Portfolio Bidding Across Doctors and Locations: Pooling Conversions for Thin-Data Clinics

This is for anyone running Google Ads for a group of clinics, a multi-branch hospital or several doctors, where each campaign sees only a handful of conversions a month. Smart Bidding needs conversion history to learn. A campaign with five enquiries a month gives it very little. The question is whether pooling campaigns under one portfolio bid strategy helps, and what you lose if you do.

The method below is a decision framework. It does not assert a gain, because the effect depends on how similar your locations are. It assumes you know target CPA bidding and the basics of campaign structure. For the wider context of paid search for hospitals, our healthcare PPC guide is the broader reference, and the cost data by specialty explains why click prices differ across the groups you may want to pool.

What a portfolio strategy is

Google's help page on portfolio bid strategies describes them as AI-powered, goal-driven bid strategies that optimise bids across multiple campaigns, with one place to change bidding settings for every campaign using the same strategy. Pairing a portfolio with a shared budget is described as best practice. You create them in the Shared library, during campaign setup or in campaign settings, or by changing the bid strategy of existing campaigns.

Two facts from that page constrain the design. A portfolio strategy is not available for Performance Max campaigns. And you cannot change an existing portfolio strategy's type, for example from target CPA to target ROAS; you would need a new strategy of the desired type. Plan the goal before you build.

Google does not spell out on that page exactly how data is pooled, so treat claims about pooled learning as reasonable expectation, not guaranteed behaviour. Test it.

Why thin data is a problem

An automated bid strategy estimates the likelihood of a conversion for each auction, using signals from past conversions. With very few conversions, the estimates are noisy. Bids swing, budgets either fail to spend or spend on poor clicks, and managers tend to react by changing targets, which resets the noise. Pooling gives the system a larger set of conversions to learn from, on the assumption that the campaigns share enough characteristics for the learning to transfer.

That assumption is the whole decision.

When pooling is likely to help

Pool campaigns when they are similar on the things that drive conversion likelihood.

• Same service. Several dental clinics offering the same consultation, or several branches of a diagnostic chain offering the same test.
• Comparable markets. Similar cities or catchments, with similar competition and click prices.
• Same conversion definition, tracked in the same way, with the same value logic.
• Comparable landing pages and booking flows.
• A common target. The business is content with the same cost per qualified enquiry for each member.

A hypothetical example, clearly illustrative: an eight-branch physiotherapy chain across tier-two cities, each branch producing a few enquiries a month, all running one landing page template and one booking flow. A single shared target CPA across the eight Search campaigns is a reasonable first test.

When pooling hides a problem

Pooling averages. An average can conceal a weak location.

• A weak branch rides on strong ones. The shared target is met overall, while one location produces enquiries at a far higher cost. Without per-campaign cost reporting, nobody notices.
• A strong branch is held back. The strongest location could take more budget at a profit, but the shared target moderates its bids.
• Mixed services. Pooling a high-value surgical line with a low-value consultation line makes the cheap line look good and the system favour it.
• Different phone handling. One branch answers calls quickly and another does not. The bid strategy sees only the conversion event, not the quality of follow-up, and rewards both the same.
• Different click prices. If one city costs far more per click, a shared target CPA can starve it of impressions.

Consider keeping separate campaigns where branches differ on service, price, capacity or follow-up. Capacity matters most: a doctor with no appointment slots for three weeks should not be pooled with a clinic that has them tomorrow.

Guardrails to set before pooling

1. Define the conversion as a qualified enquiry or booking, not a page view. Pooling a noisy conversion makes the noise louder.
2. Choose the target from your own data. Compute the current cost per qualified enquiry across the campaigns you plan to pool, and set the portfolio target near it, not far below. An aggressive target on thin data reduces delivery.
3. Set a bid limit. Google's page notes that a maximum CPC bid limit can help prevent sudden CPC spikes when moving to automated strategies, though it names Maximize Conversions as its example.
4. Use a shared budget deliberately. A shared budget lets money flow to the campaign with demand. It also lets one campaign consume it all. Decide whether that is acceptable.
5. Keep per-campaign reporting. Create a view that shows each campaign's spend, enquiries and booked appointments separately, so a hidden weak location stays visible.
6. Record the baseline for each campaign before the change: spend, conversions, cost per conversion and impression share.

Build the test

Run the change as an experiment where possible. Take half the eligible campaigns into the portfolio and leave the rest as a control, selecting at random within similar groups. If you have too few campaigns for a split, run a before-and-after with a long enough window, and accept that it is weaker evidence.

Allow for the learning period and the conversion delay. Google's help on conversion delay notes that conversions can be reported up to 90 days after a click and recommends report date ranges that end at least 30 days back. For many clinics the typical lag is shorter, but you should check it in your own account with the days-to-conversion segment before deciding the test length.

How to read the result

Compare four things between pooled and control campaigns.

• Delivery: impression share and spend against budget. Did the pool deliver more steadily?
• Volume: qualified enquiries, not raw conversions.
• Cost: cost per qualified enquiry per campaign, not only for the pool.
• Stability: how often daily spend swung wildly, and how often you had to intervene.

A pooled strategy that improves the total but leaves two campaigns clearly worse has told you which locations need separate attention, not that pooling failed. If the pooled campaigns converge, keep them. If they diverge, split the outliers back out.

Alternatives for thin data

Pooling is one answer. Others, sometimes simpler:

• Use a higher-volume micro-conversion as the optimisation target while you accumulate real enquiries, such as a call of a minimum duration or a booking-form start. The risk is optimising for a weak signal, so validate that it correlates with booked appointments.
• Consolidate campaigns. Three clinics in one city may not need three campaigns; one campaign with location-specific ad groups or assets may do.
• Use manual or capped bidding while data is scarce, reviewed on a fixed schedule. Our dental advertising article discusses where small practices waste spend.
• Feed back qualified outcomes through offline conversion import, so the system learns from enquiries that became appointments. That needs a CRM log tied to the click.

A worked example with invented numbers

Take four clinics, each producing a few qualified enquiries a month at a similar cost, and one fifth clinic in a costlier market producing a similar count at a much higher cost per enquiry. These figures are hypothetical and only illustrate the logic. Pool the first four under one target set near their combined current cost. Leave the fifth in its own campaign, with its own target reflecting its market, until it has enough history to join a different pool or stand alone.

If you had pooled all five under the average, the four cheaper clinics would face a target higher than they need, so the system would be willing to pay more per click than necessary, while the fifth would face a target below what its market can deliver and would lose impressions. The pooled number would look acceptable and neither group would be well served. The lesson is to pool by similarity of cost and capacity, not by organisational chart.

Revisit the grouping every quarter. A clinic that grows its conversion count may no longer need the pool, and a new clinic with no history may benefit from joining one. Moving a campaign between strategies resets some learning, so make these moves deliberately and one at a time.

Governance for multi-doctor setups

For campaigns advertising individual doctors, check the wording and targeting against NMC advertising rules in India and any regulator permit requirements in the UAE before you pool budgets. A shared strategy makes it easy for money to flow toward a doctor-specific campaign that has not been reviewed. Keep doctor campaigns on their own budgets until compliance has cleared them.

Where to start

1. List campaigns with fewer than a handful of conversions a month, and group them by service and market similarity.
2. Mark which groups share capacity, follow-up speed and conversion definition.
3. Build one portfolio for the most similar group, with a target near the current cost per qualified enquiry and a bid limit.
4. Keep per-campaign reporting.
5. Review after a full conversion delay window.
6. Split any location that diverges.

The healthcare media buying team at Branding Pioneers can help set the structure.