End-to-end analytics in real estate: tying ad spend to the money in your CRM

End-to-end analytics pays for itself at the point where budget gets allocated. Without it, money is split on the wrong signal: judging channels by cost per inquiry systematically pushes budget toward the shallowest contacts – price checks, price-list downloads, clicks on a "see the price" button. Those leads are several times cheaper, so the channel looks like the winner in any CPL report. Only a minority of them ever show up to a meeting.

The reverse happens just as often. Property classifieds and branded search deliver inquiries above the average cost, but by then the buyer has already picked a building and a floor plan and only needs to confirm the terms. Cost per closed deal on such a channel can be two or three times lower than on the channel that leads your CPL report. Until the chain is built through to the money, none of this is visible, and shutting down the "expensive" channel looks like a well-reasoned decision.

The second effect is that the conversation with sales moves from opinions to facts. When one table shows how many inquiries from each source reached qualification and a meeting, the argument between "marketing sends junk" and "sales cannot work the phone" ends quickly: both sides look at conversion for a specific source over a specific week. I laid out the broader logic of this system in the article on real estate marketing as a system.

The loop has six parts. Skip any one of them and the chain breaks, after which the numbers stop reconciling.

1. UTM discipline

One shared tagging reference for every campaign, every contractor, and every platform, written down in a document rather than kept in a contractor's head. The mandatory minimum: source, medium, campaign, content, term. Campaign naming deserves its own agreement – "summer_project_display_remarketing" still reads six months later, "test3_new" does not. If three agencies work on the project, the client owns the reference; otherwise the same paid search channel arrives in the report as "yandex", "yandex.direct", and "direct" at the same time.

2. Call tracking

In real estate the phone call is still the primary type of inquiry, so the loop does not close without call tracking. Dynamic call tracking swaps the phone number on the website for each visitor and ties the call to a session, source, campaign, and keyword. Static tracking assigns a dedicated number to an entire channel – the billboard by the highway exit, the banner on the facade, the line in a brochure, the listing card on a classifieds site. Both scenarios are covered by services such as Calltouch, CoMagic, and Roistat, all Russian call tracking and analytics platforms; the choice usually comes down to which CRM and which ad platforms have ready-made integrations.

The pool size for dynamic tracking is calculated from concurrent traffic. If there are fewer numbers than visitors on the site during the peak hour, the system starts reusing numbers and attributes calls to the wrong source. That is the most common technical reason reports disagree.

3. Forms and messaging apps

Website forms, chat, and inquiries via WhatsApp and Telegram have to land in the CRM with the same tags as calls. Messaging apps are a separate leak: switching to a messenger takes the person out of the browser, and without a parameter in the link the source is lost. The fix is links carrying a session identifier plus a tag inserted into the first message, or an integration that passes the contact along with its source.

4. The CRM as the single source of truth about the deal

Every deal stage – qualification, meeting, reservation, contract, payment – lives in one system. For real estate that is either a general-purpose CRM (amoCRM or Bitrix24, the two most widely used CRM platforms in Russia) with an industry add-on, or a specialized product such as Profitbase or MACRO, which ship with a unit inventory grid, reservations, deal stages, and a link to available stock out of the box. The requirement is the same for all of them: funnel stages described in words that a marketer and a head of sales understand identically.

5. Offline conversions sent back to the ad platforms

This is the step most often skipped. The fact that a deal closed has to travel back to the advertising systems as an offline conversion. Yandex Metrica, Russia's dominant web analytics platform, accepts this data via API, CRM connectors, or the conversion center inside Yandex Direct, matching a deal to a visit by ClientID, phone number, or email. From there automated bidding trains on qualified inquiries and reservations instead of raw form fills. The difference in training quality shows over time: the algorithm stops hunting for people who happily leave a phone number and starts looking for people similar to those who came to the sales office.

6. The dashboard

One page of numbers that marketing, sales, and the owner all look at. A channel-level view with spend, inquiries, qualification, meetings, reservations, and deals for the period is enough; twenty tabs work worse than one page. A dashboard opened once a quarter does not need to be built; it changes no decision.

The point of the loop is that each channel gets a whole chain of costs rather than a single number. The same CPL with different qualification rates produces cost per deal that differs several times over.

Two more metrics belong here, though they usually live in finance rather than marketing. The first is marketing spend per square meter sold: how many rubles of marketing sit behind each square meter of sold area. It translates the discussion into language the project office already speaks and shows clearly how acquisition cost drifts as a building sells out. The second is cohort payback: take the inquiries that arrived in March and track how much revenue they had produced by July, August, and September.

A home buyer's path usually looks like this: saw a banner on the facade on the way to work, opened a classifieds site a week later, searched for the project name, read reviews, came to the website through retargeting, called two weeks after that, and showed up for a meeting ten days later. Ten or more touchpoints, several devices, and part of the path never digitized at all, because a wife sent her husband a link in a messaging app.

A last-click model hands all the credit in that chain to branded search or retargeting. A first-click model hands it to out-of-home and display. Both numbers are wrong on their own and both are useful together. The approach that works on a long cycle is to view channels in two cuts at once: by first touch (who brought the person into the market and into the project) and by last touch (who drove the phone call). Channels strong in the first cut and weak in the second get funded as demand sources, not judged on cost per inquiry.

The second honest limitation is the share of unattributed deals. In projects with a heavy offline component it is normal for 15–30% of inquiries to have an unknown source: came on a friend's recommendation, saw the construction site from a window, remembered an ad from three years ago. That share should be measured and gradually reduced, but forcing it to zero through guesswork at data entry helps nobody. A dedicated "source unknown" status beats a 100% filled field where half the values were picked at random.

The perfect attribution model is therefore a false target. The loop solves a more modest problem: telling apart the channels after which people buy from the channels after which people leave a phone number. First touch, last touch, and cohorts are enough for that.

1. 01Analytics on top of an undocumented funnel. The service is connected, but the CRM has no deal stages beyond "new inquiry" and "won". There is nothing to calculate cost per meeting from. Describe the funnel with all its intermediate statuses first, connect analytics second.
2. 02Sales reps do not fill in the CRM. The call comes to a personal mobile, the deal is created by hand with no source, the meeting is never logged. Every number after that is an estimate. Persuasion does not fix this: status updates have to be built into the workflow and into the bonus.
3. 03Different source reference lists. Web analytics says "yandex / cpc", call tracking says "Yandex Direct", the CRM says "paid search", the agency report says "PPC". Merging that into one report takes manual work, repeated every month. A single shared reference is set up once and costs less than one month of that reconciliation.
4. 04Reporting for the sake of reporting. The dashboard is built and refreshed, yet no decision follows from it and budgets are still split out of habit. A useful rule: every report has an owner and a recurring meeting where the numbers produce a specific action.
5. 05Judging channels on too short a window. Two weeks after launch a channel shows zero deals and looks like a failure, even though its first reservations could not physically have appeared yet. The evaluation window should be no shorter than the project's average sales cycle.

There is no need to build the whole loop in one pass. It assembles in layers, and each layer pays off on its own.

1. 01Week 1. The funnel on paper. Describe the stages from inquiry to payment together with the head of sales and write down the definitions: what counts as a qualified lead, what counts as a meeting that took place. Without shared definitions there is nothing to measure.
2. 02Week 1–2. Source reference and UTM rules. One document with tagging rules, sent to every contractor. In parallel, audit what is already tagged incorrectly.
3. 03Week 2. Call tracking. Dynamic tracking on the website with a correctly sized number pool, static numbers on offline channels and classifieds listings. Verify the integration with test calls from each source.
4. 04Week 3. CRM and required fields. Configure funnel stages, automatic deal creation from every channel, and a rule that blocks closing a deal without a source and an amount. Run a separate training session for the sales team, without which the previous three steps lose their value.
5. 05Week 3–4. The first dashboard. Spend, inquiries, qualification, meetings, reservations, and deals by channel for the period. Start with whatever the existing data supports, gaps included.
6. 06Week 4. Offline conversions back into advertising. Set up the transfer of qualified inquiries and reservations to the ad platforms and switch automated bidding to train on those events.

The second month goes into reconciliation: dashboard numbers against a CRM export and against the ad platforms. There will always be discrepancies; the question is how large. A gap of up to 5–7% on inquiries is workable, anything above that usually points to a specific hole – the number pool, a lost source in messaging apps, or deals created manually.

Once the loop is assembled, the next question is what to do with the channel picture it produces. I covered budget allocation across performance, classifieds, and reach formats in detail in the article on marketing a residential development, and the principle of making regular decisions from numbers in the piece on systematic marketing.