),
  lr AS (
    SELECT DISTINCT
      listing_id
    FROM
      licensing_requests
  ),
  seg AS (
    SELECT
      fl.user_id,
      IF(lr.listing_id IS NULL, 'broker', 'owner') seg
    FROM
      fl
      LEFT JOIN lr ON lr.listing_id = fl.listing_id
  ),
  um AS (
    SELECT
      user_id,
      DATE(DATE_FORMAT(start_datetime, '%Y-%m-01')) m,
      SUM(budget) / 100 sar
    FROM
      b
    GROUP BY
      user_id,
      m
  ),
  f AS (
    SELECT
      user_id,
      MIN(m) cohort
    FROM
      um
    GROUP BY
      user_id
  ),
  j AS (
    SELECT
      um.user_id,
      seg.seg,
      f.cohort,
      TIMESTAMPDIFF(MONTH, f.cohort, um.m) k,
      um.sar
    FROM
      um
      JOIN f ON f.user_id = um.user_id
      JOIN seg ON seg.user_id = um.user_id
  ),
  ks AS (
    SELECT
      1 kk
    UNION ALL
    SELECT
      2
    UNION ALL
    SELECT
      3
    UNION ALL
    SELECT
      6
    UNION ALL
    SELECT
      12
    UNION ALL
    SELECT
      18
    UNION ALL
    SELECT
      24
  )
SELECT
  seg,
  kk,
  COUNT(DISTINCT j.user_id) eligible,
  COUNT(
    DISTINCT CASE
      WHEN k = kk THEN j.user_id
    END
  ) active,
  ROUND(
    COUNT(
      DISTINCT CASE
        WHEN k = kk THEN j.user_id
      END
    ) / COUNT(DISTINCT j.user_id) * 100,
    1
  ) logo_ret,
  ROUND(
    SUM(
      CASE
        WHEN k = kk THEN sar
      END
    ) / SUM(
      CASE
        WHEN k = 0 THEN sar
      END
    ) * 100,
    1
  ) rev_ret,
  ROUND(
    COUNT(
      DISTINCT CASE
        WHEN k BETWEEN 1 AND kk THEN j.user_id
      END
    ) / COUNT(DISTINCT j.user_id) * 100,
    1
  ) back_by_k
FROM
  j
  JOIN ks ON j.cohort <= DATE_SUB ('2026-09-01', INTERVAL ks.kk MONTH)
GROUP BY
  seg,
  kk
ORDER BY
  seg,
  kk

The category split uses the same b/fl/lr/seg/um/f/j CTEs, with listing_category carried through fl, grouped by cat, seg with the month 1/3/12 conditions from the category × city query above. The usage table aggregates b per user (COUNT(*), COUNT(DISTINCT sov_campaign_id), COUNT(DISTINCT listing_id), SUM(budget) / 100) and joins seg. The monthly share query:

-- Metabase database 2 (MySQL): owner share of monthly SOV (classified per charged day)
SELECT
  DATE_FORMAT(b.start_datetime, '%Y-%m') m,
  COUNT(
    DISTINCT CASE
      WHEN lr.listing_id IS NOT NULL THEN b.user_id
    END
  ) owner_users,
  COUNT(
    DISTINCT CASE
      WHEN lr.listing_id IS NULL THEN b.user_id
    END
  ) broker_users,
  ROUND(
    SUM(
      CASE
        WHEN lr.listing_id IS NOT NULL THEN b.budget
      END
    ) / SUM(b.budget) * 100,
    1
  ) owner_spend_pct
FROM
  sov_daily_biddings b
  LEFT JOIN (
    SELECT DISTINCT
      listing_id
    FROM
      licensing_requests
  ) lr ON lr.listing_id = b.listing_id
WHERE
  b.status = 4
  AND b.non_chargeable = 0
  AND b.start_datetime >= '2024-03-01'
  AND b.start_datetime < '2026-10-01'
GROUP BY
  m
ORDER BY
  m

Retention by Cohort Quarter

First-charge quarter	Advertisers	Month 1	Month 3	Month 6	Month 12
2024 Q1 (Mar only)	1,450	25.7%	14.8%	14.3%	11.2%
2024 Q2	1,909	30.5%	13.9%	10.1%	7.7%
2024 Q3	2,167	26.1%	10.0%	7.0%	8.0%
2024 Q4	1,563	24.2%	8.4%	7.8%	6.2%
2025 Q1	1,189	24.1%	10.3%	6.5%	5.7%
2025 Q2	1,115	26.0%	10.5%	5.9%	4.9%
2025 Q3	1,241	25.9%	10.3%	5.2%	4.5%
2025 Q4	1,002	26.5%	10.3%	8.0%
2026 Q1	848	28.9%	11.4%	6.2%
2026 Q2	1,003	25.1%	10.1%
2026 Q3 (Jul–Aug)	742	26.3%

Early retention (months 1 and 3) has held steady. Long-term retention was higher for the March 2024 launch cohort, which probably held the most committed early adopters. New-advertiser volume has also fallen, from about 2,000 a quarter in 2024 to about 1,000 in 2026.

SQL Query

WITH
  um AS (
    SELECT
      user_id,
      toStartOfMonth(start_datetime) m,
      sum(budget) / 100 sar
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
      AND start_datetime < '2026-10-01'
    GROUP BY
      user_id,
      m
  ),
  f AS (
    SELECT
      user_id AS uid,
      min(m) cohort
    FROM
      um
    GROUP BY
      user_id
  ),
  j AS (
    SELECT
      f.uid AS uid,
      toStartOfQuarter(f.cohort) q,
      f.cohort AS cohort,
      dateDiff('month', f.cohort, um.m) k
    FROM
      um
      JOIN f ON um.user_id = f.uid
  )
SELECT
  q,
  uniqExact(uid) users,
  round(
    uniqExactIf (uid, k = 1) / uniqExact(uid) * 100,
    1
  ) m1,
  if(
    max(cohort) <= '2026-06-01',
    round(
      uniqExactIf (uid, k = 3) / uniqExact(uid) * 100,
      1
    ),
    NULL
  ) m3,
  if(
    max(cohort) <= '2026-03-01',
    round(
      uniqExactIf (uid, k = 6) / uniqExact(uid) * 100,
      1
    ),
    NULL
  ) m6,
  if(
    max(cohort) <= '2025-09-01',
    round(
      uniqExactIf (uid, k = 12) / uniqExact(uid) * 100,
      1
    ),
    NULL
  ) m12
FROM
  j
WHERE
  cohort <= '2026-08-01'
GROUP BY
  q
ORDER BY
  q

Retention by Category and City

Segments are set by the targeting (listing_category, city_id) of the advertiser’s first charged day. The table shows segments with at least 150 advertisers.

Category	City	n (M1)	Month 1	Month 3	Month 12
Apartment for Rent	Riyadh	1,998	26.9%	11.3%	9.1%
Villa for Sale	Riyadh	1,269	31.0%	14.7%	6.6%
Apartment for Sale	Jeddah	855	30.8%	14.3%	8.4%
Land for Sale	Riyadh	812	26.7%	11.1%	5.9%
Apartment for Rent	Jeddah	599	25.4%	7.5%	6.7%
Apartment for Sale	Riyadh	586	31.4%	13.9%	9.9%
Big Flat for Sale	Riyadh	453	30.2%	14.8%	6.6%
Land for Sale	Jeddah	397	29.0%	10.5%	7.0%
Villa for Rent	Riyadh	375	25.9%	11.7%	7.7%
Villa for Sale	Jeddah	333	26.4%	8.9%	6.5%
Apartment for Rent	Khobar	235	26.4%	9.3%	2.5%
Big Flat for Rent	Riyadh	232	24.6%	6.9%	7.2%
Villa for Sale	Dammam	208	27.4%	11.1%	4.2%
Apartment for booking	Riyadh	202	14.4%	6.4%	2.7%
Building for Rent	Riyadh	197	29.9%	12.2%	12.4%
Building for Sale	Riyadh	196	27.6%	12.2%	11.3%
Apartment for Sale	Dammam	195	32.3%	12.0%	6.4%
Building for Sale	Jeddah	164	22.0%	7.1%	9.7%
Land for Sale	Madinah	161	24.2%	12.5%	5.6%

Sale listings retain better early. Apartments, villas and big flats for sale in Riyadh, Jeddah and Dammam reach 30–32% in month 1 and about 14% in month 3. Rentals sit at 25–27% and 7–11%.
Buildings (rent and sale) have the stickiest long-term core, at 10–12% in month 12. These are likely professional brokers with recurring inventory.
Bookings units retain worst: 14% in month 1 and 3% in month 12 for Riyadh apartments for booking.
The Eastern Province (Khobar rentals, Dammam villas) loses advertisers faster after month 3.

SQL Query

WITH
  um AS (
    SELECT
      user_id,
      toStartOfMonth(start_datetime) m,
      sum(budget) / 100 sar,
      argMin(listing_category, start_datetime) cat0,
      argMin(city_id, start_datetime) city0
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
      AND start_datetime < '2026-10-01'
    GROUP BY
      user_id,
      m
  ),
  f AS (
    SELECT
      user_id AS uid,
      min(m) cohort,
      argMin(cat0, m) cat,
      argMin(city0, m) city
    FROM
      um
    GROUP BY
      user_id
  ),
  j AS (
    SELECT
      f.uid uid,
      f.cat cat,
      f.city city,
      f.cohort cohort,
      dateDiff('month', f.cohort, um.m) k
    FROM
      um
      JOIN f ON um.user_id = f.uid
  ),
  cities AS (
    SELECT
      city_id,
      any(city_name) city_name
    FROM
      sadb_districts FINAL
    GROUP BY
      city_id
  )
SELECT
  cg.name_en category,
  ci.city_name city,
  uniqExactIf (uid, cohort <= '2026-08-01') n_m1,
  round(
    uniqExactIf (
      uid,
      k = 1
      AND cohort <= '2026-08-01'
    ) / n_m1 * 100,
    1
  ) m1,
  uniqExactIf (uid, cohort <= '2026-06-01') n_m3,
  round(
    uniqExactIf (
      uid,
      k = 3
      AND cohort <= '2026-06-01'
    ) / n_m3 * 100,
    1
  ) m3,
  uniqExactIf (uid, cohort <= '2025-09-01') n_m12,
  round(
    uniqExactIf (
      uid,
      k = 12
      AND cohort <= '2025-09-01'
    ) / n_m12 * 100,
    1
  ) m12
FROM
  j
  LEFT JOIN sadb_categories cg FINAL ON cg.category_id = j.cat
  LEFT JOIN cities ci ON ci.city_id = j.city
GROUP BY
  category,
  city
HAVING
  n_m1 >= 150
ORDER BY
  n_m1 DESC
LIMIT
  25

How Advertisers Use SOV

Advertisers whose first charged day was before 2026-09 (14,229):

Metric	Value
Ran 1 campaign only	42.1%
Ran 2+ campaigns	57.9%
Ran 5+ campaigns	23.9%
Median campaigns per advertiser	2
Median charged days, lifetime	5
Median charged days, first campaign	2
First campaign ≤ 3 days	78.9%
Lifetime ≤ 7 charged days	64.3%
Lifetime ≥ 90 charged days	3.2%

Activity is grouped into spells: runs of charged days with gaps of no more than 7 days. Advertisers with a first charged day before 2026-07 (13,487):

Metric	Value
Came back after a gap of more than 7 days	47.5%
…within 90 days of the first charged day	29.8%
Median days from first day to return	55
Average spells per advertiser	2.59
4+ spells	19.0%

Most advertisers try SOV for a day or two. The ones who come back use it in short bursts, most likely when they have a new listing to push, rather than as an always-on budget.

SQL Query

-- campaigns per advertiser
WITH
  c AS (
    SELECT
      user_id,
      sov_campaign_id,
      count() cdays,
      min(start_datetime) st
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
    GROUP BY
      user_id,
      sov_campaign_id
  ),
  u AS (
    SELECT
      user_id,
      count() campaigns,
      sum(cdays) tdays,
      argMin(cdays, st) first_days,
      min(st) first_st
    FROM
      c
    GROUP BY
      user_id
  )
SELECT
  count() users,
  round(countIf (campaigns = 1) / count() * 100, 1) one_campaign_pct,
  round(countIf (campaigns >= 2) / count() * 100, 1) repeat_pct,
  round(countIf (campaigns >= 5) / count() * 100, 1) ge5_pct,
  quantileExact(0.5) (campaigns) med_campaigns,
  quantileExact(0.5) (tdays) med_total_days,
  quantileExact(0.5) (first_days) med_first_campaign_days,
  round(countIf (first_days <= 3) / count() * 100, 1) first_le3days_pct,
  round(countIf (tdays <= 7) / count() * 100, 1) lifetime_le7days_pct,
  round(countIf (tdays >= 90) / count() * 100, 1) lifetime_ge90days_pct
FROM
  u
WHERE
  first_st < '2026-09-01'

-- spells (gaps > 7 days start a new spell)
WITH
  d AS (
    SELECT
      user_id,
      groupArray(DISTINCT toDate(start_datetime)) ds
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
    GROUP BY
      user_id
  ),
  s AS (
    SELECT
      user_id,
      arraySort(ds) a,
      arrayFilter(
        (x, i) -> i = 1
        OR dateDiff('day', a[i - 1], x) > 7,
        a,
        arrayEnumerate(a)
      ) spell_starts,
      a[1] first_day
    FROM
      d
  )
SELECT
  count() users,
  round(
    countIf (length(spell_starts) >= 2) / count() * 100,
    1
  ) came_back_after_7d_gap_pct,
  round(
    countIf (
      length(spell_starts) >= 2
      AND dateDiff('day', first_day, spell_starts[2]) <= 90
    ) / count() * 100,
    1
  ) back_within_90d_pct,
  quantileExactIf (0.5) (
    dateDiff('day', first_day, spell_starts[2]),
    length(spell_starts) >= 2
  ) med_days_to_return,
  round(avg(length(spell_starts)), 2) avg_spells,
  round(
    countIf (length(spell_starts) >= 4) / count() * 100,
    1
  ) ge4_spells_pct
FROM
  s
WHERE
  first_day < '2026-07-01'

Monthly Active Advertiser Mix

Month	Active	New	Continuing (active last month)	Reactivated (gap ≥ 1 month)	New share
2025-01	1,149	410	439	300	35.7%
2025-04	1,038	346	386	306	33.3%
2025-07	1,215	406	433	376	33.4%
2025-10	1,173	370	485	318	31.5%
2026-01	1,078	290	438	350	26.9%
2026-04	1,276	376	447	453	29.5%
2026-07	1,222	360	495	367	29.5%
2026-08	1,240	382	495	363	30.8%
2026-09	1,214	308	520	386	25.4%

The active base is flat at about 1,100–1,250 advertisers a month. It is now less dependent on new advertisers (36% → 25–30%). The continuing segment grew from about 440 to about 500, while new-advertiser acquisition shrank.

SQL Query

WITH
  um AS (
    SELECT DISTINCT
      user_id,
      toStartOfMonth(start_datetime) m
    FROM
      sadb_sov_daily_biddings FINAL
    WHERE
      _peerdb_is_deleted = 0
      AND status = 4
      AND non_chargeable = 0
      AND start_datetime >= '2024-03-01'
      AND start_datetime < '2026-10-01'
  ),
  x AS (
    SELECT
      user_id,
      m,
      min(m) OVER (
        PARTITION BY
          user_id
      ) first_m,
      lagInFrame(m) OVER (
        PARTITION BY
          user_id
        ORDER BY
          m ROWS BETWEEN 1 PRECEDING
          AND CURRENT ROW
      ) prev_m
    FROM
      um
  )
SELECT
  m,
  count() active,
  countIf (m = first_m) new_users,
  countIf (
    m <> first_m
    AND prev_m = addMonths(m, -1)
  ) continuing,
  countIf (
    m <> first_m
    AND prev_m < addMonths(m, -1)
  ) reactivated,
  round(countIf (m = first_m) / count() * 100, 1) new_pct
FROM
  x
WHERE
  m >= '2025-01-01'
GROUP BY
  m
ORDER BY
  m

Implications

The first 30 days decide retention. Most churn happens immediately, after a 1–3 day trial. Things to try: a first-campaign nudge (show results after day 2 and suggest extending), or a minimum first package such as the 7-day fixed price.
Win-back is a real lever. Half of churned advertisers return on their own, typically about 2 months later. A reminder when an advertiser publishes a new listing could pull that return earlier.
Revenue depends on a small core. About 7% of each cohort is still buying a year later and supplies most of the long-term revenue. Building and sale-listing brokers in Riyadh and Jeddah are the segments to protect.
Treat owners and brokers as different products. Owners are one-off buyers whose need ends when their property is sold or rented. Optimise their first purchase (packages, price) rather than retention. Retention and win-back efforts belong with brokers, who keep getting new inventory to promote.
Acquisition is falling. New advertisers per quarter have halved since 2024 while retention stayed flat, so the active base is held up by returning users, not growth.

SOV Advertiser Retention

Executive Summary

Definitions

Retention Curve (All Cohorts Pooled)

SQL Query

Owners vs Brokers

Retention Curve

Usage

By Category

Owner Share of Monthly SOV

SQL Query

Retention by Cohort Quarter

SQL Query

Retention by Category and City

SQL Query

How Advertisers Use SOV

SQL Query

Monthly Active Advertiser Mix

SQL Query

Implications